Blob Blame History Raw
2017-07-02  Jim Meyering  <>

	version 3.1
	* NEWS: Record release date.

2017-07-01  Jim Meyering  <>

	tests: avoid false failures when run in qemu user mode
	* tests/ Derive the program name that grep
	will use in diagnostics, based on a suggestion from Assaf Gordon.
	* tests/in-eq-out-infloop: Similar: accept an arbitrary "command_name: "
	prefix on checked diagnostics, rather than requiring "grep: ".
	* tests/reversed-range-endpoints: Likewise.
	* tests/write-error-msg: Likewise.
	Reported by Bruno Haible in

2017-06-25  Jim Meyering  <>

	gnulib: update to latest
	* gnulib: Update to latest for these portability fixes:
	- stat: port to xlc 12.01
	- xalloc-oversized: port to icc

	doc: fix another typo
	* doc/grep.texi (File and Directory Selection): Fix typo: s/afer/after/

2017-06-24  Jim Meyering  <>

	doc: stop calling --perl-regexp (-P) "highly" experimental
	Use wording that is less likely to make readers think that
	support for -P may be removed.
	* doc/ s/highly experimental/experimental/
	* doc/grep.texi: Likewise.
	Suggested by Evan Sheahan.

2017-06-21  Jim Meyering  <>

	doc: correct typo
	* doc/grep.texi (Performance): s/suprisingly/surprisingly/

	gnulib: update to latest

2017-06-21  Paul Eggert  <>

	grep: -m no longer cuts off trailing context
	Problem reported by Markus Jochim (Bug#26254).
	* NEWS, doc/grep.texi (General Output Control): Document this.
	* src/grep.c (prpending): Selected lines no longer cut off context.
	(usage): Say "selected" instead of "matching", where appropriate.
	* tests/foad1, tests/max-count-vs-context, tests/yesno:
	Adjust to match new behavior.

2017-05-31  Paul Eggert  <>

	Document grep performance
	* doc/grep.texi (Performance): New section.

	build: update gnulib submodule to latest

2017-05-21  Jim Meyering  <>

	maint: make the announcement template Cc the devel- list
	* (announcement_Cc_): Define.

	gnulib: update to latest; and update tests/

	maint: accommodate GCC7's -Werror=duplicated-branches
	* src/system.h (IGNORE_DUPLICATE_BRANCH_WARNING): Define.
	* src/grep.c (grepfile): Use it.
	* src/kwset.c (bmexec, acexec): Use it.

	maint: update to work with GCC7's -Werror=implicit-fallthrough=
	* src/system.h (FALLTHROUGH): Define.
	* src/grep.c (context_length_arg): Use new FALLTHROUGH macro in place
	of comments
	(fgrep_to_grep_pattern, try_fgrep_pattern, main): Likewise.

2017-05-13  Jim Meyering  <>

	gnulib: update to latest and adapt src/kwset.c
	* gnulib: Update to latest.
	* src/kwset.c: Include "verify.h" for use of assume.

2017-03-22  Jim Meyering  <>

	gnulib: update to latest for dfa [0-9] performance improvement
	This pulls in the following change that is very relevant to grep:

	    commit 6afba02d7869d39ed7f61981045ddbdcb2814101
	    Author: Paul Eggert <>
	    dfa: make [0-9] faster in non-C locales

	* gnulib: Update to latest.
	* NEWS (Improvements): Describe the effect on grep.

2017-03-05  Jim Meyering  <>

	build: use $(builddir), not $(srcdir)
	* (PATH): Use $(builddir), so this also takes effect
	in a non-srcdir build.  Also, switch ${PATH} syntax to $(PATH).

2017-03-05  Juan Manuel Guerrero  <>

	build: use $(PATH_SEPARATOR), not ":" to augment PATH
	* (PATH): Use $(PATH_SEPARATOR), for those systems that
	use something other than ":".
	* Remove name, to avoid syntax-check failure due to
	the duplicate, now that there is this commit.

2017-02-17  Jim Meyering  <>

	maint: fix distcheck failure: remove stale dosbuf.c reference
	* src/ (EXTRA_DIST): Do not attempt to distribute
	the recently deleted file, dosbuf.c.

	maint: fix new syntax-check errors
	* po/ Add lib/xbinary-io.c.
	* (FILTER_LONG_LINES): Add TODO to the list of exempt files.

2017-02-16  Paul Eggert  <>

	Fix up recent -U patches
	Inspired by a suggestion by Eric Blake (Bug#25707#17).
	* bootstrap.conf (gnulib_modules): Add xbinary-io,
	and remove binary-io and xfreopen.
	* doc/grep.texi (Other Options):
	Fix typo and reword to be a bit more general.
	* src/grep.c: Include xbinary-io.h instead of xfreopen.h.
	(grepfile): Open with O_BINARY if binary.
	(grepdesc): No need for set_binary_mode now.
	(grep_command_line_arg, main): Set stdin to binary mode if binary.
	(main): Avoid unnecessary test of stdin == NULL.
	Use xsetmode instead of xfreopen.
	* src/system.h: Do not include binary-io.h.

	build: update gnulib submodule to latest

	Simplify -U on MS-Windows by removing guesswork
	Suggested by Eric Blake (Bug#25707#11).
	* NEWS, doc/grep.texi: Document this.
	* src/dosbuf.c: Remove.
	* bootstrap.conf (gnulib_modules): Add xfreopen.
	* src/grep.c: Include xfreopen.h, not dosbuf.c.
	(fillbuf, print_line_head): Do not undossify input.
	(binary): New static var.
	(grepdesc): Apply BINARY to input file.
	(usage): Remove -u help.
	(main): Set BINARY if -U, and apply it to stdout.  Do nothing if -u.
	With -f, apply BINARY to input file.

2017-02-16  Eric Blake  <>

	grep: don't forcefully strip carriage returns
	Commit 5c92a54 made the mistaken assumption that using fopen("rt")
	on platforms where O_TEXT is non-zero makes sense.  However, POSIX
	already requires fopen("r") to open a file in text mode, vs.
	fopen("rb") when binary mode is wanted, and at least on Cygwin,
	where it is possible to control whether a mount point is binary
	or text by default (using just "r"), the use of fopen("rt") actively
	breaks assumptions on a binary mount by silently corrupting any
	carriage returns that are supposed to be preserved.

	* src/grep.c (main): Never use fopen("rt") (Bug#25707).

2017-02-13  Paul Eggert  <>

	Update TODO and doc
	* TODO: Bring up-to-date and fix formatting glitches.
	* doc/, doc/grep.texi: Fix minor glitches.
	The above patches should address the same problems that recent
	Debian doc patches address, albeit in a different way.

2017-02-12  Paul Eggert  <>

	doc: clarify default input (Bug#25651)
	* doc/
	* src/grep.c (usage): Clarify default input when -r.
	* src/grep.c (usage): Do not bother documenting egrep and fgrep;
	the manual is enough.

2017-02-09  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 3.0
	* NEWS: Record release date.

2017-02-08  Paul Eggert  <>

	grep: do not mishandle \. in multiple patterns
	Problem reported by Lars Wendler (Bug#25655).
	* NEWS: Document this.
	* src/grep.c (try_fgrep_pattern): Fix typo that prevented
	keys from being properly updated.
	* tests/foad1: Test for the bug.

2017-02-07  Paul Eggert  <>

	Do not assume PCRE 8.20 or later
	Problem reported by Zube (Bug#25647)
	* NEWS: Document this.
	* src/pcresearch.c (struct
	Declare only if PCRE_STUDY_JIT_COMPILE.

2017-02-06  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.28
	* NEWS: Record release date.

2017-02-02  Jim Meyering  <>

	gnulib: update to latest

2017-02-01  Paul Eggert  <>

	grep: tune to avoid memchr2 sometimes
	Problem noted by Norihiro Tanaka in:
	Although not enough to restore all the previous performance in the
	case he noted, it helps significantly.
	* src/kwset.c (memchr_kwset): Bring back small_heuristic,
	in a somewhat different form.

2017-01-29  Jim Meyering  <>

	gnulib: update to latest

2017-01-23  Paul Eggert  <>

	grep: simplify recent kwset change
	* src/kwset.c (acexec_trans): Simplify.

2017-01-23  Jim Meyering  <>

	tests: really add the new test name
	* tests/ (TESTS): Add fgrep-longest.

2017-01-21  Norihiro Tanaka  <>

	grep -Fo could report a match that is not the longest
	* src/kwset.c (acexec): Fix it.
	* tests/fgrep-longest: New test.
	* tests/ Add the test.
	* NEWS: Mention it.

2017-01-18  Paul Eggert  <>

	grep: speed up Aho-Corasick when at most 2 bytes
	When using Aho-Corasick and all matched strings either begin with
	the same byte, or begin with one of at most two bytes, use memchr2
	to search for these matching bytes and apply the Aho-Corasick
	algorithm only when a memchr2 match is found.  On my platform,
	this speeds up 'grep -F -e aa -e ba in' by a factor of 7, where
	the file 'in' was created by 'seq -f %040.0f 10000000 >in'.
	* src/kwset.c (struct kwset.gc1): Now int, not char.
	If negative, there is no single terminal byte.  All uses changed.
	(struct kwset.gc1help): Now int, not char.
	If negative, memchr2 cannot be used.
	(kwsprep): Set up gc1 and gc1help from kwset->next, with
	the new (slightly changed) interpretation.
	(memchr_kwset): Use memchr2 if possible.
	Adjust to match new meaning of gc1, gc1help.
	(memoff2_kwset): Remove; no longer needed.
	(acexec_trans): Use memchr_kwset when possible, for speed.
	It now supersedes memoff2_kwset.

	grep: remove Commentz-Walter code
	This code was not being used, and complicated maintenance.
	We can bring it back from the repository if it turns out
	to be useful later.
	* src/kwset.c (struct kwset.reverse): Remove.  All uses of
	FOO->reverse replaced by (FOO->kwsexec == bmexec).
	(kwsalloc): Remove 'reverse' arg, as callers outside this
	module do not care about algorithm choice.  All callers changed.
	(kwsprep): When deciding whether to use Boyer-Moore, do not worry
	about being called twice on the same kwset, as that is not allowed.
	(cwexec): Remove; it was never called.  All uses removed.

2017-01-17  Jim Meyering  <>

	maint: avoid new syntax-check failures
	* src/kwset.c (struct kwset): Split a line longer than 80.
	* bootstrap: Update from gnulib. This fixes a new syntax-check
	failure due to its use of "time stamp".

2017-01-17  Paul Eggert  <>

	* NEWS: Fix typo.

	* src/kwset.c: Fix comment typo.

	Improve -i performance in typical UTF-8 searches
	Currently ‘grep -i i’ is slow in a UTF-8 locale, because ‘i’ in
	the pattern matches the two-byte character 'ı' (U+0131, LATIN
	SMALL LETTER DOTLESS I) in data, and kwset handles only
	single-byte character translations, so grep falls back on a slower
	DFA-based search for all searches.  Improve -i performance in the
	typical case by using kwset when data are free of troublesome
	characters like 'ı', falling back on the DFA only when data
	contain troublesome characters.
	* src/dfasearch.c (GEAcompile):
	* src/grep.c (compile_fp_t):
	* src/kwsearch.c (Fcompile):
	* src/pcresearch.c (Pcompile):
	Pattern arg is now char *, not char const *, since Fcompile
	now reallocates it sometimes.
	* src/grep.c (all_single_byte_after_folding): Remove.
	All callers removed.
	(fgrep_icase_charlen): New function.
	(fgrep_icase_available, try_fgrep_pattern):
	Use it, for more-generous semantics.
	(fgrep_to_grep_pattern): Now extern.
	(main): Do not free keys, since Fexecute may use them.
	* src/kwsearch.c (struct kwsearch): New struct.
	(Fcompile): Return it.  If -i, be more generous about patterns.
	(Fexecute): Use it.  Fall back on DFA when the data contain
	troublesome characters; this should be rare in practice.
	* src/kwset.c, src/kwset.h (kwswords): New function.

	build: update gnulib submodule to latest

2017-01-15  Paul Eggert  <>

	dfa: prefer ptrdiff_t to size_t
	The code already cannot handle objects with size greater than
	SIZE_MAX / 2, so be more honest about it and use ptrdiff_t instead
	of size_t.  ptrdiff_t arithmetic is signed, which allows for more
	checking via -fsanitize=undefined.  It also makes the code a tad
	smaller on x86-64, since it can test for < 0 rather than for ==
	* src/dfasearch.c (struct dfa_comp.kwset_exact_matches):
	(kwsmusts, EGexecute):
	* src/kwsearch.c (Fcompile, Fexecute):
	* src/kwset.c (struct kwset.kwsexec, kwsincr, memchr_kwset)
	(memoff2_kwset, bmexec_trans, bmexec, cwexec, acexec_trans)
	(acexec, kwsexec):
	* src/kwset.h (struct kwsmatch.index, .offset, .size):
	Prefer ptrdiff_t to size_t where either will do.

2017-01-11  Paul Eggert  <>

	grep: improve comments, mostly in kwset
	Remove kwset.h comments that are obsolete and seemingly not
	maintained anyway; people can look in kwset.c instead.
	Update comments to reflect current behavior better.
	Cite Faro & Lecroq 2013.  Use GNU style for end-of-sentence.

2017-01-01  Jim Meyering  <>

	maint: update gnulib and copyright dates for 2017
	* gnulib: Update to latest.
	* all files: Run "make update-copyright".

2016-12-31  Paul Eggert  <>

	grep: speed up -x with many patterns
	* src/kwsearch.c (Fcompile): Improve buffer allocation overhead
	with -x and multiple patterns.  In the common case where '\n' is
	the end-of-line byte, avoid copying other than the first and last

2016-12-31  Jim Meyering  <>

	gnulib: update to latest, fixing a parallel getopt test failure

2016-12-29  Paul Eggert  <>

	maint: space before paren

	grep: int cleanup in kwset.c
	This should affect only theoretical bugs with very large inputs.
	On my platform, this patch shrinks the grep text by 136 bytes.
	* src/kwset.c: Include intprops.h, for INT_MULTIPLY_WRAPV.
	(struct trie, struct kwset, kwsalloc, kwsincr, treedelta, kwsprep)
	(bm_delta2_search, bmexec_trans, cwexec): Prefer ptrdiff_t to int
	when counts can exceed INT_MAX in large inputs, at least in theory.
	(hasevery): Use bool for booleans.
	(bmexec_trans): Avoid undefined behavior on integer overflow.

2016-12-27  Norihiro Tanaka  <>

	grep: improve performance with multiple patterns
	* src/grep.c (main): Avoid fgrep-to-grep conversion for word matching
	with multiple patterns in single byte locales.

2016-12-27  Paul Eggert  <>

	* NEWS: Fix typo.

	grep: fix bug with '... | grep pat >> /dev/null'
	Problem reported by Benno Fünfstück (Bug#25283).
	* NEWS: Document this.
	* src/grep.c (drain_input) [SPLICE_F_MOVE]:
	Don't assume /dev/null is always acceptable output to splice.
	* tests/grep-dev-null-out: Test for the bug.

2016-12-26  Paul Eggert  <>

	grep: minor performance tweak for pure functions
	* src/search.h (wordchars_size, wordchar_next, wordchar_prev):
	Declare to be pure.

2016-12-25  Zev Weiss  <>

	grep: move localeinfo to grep.c
	It's not really dfasearch-specific, and grep.c initializes it, so it
	seems like the most appropriate "owner".

	* src/dfasearch.c (localeinfo): Remove.
	* src/grep.c (localeinfo): Add.
	* src/search.h (localeinfo): Move to new commented section.

2016-12-25  Zev Weiss  <>

	pcresearch: thread safety
	* src/pcresearch.c (pcre_comp): New struct to hold previously-global
	(jit_exec): Operate on a pcre_comp parameter instead of global state.
	(Pcompile): Allocate and return a pcre_comp instead of setting global
	(Pexecute): Operate on a pcre_comp parameter instead of global state.

	kwsearch: thread safety
	* src/kwsearch.c (Fcompile): Return a kwset_t instead of setting a
	global variable.
	(Fexecute): Use a passed-in kwset_t instead of a global variable.
	(kwset): Remove global variable.

	dfasearch: thread safety
	* src/dfasearch.c (struct dfa_comp): New struct to hold
	previously-global variables.
	(dfawarn): Remove static variable.
	(kwsmusts): Operate on a dfa_comp parameter instead of global
	(GEAcompile): Allocate and return a dfa_comp struct instead of setting
	global variables.
	(EGexecute): Operate on a dfa_comp parameter instead of global
	* src/searchutils.c (kwsinit): Replace a static array with a
	dynamically-allocated one.

2016-12-25  Zev Weiss  <>

	grep: prepare search backends for thread-safety
	To facilitate removing mutable global state from search backends,
	compile() functions will return an opaque pointer to backend-specific
	data, which must then be passed back into the corresponding execute()
	function.  This is merely a preparatory step changing function
	signatures and call sites, so the pointers passed & returned are
	dummies for now and not (yet) actually used.

	* src/grep.c (compile_fp_t): Now returns an opaque pointer (the
	compiled pattern).
	(execute_fp_t): Now passed the pointer returned by a compile_fp_t.
	All call sites updated accordingly.
	(compiled_pattern): New static variable.
	* src/dfasearch.c (GEAcompile): Return a void pointer (dummy NULL).
	(EGexecute): Receive a void pointer argument (unused).
	* src/kwsearch.c (Fcompile): Return a void pointer (dummy NULL).
	(Fexecute): Receive a void pointer argument (unused).
	* src/pcresearch.c (Pcompile): Return a void pointer (dummy NULL).
	(Pexecute): Receive a void pointer argument (unused).
	* src/search.h: Update compile/execute function prototypes.

2016-12-24  Jim Meyering  <>

	maint: fix "syntax-check" failure
	* src/grep.c (SEP_STR_GROUP): Declare "static".

2016-12-23  Paul Eggert  <>

	grep: fix comment in searchutils.c

	grep: improve word checking with UTF-8
	* src/searchutils.c: Do not include <verify.h>.
	(word_start): Remove, replacing with ...
	(sbwordchar): New static var.  All uses changed.
	(wordchar_prev): Return size_t, not bool, as this generates
	slightly better code.  Go back faster if UTF-8.

	grep: standardize on localeinfo.multibyte
	* src/dfasearch.c (EGexecute):
	* src/grep.c (main):
	* src/kwsearch.c (Fexecute):
	* src/pcresearch.c (Pcompile):
	Prefer localeinfo.multibyte to (MB_CUR_MAX > 1).

	grep: speed up -wf in C locale
	Problem reported by Norihiro Tanaka (Bug#22357#100).
	This patch improves the performance on that benchmark on my
	platform so that grep is now only about 2x slower than grep 2.26,
	which means it is considerably faster than grep 2.25 and earlier.
	* src/kwsearch.c (Fexecute):
	Use wordchars_size to boost performance for this case.
	* src/search.h, src/searchutils.c (wordchars_size): New function.

	grep: specialize word-finding functions
	This improves performance a bit.
	* src/dfasearch.c, src/kwsearch.c (wordchar):
	Remove; now in searchutils.c.
	* src/grep.c (main): Call wordinit if -w.
	* src/search.h: Adjust.
	* src/searchutils.c: Include verify.h.
	(word_start): New static var.
	(wordchar): Move here from dfasearch.c and kwsearch.c.
	(wordinit, wordchars_count, wordchar_next, wordchar_prev):
	New functions.
	(mb_prev_wc, mb_next_wc): Remove.
	All callers changed to use the new functions instead.

	grep: simplify Fexecute
	* src/kwsearch.c (Fexecute): Avoid the need for a 'try' local or
	for a 'goto success'.  Update mb_start to reflect newline found.

	grep: remove C label
	* src/kwsearch.c (Fexecute): Remove label.

	maint: rewrite to avoid some macros
	These days, the dangerous powers of C macros are not needed if
	constants or functions will do just as well.
	* src/kwset.c (DEPTH_SIZE):
	Now constants, not macros.
	* src/kwset.c (link): Remove macro. Instead, rename local vars
	from 'link' to 'cur'.
	(malloc) [GREP]: Remove macro.  All uses of malloc changed to xmalloc.
	Omit double-inclusion of xalloc.h.  Do not depend on 'GREP'.
	(U): Now a function, not a macro.
	* src/kwset.c, src/searchutils.c (NCHAR): Move this macro to ...
	* src/system.h: ... here, and make it a constant.

2016-12-20  Paul Eggert  <>

	grep: fix performance with multiple patterns
	Problem reported by Jaroslav Skarvada (Bug#22357).
	* NEWS: Document this and other recent performance fixes.
	* src/grep.c (E_MATCHER_INDEX): New constant.
	New function, split out from fgrep_icase_available.
	(fgrep_icase_available): Use it.
	(try_fgrep_pattern): New function, which also uses it.
	(main): With two or more patterns, use try_fgrep_pattern to fix
	performance regression.  The number "two" here is just a heuristic.

	grep: simplify matcher configuration
	* src/grep.c (matcher, compile): Remove static vars.
	(compile_fp_t): Now takes a 3rd syntax argument.
	(Gcomppile, Ecompile, Acompile, GAcompile, PAcompile): Remove.
	(struct matcher): Now nameless, since it is used only once.
	Make 'name' a bit shorter.  New member 'syntax'.
	(matchers): Initialize it, and change removed functions to GEAcompile.
	(setmatcher): New arg MATCHER, and return new matcher index.
	Avoid unnecessary call to strcmp.
	(main): Keep matcher as a local int, not a global pointer.
	* src/kwsearch.c (Fcompile):
	* src/pcresearch.c (Pcompile): Ignore the 3rd syntax argument.

	grep: simplify line counting in patterns
	* src/grep.c (n_patterns): Rename from patfile_lineno,
	as it is now origin-zero.  Now size_t, not uintmax_t.
	(count_nl_bytes, fl_add): Simplify to just buffer and size.
	All callers changed.

2016-12-19  Paul Eggert  <>

	build: update gnulib submodule to latest

2016-12-18  Paul Eggert  <>

	build: update gnulib submodule to latest

	build: update gnulib submodule to latest

2016-12-13  Jim Meyering  <>

	tests: use just-built grep in more places
	* (PATH): Prepend $(srcdir)/src, so that we use the just-
	built grep also when running commands like those of "make distcheck".
	This would have avoided the recently-luckily-noticed infloop bug.
	Tested by running this in a just-built directory:
	f=src/grep; printf '%s\n' '#!/bin/sh' 'sleep 9h' > $f; chmod a+x $f
	and then verifying that nearly every "make syntax-check" rule hangs.

	maint: tell "syntax-check" not to worry about the NEWS update
	Whenever we change "old" NEWS, we have to update this checksum.
	Otherwise, a "make syntax-check" test that guards against a class
	of logical merge conflicts will fail.
	* (old_NEWS_hash): Update this hash to accommodate the
	recent clarification of a 2.27 NEWS entry.

2016-12-13  Arnold D. Robbins  <>

	build: update gnulib submodule to latest
	* src/dfasearch.c (GEAcompile): Remove use of flag, RE_ICASE covers it.

2016-12-12  Paul Eggert  <>

	grep: work around proc lseek glitch
	Problem reported by Andreas Schwab (Bug#25180).
	* NEWS: Document this.
	* src/grep.c (finalize_input): Ignore EINVAL lseek failures.
	* tests/ (TESTS): Add proc.
	* tests/proc: New file.

2016-12-07  Paul Eggert  <>

	grep: simplify finalize_input
	* src/grep.c (finalize_input): Simplify without changing behavior.
	It's still a bit of a rat's-nest, but it's a cozier rat's-nest.

	maint: clarify early-exit news for 2.27
	* NEWS: Mention early-exit options to avoid confusion.  See:

2016-12-06  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.27
	* NEWS: Record release date.

2016-11-29  Jim Meyering  <>

	grep: fix DFA-induced infloop
	* gnulib: Update to latest, for the DFA infloop fix.
	* tests/dfa-infloop: New test, to trigger an infinite loop
	in the DFA matcher.
	* tests/ (TESTS): Add it.

2016-11-28  Jim Meyering  <>

	tests: use "returns_ N env VAR=val ..."
	rather than "VAR=val returns_ N ..."
	Some shells do not propagate envvar settings through our use
	of the "returns_" function, so set any envvar via use of "env".
	This was an issue at least on Ubuntu and Debian-based systems,
	presumably due to their common use of "dash" as /bin/sh.
	Reported by Assaf Gordon.
	* tests/char-class-multibyte: As above.
	* tests/euc-mb: Likewise.
	* tests/false-match-mb-non-utf8: Likewise.
	* tests/pcre-infloop: Likewise.
	* tests/pcre-jitstack: Likewise.
	* tests/sjis-mb: Likewise.
	* tests/warn-char-classes: Likewise.

2016-11-28  Paul Eggert  <>

	tests: revert check for unibyte French range bug
	The test wasn't portable, as it assumed that rational ranges
	were not in effect.  Problem reported by Eric Blake (Bug#25048#8).
	There doesn't seem to be a portable way to do the test, so omit it.
	* tests/init.cfg, tests/unibyte-bracket-expr:
	Revert previous change.

	build: update gnulib submodule to latest

2016-11-27  Jim Meyering  <>

	grep: avoid false matches in non-UTF8 multibyte locales
	* gnulib: Update to latest, for the dfa.c fix.
	* NEWS (Bug fixes): Mention it.
	* tests/false-match-mb-non-utf8: New file, with tests for this.
	Based on tests from Stephane Chazelas.
	* tests/ (TESTS): Add it.
	Introduced by commit v2.18-54-g3ef4c8e, a change that made grep use
	its DFA matcher more aggressively.  The malfunction arises only with
	the DFA matcher, not with regex.
	Reported by Stephane Chazelas in

2016-11-20  Paul Eggert  <>

	tests: check for unibyte French range bug
	Problem reported by Stephane Chazelas (Bug#24973).
	This bug was fixed in Gnulib.
	* NEWS: Document the fix.
	* tests/init.cfg (require_ru_RU_koi8_r): Remove.
	* tests/unibyte-bracket-expr: Add a test for the bug.
	Call get-mb-cur-max directly instead of bothering with

	build: update gnulib submodule to latest

2016-11-19  Paul Eggert  <>

	grep: further -P performance fix
	Problem reported by Stephane Chazelas in:
	* src/pcresearch.c (Pexecute): Set the subject to the start of
	each line as it is found.

	grep: -P no longer uses PCRE_MULTILINE
	This reverts commit f6603c4e1e04dbb87a7232c4b44acc6afdf65fef,
	as the extra performance is not worth the trouble for PCRE users.
	Problem reported by Stephane Chazelas in:
	* NEWS: Document this and the next patch.
	* src/dfasearch.c (EGexecute):
	* src/grep.c (execute_fp_t):
	* src/kwsearch.c (Fexecute):
	* src/pcresearch.c (Pexecute):
	First arg is now a const pointer again.
	* src/grep.c (buf_has_encoding_errors): Now static.
	* src/grep.h (buf_has_encoding_errors): Remove decl.
	* src/search.h: Adjust decls.
	* src/pcresearch.c (reflags): Remove.  All uses removed.
	(Pcompile, Pexecute): Do not use PCRE_MULTILINE.

2016-11-19  Jim Meyering  <>

	doc: fix a doubled "the"
	* doc/grep.texi (--perl-regexp): s/the\nthe/the/

2016-11-19  Paul Eggert  <>

	grep: fix -zxP bug
	* NEWS: Document this.
	* src/pcresearch.c (Pcompile): Search a line at a time if -x is
	used, since -x uses ^ and $.
	* tests/pcre: Test this.

	grep: simplify by using PRIuMAX
	* (HAVE_PRINTF_C99_SIZES): Remove; no longer needed.
	* src/grep.c (print_offset): Simplify (Bug#24451).

	grep: -T now adjusts number widths for worst case
	* NEWS, doc/grep.texi (Output Line Prefix Control):
	Document this (Bug#24451).
	* src/grep.c (offset_width): New static var.
	(print_offset): Use it instead of arg.  All callers changed.
	(grep): Set it.
	* tests/initial-tab: Test this.

	grep: -T no longer outputs BS
	* NEWS: Document this (Bug#24451).
	* src/grep.c (print_line_head): Do not attempt to backspace output.
	* tests/initial-tab: New test.
	* tests/ (TESTS): Add it.

	grep: document -oz better
	* doc/grep.texi (General Output Control, Usage): Tweak (Bug#24961).

	grep: fix performance typo with -P
	Reported by Zev Weiss in:
	* src/pcresearch.c (Pcompile): Initialize reflags.

	tests: use "returns_" rather than "$?"
	* tests/grep-dev-null-out: Use "returns_ 124" rather than testing
	$? = 124.

	grep -f /dev/null -L PAT FILE outputs FILE
	* NEWS: Document this.
	* src/grep.c (main): Do not exit right away with -L.
	* tests/skip-read: Test for the fix.

	grep: tune -f /dev/null
	* src/grep.c (main): Do the -f /dev/null early-exit checks before
	more-expensive tests that involve syscalls.

	grep: treat -f /dev/null like -m0
	* NEWS: Document this.
	* src/grep.c (main): With -f /dev/null, don't bother to read the
	input.  This is what FreeBSD grep does.
	* tests/ (TESTS): Add skip-read.
	* tests/skip-read: New file.

	grep: avoid O(N**2) buffer reallocation
	* src/grep.c (main): Use x2realloc to avoid O(N**2) performance as
	pattern buffers grow.

	grep: avoid unnecessary gettext call
	Translate "(standard input)" lazily.
	* src/grep.c (input_filename): New function.
	(suppressible_error): Remove 1st arg, since it is always
	input_filename ().  All callers changed.
	(suppressible_error, print_filename, grep, grepdesc): Use it.
	(grep_command_line_arg): Set filename to NULL if standard
	input has no label.  Often, this avoids all calls to gettext,
	which can be a win as the first call can be expensive.

	grep: drain the input pipe faster
	* src/grep.c (dev_null_output): Now static.
	(drain_input): New function, using 'splice' if that makes sense.
	(finalize_input): Use it.
	(main): Omit now-unnecessary initialization.

	grep: scale back /dev/null speedup
	The performance improvement when output is /dev/null (commit
	af6af288eac28951b5eee1eaaf373e22b2193b7b dated 2016-05-01)
	breaks scripts that run "PROGRAM | grep PATTERN >/dev/null"
	where PROGRAM dies when writing into a broken pipe.
	Suppress the improvement if standard input is not seekable.
	Problem reported by Gary Johnson (Bug#24941).
	* NEWS: Document this.
	* src/grep.c (seek_failed): New static var.
	(seek_data_failed): Move decl earlier, to be next to seek_failed.
	(file_must_have_nulls): Skip useless syscalls if seek_failed.
	Lessen source-code nesting.
	(reset): Set seek_failed and seek_data_failed.
	Try lseek even on non-regular files.
	(grep): New arg INEOF.  All callers changed.
	Do not clear seek_data_failed here, since 'reset' now does this.
	(finalize_input): New static function.
	(grepdesc): Use it.
	(main): Do not exit on first match merely because output is
	* tests/grep-dev-null-out: Adjust to new behavior.

	grep: improve diagnostic on lseek failure
	* src/grep.c (reset): Mention the file name in the (unlikely)
	chance of an lseek failure.

	grep: avoid unnecessary isatty calls
	This fixes an inefficiency that was mistakenly introduced a while
	back, when the macro SET_BINARY became defined on all platforms.
	* src/grep.c (grepdesc, main): Do not unecessarily call isatty on
	POSIXish platforms.

	grep: -Pz no longer rejects ^, $
	Problem reported by Stephane Chazelas (Bug#22655).
	* NEWS: Document this.
	* doc/grep.texi (grep Programs): Warn about -Pz.
	* src/pcresearch.c (reflags): New static var.
	(multibyte_locale): Remove static var; now local to Pcompile.
	(Pcompile): Check for (? and (* too.  Set reflags instead of
	dying when problematic operators are found.
	(Pexecute): Use reflags to decide whether searches should
	be multiline.
	* tests/pcre: Test new behavior.

2016-11-14  Jim Meyering  <>

	tests: use "returns_" rather than explicit comparison with "$?"
	* tests/sjis-mb (encode): Rearrange to emit desired input into
	a file, rather than piping directly into grep.  That permits
	the use of returns_ 1 to verify timeout's exit status.
	* tests/euc-mb: Use "returns_ 1" rather than testing $? = 1
	* tests/char-class-multibyte: Likewise.
	* tests/dfa-heap-overrun: Likewise.
	* tests/encoding-error: Likewise.
	* tests/fedora: Likewise.
	* tests/grep-dev-null: Likewise.
	* tests/init.cfg (envvar_check_fail): Likewise.
	* tests/kwset-abuse: Likewise.
	* tests/mb-non-UTF8-overrun: Likewise.
	* tests/multibyte-white-space: Likewise.
	* tests/pcre-infloop: Likewise.
	* tests/surrogate-pair: Likewise.
	* tests/warn-char-classes: Likewise.
	Do the same for other values:
	* tests/backref-multibyte-slow: Likewise.
	* tests/euc-mb: Likewise.
	* tests/pcre-abort: Likewise.
	* tests/pcre-jitstack: Likewise.
	* tests/repetition-overflow: Likewise.
	* tests/reversed-range-endpoints: Likewise.
	* tests/warn-char-classes: Likewise.

2016-10-26  Jim Meyering  <>

	doc: grep builds on HP-UX once again
	* NEWS (Bug fixes): Mention the HP-UX fix.

	gnulib: update to latest, for getprogname HPUX port

2016-10-22  Mark Veltzer  <>

	ignore coverage generated files

	ignore ar-lib in build-aux

2016-10-20  Zev Weiss  <>

	grep: use 'j' intmax_t printf length modifier if supported
	* Use gl_PRINTF_SIZES_C99 to test printf and
	(conditionally) define HAVE_PRINTF_C99_SIZES.
	* src/grep.c (print_offset): Use printf("%j...") for printing
	[u]intmax_t if HAVE_PRINTF_C99_SIZES is defined; otherwise continue
	using the existing hand-rolled loop.

2016-10-15  Jim Meyering  <>

	build: distribute new file, die.h, so "make distcheck" passes
	* src/ (grep_SOURCES): Add die.h.
	Also, sort these file names.

2016-10-10  Paul Eggert  <>

	build: update gnulib submodule to latest

2016-10-09  Jim Meyering  <>

	maint: die.h: add the "#define ..." part of double inclusion guard
	* src/die.h (DIE_H): Define to 1.

2016-10-04  Paul Eggert  <>

	grep: don't assume stdbool.h before die call
	* src/die.h: Include stdbool.h, since 'die' uses 'false'

	grep: die more systematically
	* src/die.h: New file.
	* src/dfasearch.c, src/grep.c, src/pcresearch.c: Include die.h.
	* src/dfasearch.c (dfaerror):
	* src/grep.c (context_length_arg, add_count, prline, setmatcher, main):
	* src/pcresearch.c (jit_exec, Pcompile, Pexecute):
	Use 'die' instead of 'error' when exiting.
	* src/pcresearch.c: Do not include verify.h.
	(die): Remove; now in die.h.
	* src/search.h: Do not include error.h here, since this file does
	not use anything defined in error.h.  Instead, dfasearch.c, which
	uses error.h's symbols, now includes error.h directly.

2016-10-02  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.26
	* NEWS: Record release date.

2016-10-01  Jim Meyering  <>

	gnulib: update to latest; for getprogname fix

2016-10-01  Paul Eggert  <>

	tests/grep-dir: port to Solaris 10
	* tests/grep-dir: Port to Solaris 10 'cat', which
	exits with status 0 even after 'read' fails from a directory.

2016-09-28  Jim Meyering  <>

	build: placate GCC 7's -Wimplicit-fallthrough
	* src/pcresearch.c (die): New macro.
	(Pexecute): Use it in place of offending uses of error,
	to placate GCC 7's -Wimplicit-fallthrough.
	Include verify.h.  Since this is grep's first explicit use of this
	gnulib module, ...
	* bootstrap.conf (gnulib_modules): Add verify.

	gnulib: update to latest; for ...
	This includes the following:
	 - a getprogname-vs-openbsd-5.1 portability fix
	 - "fallthru" comment-adding changes for dfa and unistr/u8-uctomb-aux.c
	 - another getprograme fix to avoid breaking newer glibc

2016-09-27  Paul Eggert  <>

	build: reword .git old-GCC warning
	* (gl_gcc_warnings): Reword diagnostic.
	Suggested by Assaf Gordon in:

	build: port .git builds to newer GCC
	* (gl_gcc_warnings): Omit duplicate copy of 'main'.
	Problem reported by Assaf Gordon in:

	build: port .git builds to older GCC
	Problem reported by Dagobert Michelsen in:
	* (gl_gcc_warnings): Default to false if .git
	exists but GCC is too old.

2016-09-27  Jim Meyering  <>

	tests/long-pattern-perf: avoid false-failure due to cache speed
	* tests/long-pattern-perf: This test would fail semi-consistently
	on some systems, probably because the smaller regexp fit well
	within cache, yet the larger one did not.  In that case, there
	was a relative speed difference greater than 20x and the test
	would fail.  Quadruple the sizes, to make that less likely.
	Also, construct the 10x larger regexp directly from the smaller,
	rather than relying on seq with endpoints to induce that
	approximate size ratio.  Reported by Bruce Dubbs in

2016-09-24  Jim Meyering  <>

	build: avoid "./configure && make dist" missing-dep. failure
	* (run-syntax-check): Depend on "all", to avoid a
	parallel build failure due to a missing dependency. Reported by
	Paul Eggert in

2016-09-24  Paul Eggert  <>

	build: update gnulib submodule to latest

2016-09-24  Jim Meyering  <>

	tests/fmbtest: avoid false-failure due to reliance on MB-correct sed
	* tests/fmbtest: Several of these tests would mistakenly fail due to
	postprocessing with a combination of sed and locale support that failed
	to handle some multibyte characters in the cs_CZ.UTF-8 locale.  Instead
	of relying on sed's multibyte support or anything locale-related to
	perform this simple filtering, just use this: tr -cs '0-9' '[ *]'
	Also, rather than exporting LC_ALL, just set it for each command.
	Reported by Nelson H. F. Beebe.

	tests: revamp multibyte-white-space test to be more permissive
	This test elicits too many failures.  Whether a system has accurate
	unicode "whitespace" attributes should not influence whether grep's
	test suite passes.  In many cases, now you will see a warning that
	some multibyte characters do not pass whitespace-related tests, but
	this test no longer fails.  However, if you run this test on a modern
	enough system, it does require that \s and \S do work properly with
	most of the listed characters.
	* tests/multibyte-white-space: Confirm that Fedora 24's locale
	tables still declare those four Unicode code points *not* whitespace.
	Honor a new column telling how to handle failure.  Provide more
	information in each diagnostic.
	Reported by Nelson H. F. Beebe.

	tests: avoid erroneous failure of pcre-jitstack test
	On some systems (*BSD), 'ulimit -s unlimited' would fail, yet the
	test for that mistakenly masked the failure, so the following grep
	command ended up failing with a segfault.
	* tests/pcre-jitstack: Don't mask the ulimit failure.
	Reported privately by Nelson H. F. Beebe.

2016-09-23  Jim Meyering  <>

	grep: avoid unwarranted "input file 'F' is also the output" on *BSD
	On *BSD systems, any command like "echo y | grep x", where grep reads
	from a pipe and writes to standard output, would mistakenly emit this:
	  grep: input file '(standard input)' is also the output
	* src/grep.c (grepdesc): Ensure that the file descriptor we're
	reading is a regular one before using SAME_INODE to test whether
	it is the same as the descriptor open on standard output.
	Nelson Beebe reported privately that the foad1 tests failed on many
	BSD systems.  Exposed by commit v2.25-2-gaf6af28.

	tests: avoid backref-multibyte-slow false failure
	* tests/backref-multibyte-slow (max_seconds): If we calculate
	a max duration of 1 second, use 5.  Otherwise, on high-latency
	systems, it would be way too easy for the duration of the final
	test run to exceed that limit.  Reported by Nelson H. F. Beebe.

2016-09-22  Jim Meyering  <>

	gnulib: update to latest; for getprogname-vs-AIX fix

2016-09-18  Norihiro Tanaka  <>

	grep: add news entry for fix to bug#24233
	* NEWS (Bug fixes): Add an entry describing bug#24233.
	The bug was fixed by commit v2.25-77-gad468bb, by chance.

2016-09-15  Jim Meyering  <>

	gnulib: update to latest

2016-09-10  Jim Meyering  <>

	dfa: reflect move of grep's DFA code to gnulib
	Now that the core DFA code and tests reside in gnulib,
	remove the copies here and use what gnulib provides.
	* bootstrap.conf: Use the dfa module.
	* Remove settings involving files that have moved.
	(_gl_TS_unmarked_extern_functions): Add dfaerror and dfawarn.
	It is wrong/ugly to have to define these global symbols to use
	the dfa module, but we'll adjust that separately.
	* po/ Apply s/src/lib/ to src/dfa.c.
	* src/ Remove mention of dfa.[ch] and localeinfo.[ch].
	* tests/ Remove mention of the tests that we have
	moved to the gnulib module.
	* src/dfa.c: Remove file.
	* src/dfa.h: Likewise.
	* src/localeinfo.c: Likewise.
	* src/localeinfo.h: Likewise.
	* tests/dfa-match: Likewise.
	* tests/dfa-match-aux.c: Likewise.
	* tests/invalid-char-class: Likewise.

	gnulib: update to latest, for new dfa module

2016-09-08  Paul Eggert  <>

	grep: encoding errors suppress just their line
	From a suggestion by Marcello Perathoner (Bug#22838).
	* NEWS, doc/grep.texi (File and Directory Selection): Document this.
	* src/grep.c (print_line_head): Do not suppress later output lines
	merely because an earlier output line would have had an encoding error.
	* tests/encoding-error: Test for the new behavior.

2016-09-08  Jim Meyering  <>

	gnulib: update to latest, for getprogname fixes

2016-09-08  Norihiro Tanaka  <>

	dfa: additional change new option for anchored searches
	* src/dfa.c (dfaexec_main): Do it.

2016-09-07  Paul Eggert  <>

	doc: define "context lines"
	Reported by Igor Bogomazov via Santiago Ruano Rincón (Bug#24024).
	* doc/grep.texi (Context Line Control): Define "context lines".

	build: update gnulib submodule to latest

2016-09-05  Jim Meyering  <>

	maint: switch from gnulib's progname to getprogname module
	* gnulib: Update to latest, for its new getprogname module.
	* bootstrap.conf (avoided_gnulib_modules): Include the getprogname
	module rather than the now-obsolescent progname.
	* src/grep.c: Include "getprogname.h" rather than "progname.h"
	and remove any use of set_program_name.
	* tests/dfa-match-aux.c (main): Likewise.
	* tests/get-mb-cur-max.c (main): Likewise.
	* src/grep.c (usage, main): Use getprogname() in place of program_name.

2016-09-02  Paul Eggert  <>

	dfa: minor cleanup of previous change
	* src/dfa.c (dfaexec_main): Omit redundant code and reindent.

2016-09-02  Norihiro Tanaka  <>

	dfa: additional change new option for anchored searches
	* src/dfa.c (dfaexec_main): Do it.

	dfa: use single-byte algorithm even in non-UTF-8
	* src/dfa.c (dfaexec_main): Do it.  (This was inadvertently
	omitted in a recent patch.)

2016-09-02  Paul Eggert  <>

	dfa: merge xalloc.h changes from Gawk
	* src/dfa.h (_GL_ATTRIBUTE_MALLOC): Define here, as other
	Gnulib .h files do.  This is more consistent with Gawk.
	* src/dfa.c: Include xalloc.h, since dfa.h no longer does so.
	Include localeinfo.h later; we don't care about order, but Gawk does.

2016-09-02  Arnold Robbins  <>

	dfa: port to C90
	* src/dfa.c (dfamust): Avoid declarations after statement (Bug#21486).

2016-09-02  Paul Eggert  <>

	dfa: new option for anchored searches
	This follows up on a suggestion by Norihiro Tanaka (Bug#24262).
	* src/dfa.c (struct regex_syntax): New member 'anchor'.
	(char_context): Use it.
	(dfasyntax): Change signature to specify it, along with the old
	FOLD and EOL args, as a single DFAOPTS arg.  All uses changed.
	* src/dfa.h (DFA_ANCHOR, DFA_CASE_FOLD, DFA_EOL_NUL): New constants
	for dfasyntax new last arg.

2016-09-02  Norihiro Tanaka  <>

	dfa: simplify and optimize at initial state in execution
	* src/dfa.c (skip_remains_mb): Remove argument *pwc.  Update calller.
	(dfaexec_main): Simplify and optimize at initial state (Bug#24261).

	dfa: simplify to find state index for state 0
	* src/dfa.c (dfastate): Simplify to find state index for state 0.

2016-09-01  Norihiro Tanaka  <>

	tests: add a new test for SJIS locale
	* tests/sjis-mb: Add a new test.  It fails in grep-2.25 or prior.

2016-09-01  Paul Eggert  <>

	grep: update NEWS
	* NEWS: Describe previous change.

	grep: use regex fastmap unless -i
	This builds on a suggestion by Norihiro Tanaka (Bug#24009).
	* src/dfasearch.c (GEAcompile): Use a fastmap unless -i.
	This improves performance 20x for me using the first benchmark
	given in Bug#24009.

	grep: improve dfasearch storage management
	This patch is mostly refactoring, with a bit of performance tweaking.
	It is done in preparation for a fix for Bug#24009.
	* src/dfasearch.c (patterns): Now of type struct re_pattern_buffer *
	instead of an anonymous struct pointer, since there is no longer
	any need to keep regs here.  All uses changed.
	(GEAcompile): Use patlim instead of a hard-to-follow "total".
	Use x2nrealloc to avoid potential O(N**2) reallocation algorithm.
	Initialize just the pattern members that need clearing.
	(EGexecute): Put regs into a static variable, as this code did
	before 2001-02-18, as there is no need to have a separate set of
	regs for each pattern.  Explain the "Q@#%!#" comment better.

2016-09-01  Norihiro Tanaka  <>

	dfa: remove separation by context in transition in non-UTF8 multibyte locales
	* src/dfa.c (struct dfa): Remove member curr_dependent.  All uses

2016-09-01  Paul Eggert  <>

	dfa: document previous change
	* NEWS: Adjust to match previous change.

2016-09-01  Norihiro Tanaka  <>

	dfa: avoid invalid character matching period
	* dfa.c (transit_state): Avoid invalid character matching period.

	dfa: use single-byte algorithm even in non-UTF-8
	Even in non-UTF8 locales, if the current input character
	is single byte, we can use CSET to match ANYCHAR.
	* src/dfa.c (struct dfa): New member canychar.
	Cache index of CSET for ANYCHAR.
	(lex): Make CSET for ANYCHAR.
	(state_index): Simplify.
	(dfastate): Consider CSET for ANYCHAR.
	(transit_state_singlebyte, transit_state): Remove handling for eolbyte,
	as we assume that eolbyte does not appear at current position.
	(dfaexec_main): Use algorithm for single byte character to any single
	byte character in input text always.
	(dfasyntax): Initialize canychar.

2016-09-01  Paul Eggert  <>

	grep: avoid code duplication with -iF
	This follows up on the -iF performance improvement (Bug#23752).
	* NEWS: Simplify description of -iF improvement.
	* src/dfa.c: Do not include wctype.h.
	(lonesome_lower, case_folded_counterparts): Move to localeinfo.c.
	(CASE_FOLDED_BUFSIZE): Move to localeinfo.h.
	* src/grep.c: Do not include wctype.h.
	(lonesome_lower): Remove.
	(fgrep_icase_available): Use case_folded_counterparts instead.
	Do not call it for the same character twice.
	Return false on wcrtomb failures (which should never happen).
	(fgrep_to_grep_pattern, main): Simplify.  Let fgrep_to_grep’s
	caller fiddle with the global variables.
	* src/localeinfo.c: Include <wctype.h>
	(lonesome_lower, case_folded_counterparts):
	Move here from src/dfa.c.  Return int, not unsigned int.
	Verify that CASE_FOLDED_BUFSIZE is big enough.
	* src/localeinfo.h (CASE_FOLDED_BUFSIZE): Now 32, so that
	we don’t expose lonesome_lower’s size.
	* src/searchutils.c (kwsinit): Return new kwset instead of
	storing it via a pointer.  All callers changed.  Simplify a bit.

2016-09-01  Norihiro Tanaka  <>

	grep: speed up -iF in multibyte locales
	In a multibyte locale, if a pattern is composed of only single byte
	characters and their all counterparts are also single byte characters
	and the pattern does not have invalid sequences, grep -iF uses the
	fgrep matcher, the same as in a single byte locale (Bug#23752).
	* NEWS: Mention it.
	* src/grep.c (lonesome_lower): New constant.
	(fgrep_icase_available): New function.
	(fgrep_to_grep_pattern): Simplify it.
	(main): Use them.
	* src/searchutils.c (kwsinit): New arg MB_TRANS; all uses changed.
	Try fgrep matcher for case insensitive matching by grep -F in multibyte

2016-08-31  Paul Eggert  <>

	build: update gnulib submodule to latest

2016-08-31  Jim Meyering  <>

	maint: avoid new 'make syntax-check' failure
	* src/dfa.c (using_simple_locale): Prefer STREQ(a,b) over
	strcmp(a,b) == 0.

	gnulib: update to latest

2016-08-31  Paul Eggert  <>

	dfa: make dfa.c fully thread-safe
	This follows up on Zev Weiss’s recent patches to make the DFA code
	thread-safe (Bug#24249).  It removes the remaining static
	variables used by dfa.c.  These variables are locale-dependent, so
	they would cause problems in multithreaded code where different
	threads are in different locales (e.g., via uselocale).  I
	abstracted most of the variables into a new localeinfo module.
	* src/ (grep_SOURCES): Add localeinfo.c.
	(noinst_HEADERS): Add localeinfo.h.
	* src/dfa.c: Include localeinfo.h.
	(struct dfa): Remove multibyte member, as it is now part of
	localeinfo.  New members simple_locale and localeinfo.
	Put locale-related members at the end.
	(mbrtowc_cache): Remove; now part of dfa->localeinfo.
	(charclass_index): Rename back from dfa_charclass_index,
	since it's private.
	(unibyte_word_constituent): New arg DFA; use its sbctowc member.
	(using_utf8, dfa_using_utf8, init_mbrtowc_cache, check_utf8):
	Remove; now done by localeinfo members.  All uses changed.
	(dfasyntax): New localeinfo arg.  Move to end to avoid forward decls.
	Initialize the entire DFA.
	(unibyte_c, check_unibyte_c): Remove; now in simple_locale member.
	(using_simple_locale): Now takes bool instead of DFA.
	Do the locale check here, rather than in the caller,
	as the result is now cached in dfa->simple_locale.
	(dfaalloc): Just allocate the DFA.  dfasyntax now initializes it.
	* src/dfa.h: Add forward decl of struct localeinfo.
	Adjust to new dfa.c API.
	* src/dfasearch.c (localeinfo): New var, replacing former static
	vars like mbrtowc_cache.
	* src/localeinfo.c, src/localeinfo.h: New files.
	* src/search.h: Include localeinfo.h.
	(localeinfo): New decl.
	* src/searchutils.c (mbclen_cache, build_mbclen_cache):
	Remove.  All uses changed to localeinfo.
	* tests/ (dfa_match_aux_LDADD): Add localeinfo.o.
	* tests/dfa-match-aux.c: Include localeinfo.h.
	(main): Adjust to changes in DFA API.

2016-08-28  Paul Eggert  <>

	build: update gnulib submodule to latest
	This should fix Bug#24323 reported by Dennis Clarke, where grep
	does not build on Solaris 10 when compiled with Solaris Studio 12.4.

2016-08-23  Paul Eggert  <>

	dfa: minor thread-safety cleanups
	* src/dfa.c (struct lexer_state): Rename lexptr to ptr and lexleft
	to left, for brevity.  All uses changed.
	(struct dfa): Rename lexstate to lex and parsestate to parse,
	for brevity.  All uses changed.
	(using_simple_locale): Simplify boolean expression.
	(FETCH_WC): Parenthesize uses of dfa macro arg.
	(FETCH_WC, parse_bracket_exp, addtok_mb): Prefer suffix operators
	on structure members when possible, for clarity.
	(parse_bracket_exp): Check for buffer exhaustion before
	dereferencing buffer pointer.
	(struct lexptr): New type.
	(push_lex_state, pop_lex_state): Use it.  Change from macros
	PUSH_LEX_STATE and POP_LEX_STATE to static functions, and add
	parameters to make them proper C functions.  All uses changed.
	(lex): Simplify tests for \) and \|.  Avoid some string
	duplication by using &"^..."[boolean].
	(dfaalloc): Use xzalloc, not xcalloc with 1.

2016-08-21  Paul Eggert  <>

	grep: minor tweaks of initial buffer alloc
	* src/grep.c (main): Allocate input buffer only when about
	to do I/O.  Avoid int overflow on systems with 2 GiB pages.
	Fix size_t overflow check.

2016-08-20  Zev Weiss  <>

	dfa: constify some function parameters
	* src/dfa.c (char_context): Mark dfa parameter const.
	(charclass_context): Likewise.

	dfa: thread-safety: initialize mbrtowc_cache in dfa_init
	* src/dfa.c (dfasyntax): Remove initialization of mbrtowc_cache.
	(init_mbrtowc_cache): New function.
	(dfa_init): Call it.

	dfa: thread-safety: eliminate static local variables
	* src/dfa.c: Replace utf8 and unibyte_c static local variables with
	static globals initialized by a new function dfa_init() which must be
	called before any other dfa*() functions.
	(dfa_using_utf8): Rename using_utf8() to dfa_using_utf8() for
	consistency with other exported functions.
	* src/dfa.h (dfa_using_utf8): Rename using_utf8() to dfa_using_utf8();
	(dfa_init): New function.
	* src/grep.c (main), tests/dfa-match-aux.c (main): Call dfa_init().
	* src/dfasearch.c (EGexecute): Replace using_utf8 with dfa_using_utf8.
	* src/kwsearch.c (Fexecute): Likewise.
	* src/pcresearch.c (Pcompile): Likewise.

	dfa: thread-safety: move regex syntax configuration into struct dfa
	* src/dfa.c: move global variables holding regex syntax configuration
	into a new struct (`struct regex_syntax') and add an instance of it to
	struct dfa.  All references to the globals are replaced with
	references to the dfa struct's new member.  As a side effect, a
	`struct dfa' must be allocated with dfaalloc() and passed to
	* src/dfa.h (dfasyntax): Add new struct dfa* parameter.
	* src/dfasearch.c (GEAcompile): Allocate `dfa' earlier and pass it to
	* tests/dfa-match-aux.c (main): Pass `dfa' to dfasyntax().

	dfa: thread-safety: move parser state into struct dfa
	* src/dfa.c: move global variables holding parser state (`tok' and
	`depth') into a new struct (`struct parser_state') and add an instance
	of it to struct dfa.  All references to the globals are replaced by
	references to the dfa struct's new member.

	dfa: thread-safety: move lexer state into struct dfa
	* src/dfa.c: move global variables holding lexer state into a new
	struct (`struct lexer_state') and add an instance of this struct to
	struct dfa.  All references to the globals are replaced with
	references to the dfa struct's new member.

2016-08-19  Zev Weiss  <>

	dfa: thread-safety: remove dfa.c's "dfa" global
	Remove the global dfa struct.  Instead, add a struct dfa pointer
	parameter to each function that had been using the global.
	* src/dfa.c (dfa): Remove file-scoped global.
	(charclass_index): Remove now-unnecessary function.
	(using_simple_locale): Add a dfa parameter and update all callers.
	(FETCH_WC, parse_bracket_exp, lex, addtok_mb, addtok): Likewise.
	(addtok_wc, add_utf8_anychar, atom, nsubtoks, copytoks): Likewise.
	(closure, branch, regexp): Likewise.
	(dfaparse): No longer set the global.

2016-08-18  Paul Eggert  <>

	grep: tune list_files conversion to enum
	* src/grep.c (grepdesc): Use a slightly more-efficient way to test

	grep: prefer bitwise to short-circuit when shorter
	* src/grep.c (skip_devices, initialize_unibyte_mask, fillbuf, main)
	* src/kwsearch.c (Fexecute): Prefer bitwise to short-circuit ops
	when they are logically equivalent and the bitwise ops generate
	shorter code on GCC 6.1 x86-64.
	* src/grep.c (get_nondigit_option, parse_grep_colors):
	Use c_isdigit instead of spelling it out with a short-circuit op.

2016-08-17  Paul Eggert  <>

	dfa: use 64-bit when ulong is at least that wide
	* src/dfa.c (charclass_word): Now unsigned long instead of unsigned.
	(CHARCLASS_WORD_BITS): Now 64 on 64-bit platforms.
	(CHARCLASS_WORD_MASK): Now a static const, since it no longer
	needs to be a macro.
	(equal): Open-code rather than calling memcmp.
	(add_utf8_anychar): Use CHARCLASS_INIT.

	dfa: avoid uninitialized constants
	Some compilers warn about 'static int const x;' on the grounds
	that X should have an initializer.  Instead of worrying about
	this, rewrite to avoid this sort of thing.
	* src/dfa.c (emptyset): New function.
	(parse_bracket_exp): Use it instead of 'equal' and a zero constant.
	* src/dfasearch.c (struct patterns): Remove tag 'patterns'.
	(patterns0): Remove zero constant.
	(GEAcompile): Use memset instead of the zero constant.

2016-08-17  Jim Meyering  <>

	maint: avoid new "make syntax-check" failure
	* src/dfa.c: Adjust comment not to go past column 80.

	tests: pcre-jitstack: avoid false failure without base64 -d support
	* tests/pcre-jitstack: Try harder to find a base64 decoder:
	try 'base64 -d', 'base64 -D', 'openssl base64 -d' and perl's
	MIME::Base64 decode_base64.  The old code would fail at least on
	OS X, for which base64 expects -D or --decode.
	Reported by Jack Howarth in

2016-08-16  Paul Eggert  <>

	dfa: minor refactoring and doc fixes
	* NEWS: Improve description of recent change.
	* src/dfa.c: Improve commentary.  Indent new code (and some
	long-existing howlers) more in GNU style.
	(dfa_state): Reorder members to make struct smaller on x86.
	mb_trindex member is now state_num, not size_t, so that -1 is more
	natural; all uses changed.
	(struct dfa): Similarly for mb_trcount member.
	(state_index): Compute values for new state components before
	allocating the state, to make the code easier to understand.
	(state_index, dfastate): Prefer A & ~B to other forms like (A & B)
	!= A.
	(dfastate, build_state, transit_state): In new code, prefer i++ to
	++i in for-loop control.
	(build_state, transit_state): In new code, prefer < to >.
	(transit_state): Add to *PP in one assignment, rather than in a
	loop.  Prefer !x to x == NULL.  Use xmalloc instead of xnmalloc,
	since the size is a constant.  Do the size calculation as a signed
	integer constant expression, so that the compiler diagnoses any
	(transit_state, free_mbdata): Tune by looping from -1 to N - 1,
	rather than from 0 to N - 1 with a separate instance for -1.
	(dfaexec_main): Rewrite to avoid side effects in if-part.
	(free_mbdata): Simplify.

	dfa: port to C90
	* src/dfa.c (transit_state, dfa_supported, dfamust):
	Don't use declarations after statements.
	If I recall correctly, gawk still wants to port to C90.

	dfa: fix context newline confusion
	* src/dfa.c (transit_state): Fix "... & ~0" that was evidently
	intended to be "... & ~1".  Do index calculation in a simpler way,
	that uses just addition (Bug#21486).

2016-08-16  Norihiro Tanaka  <>

	dfa: improve leading "." with non-UTF8 multibyte
	In non-UTF8 multibyte locales, matching the dot expression is very
	slow, as the next state is calculated on demand.  This change caches
	the result for the typical case (Bug#21486).

	Compare the run times of this command before and after this change,
	on a i5-4570 CPU @ 3.20GHz using rawhide (~fedora 22) and compiled
	with gcc 5.1.1 20150618:
	yes "$(printf 'a%38db\n' 0)" | head -1000000 >in
	env LC_ALL=ja_JP.eucJP time -p \
	src/grep .......................................... in
	  Before: 19.10
	  After :  0.55

	* NEWS: Document this.
	* src/dfa.c: (struct dfa_state): New members curr_dependent, mb_trindex.
	(MAX_TRCOUNT): New constant.
	(struct dfa): New members mb_trans, mb_trcount.
	(state_index): Initialize new members of struct dfa_state and calculate
	dependency on context of next character for positions for dot.
	(dfastate): Calculate follows positions for dot if enabled.
	(realloc_trans_if_necessary): Allocate transition tables.
	(build_state): Use new constant and reset transition tables.
	(transit_state): Use cache for transition from a state with the dot
	(free_mbdata): Deallocate transition tables.

2016-08-06  Jim Meyering  <>

	tests: standardize on 10-second timeouts to avoid rare false failure
	In a parallel test run, it is not unusual to exceed a timeout of
	1-3 seconds.  Increase several from 3 or fewer to 10 seconds.
	* tests/skip-device: Increase timeout from 2 to 10 seconds.
	* tests/grep-dev-null-out: Likewise, but s/1/10/.
	* tests/pcre-invalid-utf8-input: Likewise, but s/3/10/.
	* tests/dfa-match: Likewise.
	* tests/pcre-invalid-utf8-infloop: Likewise.
	* tests/pcre-infloop: Likewise.
	* tests/max-count-overread: Likewise.
	* tests/invalid-multibyte-infloop: Likewise.
	Prompted by

	tests/backref-multibyte-slow:: avoid false positive
	* tests/backref-multibyte-slow: When redirecting the "fast" LC_ALL=C
	run's output to /dev/null, we got an artificially low timing (of 0),
	due to grep's own stdout-vs-/dev/null optimization.  With an initial
	timing of 0 on that first run, the derived timeout for the UTF-8 run
	(which redirects to a file) would be a mere 1 second.  The fix: also
	redirect that first run's output to a file, not to /dev/null.

2016-08-05  Norihiro Tanaka  <>

	dfa: minor fix for whether dfa is "fast"
	* src/dfa.c (dfaoptimize): When a UTF-8 optimization succeeds for
	a DFA (it can use single-byte code paths), record that by setting
	its ->fast flag.

2016-07-25  Jim Meyering  <>

	grep: print "filename:lineno:" in invalid-regex diagnostic
	Determining the file name and line number is a little tricky because
	of the way the regular expressions are all concatenated onto a newline-
	separated list.  By the time grep would compile regular expressions,
	the <filename,lineno> origin of each regexp was no longer available.
	This patch adds a list of filename,first_lineno pairs, one per input
	source, by which we can then map the ordinal regexp number to a
	filename,lineno pair for the diagnostic.

	* src/dfasearch.c (GEAcompile): When diagnosing an invalid regexp
	specified via -f FILE, include the "FILENAME:LINENO: " prefix.
	Also, when there are two or more lines with compilation failures,
	diagnose all of them, rather than stopping after the first.
	* src/grep.h (pattern_file_name): Declare it.
	* src/grep.c: (struct FL_pair): Define type.
	(fl_pair, n_fl_pair_slots, n_pattern_files, patfile_lineno):
	Define globals.
	(fl_add, pattern_file_name): Define functions.
	(main): Call fl_add for each type of the following: -e argument,
	-f argument, command-line-specified (without -e) regexp.
	* tests/ New file.
	* tests/ (TESTS): Add it.
	* NEWS (Improvements): Mention this.
	Initially reported by Gunnar Wolf in
	Forwarded to grep's bug list by Santiago Ruano Rincón as

2016-07-24  Jim Meyering  <>

	tests: add coreutils' perl-driven test framework
	* Set the AM_CONDITIONAL variable, HAVE_PERL.
	* tests/ New file.
	* tests/ New file.
	* tests/ New file.
	* tests/no-perl: New file.
	* tests/ Set up to use .pl tests:
	(EXTRA_DIST): Add the four new file names.

	doc: omit an excess word in HACKING

2016-07-21  Norihiro Tanaka  <>

	grep: always match single line only with DFA superset
	\n cannot occur inside a multibyte character.  So an input always
	matches single line only with DFA superset.

	* src/dfasearch.c (EGexecute): Simplify it with above.

2016-07-15  Norihiro Tanaka  <>

	dfa: fix whitespace problems
	* src/dfa.c: Use GNU style for pointer decls.

2016-07-15  Paul Eggert  <>

	maint: modernize HACKING a bit
	* HACKING: Remove some ancient history to simplify maintenance.

2016-07-14  Paul Eggert  <>

	grep: minor style changes for -F crash fix
	* src/kwset.c (memoff2_kwset): Use ?: instead of if-else.

2016-07-14  Norihiro Tanaka  <>

	grep: fix -F crash when alternating duplicates
	grep -F crashes with a pattern like 0\n0.
	This bug was introduced in 966f6586fbce3081ce6e5e2f9b55301b0ec3d2b4.

	* src/kwset.c (memoff2_kwset): If two characters are the same,
	use memchr instead of memchr2.
	* tests/two-chars: New test.
	* tests/ (TESTS): Add it.

2016-07-07  Paul Eggert  <>

	dfa: fix comments to match code better
	* src/dfa.c: Fix comments.

2016-07-06  Paul Eggert  <>

	dfa: don't treat null bytes specially
	* src/dfa.c (transit_state): Do not treat null byte specially
	when eolbyte == '\n'.

2016-07-06  Norihiro Tanaka  <>

	dfa: don't distingish letter in non-POSIX locales
	For non-POSIX locales, dfa does not support word delimiter
	support, so remove distinction between letters and non-letters.
	* src/dfa.c (struct dfa): Remove members initstate_letter,
	initstate_others.  All uses removed.  New member initstate_notbol.
	(dfaanalyze, dfaexec_main): Replace old members with new member.
	(wchar_context): Remove.  Update callers.

2016-07-06  Paul Eggert  <>

	dfa: minor cleanups for non-POSIX simplification
	* src/dfa.c (transit_state_singlebyte): Remove unnecessary 'const'
	from arg; we usually don't bother with 'const' on locals.
	(transit_state_singlebyte): Omit '!= NULL' in boolean context.
	Use assert rather than abort.

2016-07-06  Norihiro Tanaka  <>

	dfa: simplify for non-POSIX locales
	Simplify the dfa code, since it no longer supports ranges,
	collating elements, and equivalent classes in non-POSIX locales.
	* src/dfa.c (struct dfa): Remove mb_match_lens.
	(enum status_transit_state, match_anychar)
	(check_matching_with_multibyte_ops, transit_state_consume_1char):
	(State_transition): Remove.
	(transit_state_singlebyte): Accepts pointer-to-pointer position,
	instead of pointer, and no longer accept pointer to next state.
	Return next state instead of status_transit_state.  All callers
	(transit_state_singlebyte, transit_state): Simplify.
	(dfaexec_main): Now transit_state is called only when next character
	matches with ANYCHAR.

2016-06-14  Paul Eggert  <>

	doc: propagate more changes from grep.texi
	Problem reported by Björn Voigt in:
	* doc/ Fix more inconsistencies with grep.texi.

2016-06-13  Paul Eggert  <>

	doc: remove obsolete MS-DOS mention
	* doc/ Remove obsolete discussion of MS-DOS heuristics.
	Problem reported by Björn Voigt in:

2016-06-09  Zev Weiss  <>

	grep: do pagesize initialization and buffer allocation earlier
	* src/grep.c (reset, main): We're going to need pagesize and buffer
	initialized anyway, so we might as well do so unconditionally early on
	rather than checking on every call to reset().

	grep: remove unnecessary dirdesc variable.
	* src/grep.c (grepdirent): Remove dirdesc variable and just use
	fts_cwd_fd directly, since the fts_options test was guaranteed to
	succeed (and fts_cwd_fd was already being used directly in fstatat()

	grep: convert list_files to an enum
	* src/grep.c: Make list_files a tristate enum instead of an int.

	grep: correct a stale comment and remove dead code
	* src/grep.c (grepdesc): The `grep()' function no longer has
	special-case negative return values, since it no longer handles
	directories, so don't bother checking for them.

	maint: replace bitwise with logical OR
	* src/grep.c (main): replace bitwise ORs with logical ORs where it
	makes sense (when dealing with boolean conditions as opposed to

	maint: mark a couple of static variables const
	* src/dfa.c (parse_bracket_exp): mark zeroclass const.
	* src/dfasearch.c: mark patterns0 const.

2016-06-03  Paul Eggert  <>

	tests: fix similar bug in exit status test
	* tests/grep-dir (status_range): New shell function.
	Use it to fix bug where $? was not saved properly.

2016-06-03  Zev Weiss  <>

	tests: fix bug in exit status test
	When checking $? against multiple values, save its value in another
	variable and check that so as to avoid tests beyond the first seeing a
	$? clobbered by earlier ones.

	* tests/status: save $? in a temporary variable before testing it.

2016-06-02  Norihiro Tanaka  <>

	dfa: more simplification of dfaexec_main
	* src/dfa.c (dfaexec_main): Failure at an acceptable position and demand
	to build state is unlikely.  So go next loop without checking them after
	a newline.  This commit induces no semantic change.

2016-06-02  Paul Eggert  <>

	maint: correct attribution
	* build-aux/git-log-fix: Fix attribution of primary Aho-Corasick patch

2016-06-02  Paul Eggert  <>

	grep: simplify -F Aho-Corasick a bit
	This removes some tuning that complicates the code without providing
	performance benefits that I could measure (GCC 6.1, x86-64).
	(acexec_trans): Do not hand-unroll.  Unduplicate the code for a
	transition step.

	* src/kwset.c (struct kwset.kwsexec, bmexec, acexec_trans, acexec)

2016-06-02  Paul Eggert  <>

	grep: minor cleanups for -F Aho-Corasick
	* NEWS: Don't claim 7x, as the value seems to be system-dependent.
	* src/kwset.c (struct kwset.kwsexec, bmexec, acexec_trans, acexec):
	* src/kwset.c, src/kwset.h (kwsalloc, kwsexec):
	Don't put 'const' into the declaration when that is irrelevant to
	the API.  More generally, don't bother with 'const' when it's only
	a local so it is reasonably obvious to a reader that it is 'const'
	anyway.  It would be overkill to add 'const' to all locals that
	never change.
	* src/kwset.c (U): Avoid unnecessary parens.
	(treefails, memoff2_kwset, bmexec_trans, bmexec, cwexec, acexec_trans):
	Prefer SIZE_MAX to (size_t) -1.
	(bmexec_trans, cwexec, acexec_trans):
	Remove attributes for static functions that no longer seem needed.
	(memoff2_kwset): Rename from memchr2_kwset, since it returns
	an offset, not a pointer.  All uses changed.
	(cwexec, acexec_trans) [lint]: Remove initialization that is no
	longer needed; at least, GCC 6.1 x86-64 does not need it.
	(acexec_trans): Clarify code by using nesting rather than 'continue'.

2016-06-02  Norihiro Tanaka  <>

	grep: use memchr2 for two patterns of a character
	* src/kwset.c (memchr2_kwset): Add a new function.  grep uses memchr2 to
	search just two letters.
	(cwexec, acexec_trans): Use it.

	grep: -F multiword longest match not always needed
	Searching multiple fixed words, grep immediately returns without longest
	match if not needed.  Without this change, grep tries longest match for
	multiple words even if not needed.
	* src/kwset.c (kwsexec, acexec, cwexec, bmexec): Add a bool argument
	for whether longest match is needed.  All callers changed.
	* src/kwset.h (kwsexec): Update prototype.

2016-06-02  Norihiro Tanaka  <>

	grep: use Aho-Corasick algorithm to search multiple fixed words
	Searching multiple fixed words, grep used the Commentz-Walter
	algorithm, but this was O(m*n) and was very slow in the worst case.
	For example:

	  - input: yes `printf %040d` | head -10000000
	  - word1: x0000000000000000000
	  - word2: x

	This change instead uses the Aho-Corasick algorithm to search multiple
	fixed words.  It uses a high-quality trie-building function that is
	already defined for Commentz-Walter in kwset.c.

	I see 7x speed-up even for a typical case on Fedora 21 with a 3.2GHz i5
	by this change.  Using best-of-5 trials for the benchmark:

	    find /usr/share/doc/ -type f |
	    LC_ALL=C time -p src/grep -Ff /usr/share/dict/linux.words >/dev/null

	The results were:

	    real 11.37  user 11.03  sys 0.24  [without the change]
	    real 1.49   user 1.31   sys 0.15  [with the change]

	* src/kwset.c (struct kwset): Add a new member 'mode'.
	(kwsalloc): Use it.
	All callers are changed.
	(kwsincr): Using Aho-Corasick algorithm, build tries in normal order.
	(acexec_trans, acexec): Add a new function.
	(kwsexec): Use it.
	* src/kwset.h (kwsalloc): Update a prototype.
	* NEWS (Improvements): Mention it.

2016-05-13  Jim Meyering  <>

	maint: do not let a LANGUAGE envvar setting perturb tests
	E.g., running "LANGUAGE=eo make check" would provoke a failure
	of the encoding-error test, on systems that mistakenly let that
	envvar trump the setting of LC_ALL.
	* tests/envvar-check: New file, copied from coreutils.
	* tests/ (EXTRA_DIST): Add it.
	Also select TMPDIR as we do for coreutils tests.
	Reported by Benno Schulenberg in

2016-05-02  Jim Meyering  <>

	maint: avoid NEWS syntax-check failure
	* NEWS: Move the mention of the /dev/null speed-up from the
	block for 2.25 into the current, in-preparation block.

2016-05-01  Paul Eggert  <>

	dfa: prefer bool for boolean
	* src/dfa.c (syntax_bits_set, dfasyntax, using_utf8, FETCH_WC)
	(POP_LEX_STATE, State_transition):
	* src/dfa.h (using_utf_8):
	Use bool for boolean.

2016-05-01  Norihiro Tanaka  <>

	dfa: stop exporting internal functions
	* src/dfa.c, src/dfa.h (dfaparse, dfaanalyze, dfastate, dfainit):
	Now static.

	dfa: prefer bool at DFA interfaces
	* src/dfa.c (struct dfa, dfasyntax, dfaanalyze, dfaexec_main)
	(dfaexec_mb, dfaexec_sb, dfaexec_noop, dfaexec, dfacomp):
	* src/dfa.h (dfasyntax, dfacomp, dfaexec, dfaanalyze):
	* src/dfasearch.c (EGexecute):
	Use bool for boolean.

2016-05-01  Paul Eggert  <>

	dfa: speed up checking for character boundary
	This should help performance with gawk; not so much with grep.
	Suggested by Norihiro Tanaka in:
	* src/dfa.c (never_trail): New static var.
	(dfasyntax): Initialize it.
	(skip_remains_mb): Use it to speed up a common case in Gawk.

	grep: /dev/null output speedup
	This sped up 'seq 10000000000 | grep . >/dev/null' by a factor of
	380,000 on my platform (Fedora 23, x86-64, AMD Phenom II X4 910e,
	en_US.UTF-8 locale).
	* NEWS: Document this.
	* src/grep.c (grepbuf): exit_on_match no longer implies that -q
	was specified, so when a match is found, exit with exit_failure if
	an error was also found.
	(grepdesc): Omit unnecessary S_ISREG and st_ino checks.
	out_stat.st_ino is zero if stdout is not a regular file,
	and this cannot possibly equal st->st_ino.
	(main): Omit duplicate initialization of exit_failure.  Do not
	bother with isatty unless -q is not used and stdout is a character
	special file and --color=auto and TERM says colorization is
	possible.  Most importantly, set exit_on_match if the output is
	* tests/grep-dev-null-out: New test.
	* tests/ (TESTS): Add it.
	* tests/status: Do not require grep to actually read all the input
	files when the output is /dev/null and a matching line has been

2016-04-21  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.25
	* NEWS: Record release date.

2016-04-19  Paul Eggert  <>

	dfa: remove dependency on btowc
	MirOS BSD btowc is a macro that (when GCC is being used) hardcodes
	btowc (0x80) == WEOF regardless of locale, which contradicts
	future POSIX in the C locale.  Instead of bothering to develop a
	Gnulib workaround for the btowc incompatibility, use mbrtowc,
	which we are using elsewhere and fixing anyway, and are caching so
	it is fast here.  Problem reported by Nelson H. F. Beebe via Jim
	Meyering in:
	* bootstrap.conf (gnulib_modules): Remove btowc.
	* src/dfa.c (struct dfa): Remove mbrtowc_cache member, replacing with ...
	(mbrtowc_cache): ... this new static var.  All uses changed.
	(dfambcache): Remove; now done by setsyntax.  Call removed.
	(is_valid_unibyte_character): Remove.
	(IS_WORD_CONSTITUENT): Remove this macro, replacing it with ...
	(unibyte_word_constituent): ... this new function.  It uses
	mbrtowc_cache rather than btowc.
	(dfasyntax): Initialize mbrtowc_cache before using it.

2016-04-10  Paul Eggert  <>

	grep: minor doc tweaks inspired by Debian
	Problem reported by Santiago Ruano Rincón in:
	* doc/
	* doc/grep.texi (Matching Control, grep Programs)
	(Regular Expressions):
	Document -e, -f, and PCRE more carefully.

2016-04-10  Jim Meyering  <>

	maint: remove unused mbtoupper function
	* src/searchutils.c (mbtoupper): Remove now-unused function.
	Also remove inclusion of <assert.h>, since this change removed
	the final use of assert.
	* src/search.h (mbtoupper): Remove declaration.

2016-04-10  Paul Eggert  <>

	grep: in C locale, all bytes are valid characters
	This works around glibc bug 19932:
	The actual bug fix was the update to the current version of Gnulib.
	grep problem reported by Björn Jacke in:
	* NEWS: Mention this.
	* doc/grep.texi (File and Directory Selection): Crossref to LC_*
	section.  Suggest why -a or LC_ALL=C might be useful.
	(Environment Variables): Mention 'locale -a'.
	Say that LC_CTYPE also specifies encoding, and that every
	byte is a valid character in the C or POSIX locale.
	* tests/c-locale: New test.
	* tests/ (TESTS): Add it.

	build: update gnulib submodule to latest

2016-04-05  Paul Eggert  <>

	Give another example of binary file processing
	Problem reported by Shlomi Fish
	* doc/grep.texi (File and Directory Selection):
	Document that 'q$' might match 'q' followed by a NUL
	if --binary-files=binary is in effect.

2016-04-03  Paul Eggert  <>

	tests: test egrep/fgrep help only if our grep
	Problem reported by Christian Weisgerber in:
	Test egrep and fgrep only if they use our grep.

2016-03-29  Jim Meyering  <>

	tests: remove spurious test of egrep
	* tests/reversed-range-endpoints: Do not test egrep here.
	There is already a test of grep -E.
	Prompted by

2016-03-23  Paul Eggert  <>

	grep: -Pz no longer misdiagnoses [^a]
	Problem reported by Michael Jess.
	* NEWS: Document this.
	* src/pcresearch.c (Pcompile): Do not diagnose [^ when [ is unescaped.
	* tests/pcre: Test for the bug.

2016-03-22  Jim Meyering  <>

	maint: move new 'Improvements' blurb into proper section
	* NEWS (Improvements): Move this new section from within the block
	for the already-released 2.24 into the proper "next-release" block.
	Also, retain the 2-blank-line separator between blocks.

2016-03-18  Jim Meyering  <>

	maint: avoid spurious "binary file ... matches" in generated THANKS
	* (THANKS): Don't apply grep to a stream containing
	NUL bytes.  Sync this rule from the one in coreutils: it was missing
	some improvements.
	Reported by Bailes Magio in

2016-03-18  Paul Eggert  <>

	grep: -oz now outputs null bytes, not newlines
	* NEWS: Document this.
	* doc/grep.texi (Other Options): Clarify that -z affects output
	as well as input data.
	* src/grep.c (print_line_middle): Output eolbyte, not newline, if -o.
	* tests/null-byte: Test -o too.
	* tests/pcre-context: Adjust test to match new behavior.

2016-03-17  Paul Eggert  <>

	grep: use errno consistently in write diagnostics
	Feature request and initial version reported by Assaf Gordon in:
	* NEWS: Document this.
	* src/grep.c: Include <stdarg.h>.
	(stdout_errno): New static var.
	(write_error_seen): Remove; superseded by stdout_errno.
	All uses changed.
	(putchar_errno, fputs_errno, printf_errno, fwrite_errno)
	(fflush_errno): New static functions.
	(print_filename, print_sep, print_offset, print_line_head)
	(print_line_middle, print_line_tail, prline, prtext, grep)
	(grepdesc): Use them.
	* tests/write-error-msg: New file.
	* tests/ (TESTS): Add it.

2016-03-10  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.24
	* NEWS: Record release date.

2016-02-28  Jim Meyering  <>

	maint: add
	This file augments "make distcheck" rules.
	* New file, from coreutils via gzip.
	* (EXTRA_DIST): Add it.
	* Include it.

2016-02-21  Paul Eggert  <>

	grep: -Pz is incompatible with ^ and $
	Problem reported by Sergei Trofimovich in:
	* NEWS: Document this.
	* src/pcresearch.c (Pcompile): Warn with -Pz and anchors.
	* tests/pcre: Test new behavior.

2016-02-21  Jim Meyering  <>

	tests: test cleanup
	* tests/z-anchor-newline: Remove test artifact that would write
	to /t/x.

2016-02-20  Jim Meyering  <>

	grep -z: avoid erroneous match with regexp anchor and \n in text
	* src/dfasearch.c (EGexecute): Clear the newline_anchor bit when
	eolbyte is not '\n'.
	* tests/z-anchor-newline: New file.
	* tests/ (TESTS): Add it.
	* NEWS (Bug fixes): Describe it.
	Originally reported by Ulrich Mueller in
	Reported to us by Sergei Trofimovich as

	tests: convert "cmd && fail=1" to "returns_ 1 cmd || fail=1"
	The latter is robust, while the former can silently ignore
	failure due to signals.
	* (sc_prohibit_and_fail_1): New rule, copied from coreutils.
	* tests/long-pattern-perf: Perform the above substitution.
	* tests/mb-non-UTF8-performance: Likewise.
	* tests/help-version: Merge from coreutils.

2016-02-09  Jim Meyering  <>

	maint: add a check-very-expensive target
	* (check-very-expensive): New convenience rule,
	currently merely equivalent to check-expensive.

2016-02-04  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.23
	* NEWS: Record release date.

2016-02-02  Jim Meyering  <>

	gnulib: update to latest
	Update for this "make distcheck"-fixing change:
	> verify-tests: also remove stray test-verify.Tpo

2016-02-01  Jim Meyering  <>

	tests/null-byte: test another code path
	* tests/null-byte: Also exercise the case in which there is
	a match in the block along with the NUL byte.

2016-01-31  Paul Eggert  <>

	Omit excess "Binary file ... matches"
	Problem reported in:
	* src/grep.c (grep): Don't report "Binary file ... matches"
	merely because the file contained both matches and binary data.
	Insist that the binary data contained a match.
	* tests/null-byte: Add a test for this.

2016-01-28  Jim Meyering  <>

	gnulib: update to latest

2016-01-23  Jim Meyering  <>

	gnulib: update to latest

	maint: fix typo in NEWS: s/a/an/

2016-01-15  Paul Eggert  <>

	grep: -x now supersedes -w more consistently
	* NEWS, doc/grep.texi (Matching Control): Mention this.
	* src/dfasearch.c (EGexecute):
	* src/pcresearch.c (Pcompile):
	Don't get confused by -w if -x is also present.
	* src/pcresearch.c (Pcompile): Remove misleading comment about
	non-UTF-8 multibyte locales, as PCRE doesn't support them.
	Calculate buffer sizes more carefully; the old method
	allocated a buffer slightly too big, seemingly due to luck.
	* tests/backref-word, tests/pcre: Add tests for this bug.

	tests: omit update-copyright-tests
	This test does not check how 'grep' itself operates, so it is
	out of place for grep's 'make check'.  Problem reported by Sam Razavi in:
	* bootstrap.conf (avoided_gnulib_modules): Add update-copyright-tests.

2016-01-11  Jim Meyering  <>

	tests: do use "yes" but via an AWK replacement
	Also, use sed Nq in place of head -N
	* tests/init.cfg (yes): Define.
	Thanks to Paul Eggert for this definition.
	* tests/max-count-overread: Revert to using "yes".
	* tests/mb-non-UTF8-performance: Likewise, and use
	"sed Nq" in place of head -N.

2016-01-11  Paul Eggert  <>

	* tests/pcre-count: Don't assume the page size is 32kB.

2016-01-08  Paul Eggert  <>

	tests: port to other POSIXish platforms
	I tested this on Solaris 10 and AIX 7.1.
	* tests/max-count-overread:
	* tests/mb-non-UTF8-performance:
	Don't assume 'yes' exists, as 'yes' is not in POSIX.
	* tests/mb-non-UTF8-performance:
	Don't rely on 'head -1000', as that option syntax is not POSIX.
	* tests/pcre-count: Don't rely on "printf '\x0'".
	* tests/unibyte-binary: Don't assume \200 is an encoding error
	in every unibyte locale.

2016-01-08  Jim Meyering  <>

	tests: fix encoding-error test failure to use of printf '\xHH'
	* tests/encoding-error: Don't rely on printf having support for \xHH
	hexadecimal.  That is not portable.  Use \OOO octal, instead.

	maint: fix typo in NEWS: s/a/an/

2016-01-07  Jim Meyering  <>

	mb-non-UTF8-performance: avoid FP test failure on fast hardware
	* tests/mb-non-UTF8-performance: Don't use a fixed size.
	Otherwise, on a fast system, the fixed-size unibyte test
	would complete in a nominal 0 ms, which might well be
	smaller than 1/30 of the multibyte duration, provoking
	a false positive test failure.  Instead, increase the
	size of the input until we obtain a unibyte duration of
	at least 10ms.

2016-01-07  Paul Eggert  <>

	doc: mention unibyte encoding fix
	* NEWS: Document recent fix for encoding errors in unibyte locales.

	grep: improve unibyte -P performance
	This is a followon to the recent changes prompted by Bug#20526.
	In <> Norihiro Tanaka pointed out
	that grep mistakenly assumed that unibyte locales cannot have
	encoding errors.  Here, the mistake hurt performance significantly.
	On Fedora 23 x86-64 in the C locale, this patch improved grep's
	performance by a factor of 7 when run as "grep -P 'z.*a'" on the
	output of "yes $(printf '\200\n') | head -n 1000000000".
	* src/pcresearch.c (multibyte_locale) [HAVE_LIBPCRE]: New static var.
	(Pcompile): Set it.
	(Pexecute): Use it to avoid the need to call
	buf_has_encoding_errors in unibyte locales.

2016-01-06  Paul Eggert  <>

	Improve on fix for Bug#22181
	* src/pcresearch.c (Pexecute): Update subject when skipping past
	easily-determined encoding errors, as this is faster than letting
	pcre_exec skip them.  On my platform this improves performance
	4.7x on a benchmark created via "yes $(printf '\200\200\200\200
	| head -n 1000000 >j; grep -oP y j" in a UTF-8 locale.  Rework
	code that deals with PCRE_ERROR_BADUTF8 return, to avoid an
	incorrect (albeit currently harmless) 'bol = false' assignment.

	grep: restore -P optimization (followup fix)
	* src/search.h (EGexecute, Fexecute, Pexecute):
	Change decls to match new implementations.
	I forgot to add this file to the previous commit.

	grep: restore -P PCRE_NO_UTF8_CHECK optimization
	On my platform in the en_US.utf8 locale, this makes 'grep -P "z.*a" k'
	220x faster, where k is created by the shell command:
	yes 'abcdefg hijklmn opqrstu vwxyz' | head -n 10000000 >k
	* src/dfasearch.c (EGexecute):
	* src/grep.c (execute_fp_t):
	* src/kwsearch.c (Fexecute):
	* src/pcresearch.c (Pexecute):
	First arg is now char *, not char const *, since Pexecute now
	temporarily modifies this argument.
	* src/grep.c, src/grep.h (buf_has_encoding_errors): Now extern.
	* src/pcresearch.c (Pexecute): Use it.  If the input is free of
	encoding errors, use a multiline search and the PCRE_NO_UTF8_CHECK
	option, as this is typically way faster.  This restores an
	optimization that was removed with the recent changes for binary
	file detection.

2016-01-05  Paul Eggert  <>

	Fix calculation of unibyte_mask
	* src/grep.c (initialize_unibyte_mask): The old method worked for
	UTF-8 and other typical encodings, but did not work for weird
	encodings, e.g., one where all bytes other than 0x7f and 0x80 are
	unibyte characters.

2016-01-01  Paul Eggert  <>

	grep: fix bug with with invalid unibyte sequence
	This was introduced by the recent binary-data-detection changes.
	Problem reported by Norihiro Tanaka in:
	* src/grep.c (HIBYTE, easy_encoding, init_easy_encoding): Remove,
	replacing with ...
	(uword_max, unibyte_mask, initialize_unibyte_mask): ... this new
	constant, static var, and function.  All uses changed.  The
	unibyte_mask var generalizes the old local var hibyte_mask, which
	worked only for encodings where every byte with 0x80 turned off is
	a single-byte character.
	(buf_has_encoding_errors): Return false immediately if
	unibyte_mask is zero, not whether the current encoding is unibyte.
	The old test was incorrect in unibyte locales in which some bytes
	were encoding errors.
	* tests/pcre-z: Require UTF-8 locale, since the grep -z . test now
	needs this.  Use printf \0 rather than tr.  Port the 'grep -z .'
	test to platforms where the C locale says '\200' is an encoding
	error.  Use cmp rather than compare, as the file is binary and
	so non-GNU diff might not work.
	* tests/unibyte-binary: New file.
	* tests/ (TESTS): Add it.

2016-01-01  Jim Meyering  <>

	maint: update copyright year, bootstrap,
	Run "make update-copyright" and then...

	* gnulib: Update to latest.
	* tests/ Update from gnulib.
	* bootstrap: Likewise.

2015-12-31  Paul Eggert  <>

	doc: clarify text vs binary match output
	* NEWS:
	* doc/grep.texi (File and Directory Selection):
	Make it clearer that grep can now output matching text before
	reporting a binary match.  Problem reported by Norihiro Tanaka in:

	doc: minor clarifications
	* doc/, doc/grep.texi: Minor clarifications suggested by
	Debian documentation patches.  Problem reported by Santiago Ruano
	Rincón in:

	grep: fix -l --line-buffer bug
	Problem reported by Louis Sautier in:
	* NEWS: Document this.
	* src/grep.c (grep, grepdesc): If --line-buffered, flush
	stdout after outputting newline (or null byte, if applicable).

2015-12-30  Paul Eggert  <>

	grep: remove duplicate init
	* src/grep.c (print_line_middle): Remove duplicate initialization.

	grep: report line-buffered write error right away
	* src/grep.c (prline): When line buffered, if there is a write
	error, report it immediately rather than waiting until the next
	line of output.

	grep: -c should keep counting after binary data
	Problem and fix reported by Jaroslav Škarvada, and test case
	reported by Norihiro Tanaka, in:
	* NEWS: Document this.
	* src/grep.c (grep): Don't stop counting merely because nulls seen.
	* tests/pcre-count: New file.
	* tests/ (TESTS): Add it.

	dfa: port to tinycc
	* src/dfa.c (add_utf8_anychar): Put 'const' after type.
	Problem reported by Aharon Robbins in:

	grep: be less picky about encoding errors
	This fixes a longstanding problem introduced in grep 2.21,
	which is overly picky about binary files.
	* NEWS:
	* doc/grep.texi (File and Directory Selection): Document this.
	* src/grep.c (input_textbin, textbin_is_binary, buffer_textbin)
	Remove.  All uses removed.
	(encoding_error_output): New static var.
	(buf_has_encoding_errors, buf_has_nulls, file_must_have_nulls):
	New functions, which reuse bits
	and pieces of the removed functions.
	(lastout, print_line_head, print_line_middle, print_line_tail, prline)
	(prpending, prtext, grepbuf):
	Avoid use of const, now that we have
	functions that require modifying a sentinel.
	(print_line_head): New arg LEN.  All uses changed.
	(print_line_head, print_line_tail):
	Return indicator whether the output line was printed.
	All uses changed.
	(print_line_middle): Exit early on encoding error.
	(grep): Use new method for determining whether file is binary.
	* src/grep.h (enum textbin, TEXTBIN_BINARY, TEXTBIN_UNKNOWN)
	(TEXTBIN_TEXT, input_textbin): Remove decls.  All uses removed.
	* src/pcresearch.c (Pexecute): Remove multiline optimization,
	since the main program no longer checks for encoding errors on input.
	* tests/encoding-error: New file.
	* tests/ (TESTS): Add it.

2015-12-29  Jim Meyering  <>

	maint: correct (make sorted) order of test file names
	* tests/ (TESTS): Insert new test name in sorted order.

2015-12-28  Paul Eggert  <>

	grep: --exclude matches trailing parts of args
	Problem reported by Vincent Lefevre in:
	* NEWS:
	* doc/grep.texi (File and Directory Selection): Document this.
	* src/grep.c (excluded_patterns, excluded_directory_patterns):
	Now 2-element arrays, with one element for subfiles and another
	for command-line args.  All uses changed.  This implements the change.
	(exclude_options): New function.
	* tests/include-exclude: Test the change.

2015-12-18  Jim Meyering  <>

	grep -oP: don't infloop when processing invalid UTF8 preceding a match
	* src/pcresearch.c (Pexecute): When advancing SUBJECT past an
	encoding error, don't blindly set P to that new value, since we
	will soon compute SEARCH_OFFSET = P - SUBJECT, and mistakenly
	making that difference too small would allow us to match some
	previously-processed text, resulting in an infinite loop.
	* NEWS (Bug fixes): Mention it.
	* Add Christian's name and email address.
	* tests/pcre-invalid-utf8-infloop: New file.
	* tests/ (TESTS): Add it.
	Reported by Christian Boltz in
	Introduced by commit, v2.21-37-g14f8e48.

2015-11-04  Jim Meyering  <>

	tests: mark performance-related tests as expensive
	These performance-related tests are slightly failure prone due to
	varying system load during the two runs.
	Marking these tests as "expensive" makes it so they are no longer run
	via "make check". You can still run them via make "check-expensive".
	This makes them less likely to be run by regular users.
	* tests/long-pattern-perf: Use expensive_.
	* tests/mb-non-UTF8-performance: Likewise.
	Reported by Jaroslav Skarvada in
	and by Andreas Schwab in

2015-11-01  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.22
	* NEWS: Record release date.

	tests: pcre-jitstack: upon failure, retry with no stack size limit
	* tests/pcre-jitstack: Don't let an example that provokes inordinate
	stack space use cause a test failure.  Thanks to reports from and
	analysis by Bruce Dubbs; see

2015-10-27  Jim Meyering  <>

	maint: update
	* Add name+email of those who found and reported
	the bug that made grep -E '^x|x$' match any "x".

2015-10-25  Zev Weiss  <>

	dfa: plug a memory leak in dfamust
	* src/dfa.c (dfamust): Ensure MP is freed, by refraining
	from returning early when, at "done:" *RESULT is NULL.

2015-10-25  Jim Meyering  <>

	gnulib: update to latest
	* gnulib: Pull in one more portability fix:
	  stdalign: port to Sun C 5.9

2015-10-24  Jim Meyering  <>

	gnulib: update to latest, for portability fixes
	* gnulib: Pull in changes like these:
	  fts: port to C11 alignof
	  stdalign: work around pre-4.9 GCC x86 bug

	maint: NEWS: correct/amend
	* NEWS: Move the long-regexp-performance-improvement from
	"Bug fixes" to "Improvements."  Say more and include an example.
	The -Fw degradation was introduced in commit v2.18-125-g94555dd

	tests: avoid spurious failure on OpenBSD 5.8
	* tests/fedora: Don't rely on "diff - FILE" reading from stdin.
	Reported privately by Nelson Beebe.

2015-10-17  Jim Meyering  <>

	gnulib: update to latest; also bootstrap and tests/
	* bootstrap: Update from gnulib.
	* tests/ Likewise.
	* gnulib: Update submodule to latest.

	build: avoid spurious bootstrap failure involving pkg.m4
	Running ./bootstrap could fail mistakenly at the very end in
	its attempt to obtain a copy of pkg.m4.  It would search only
	$(aclocal --print-ac-dir) and some other directories, but not
	those listed in $(aclocal --print-ac-dir)/dirlist.
	* bootstrap.conf (bootstrap_post_import_hook): Also search the
	directories named in $(aclocal --print-ac-dir)/dirlist when that
	file exists with nonzero size.

2015-10-16  Paul Eggert  <>

	maint: add news item
	* NEWS: Document grep -Fw speedup.

	grep: simplify previous change
	* src/grep.c (main): Simplify recently-changed grep -Fw test.

2015-10-16  Norihiro Tanaka  <>

	grep: use grep matcher for grep -Fw when unibyte
	In single byte locales with grep -Fw, prefer the grep matcher to the
	kwset matcher, as the former uses KWset and a DFA, whereas the latter
	calls kwsexec many times until it matches a word.
	* src/grep.c (main): Change pattern for fgrep into grep for grep -Fw in
	single byte locales.

2015-10-16  Paul Eggert  <>

	grep: use memchr/memrchar
	* src/kwsearch.c (Fexecute): Prefer memchr and memrchr to doing it
	by hand.

2015-10-16  Norihiro Tanaka  <>

	grep: improve performance of grep -Fw
	* src/kwsearch.c (Fexecute): grep -Fw examined whether the previous
	character is a word character after matching from the head of the
	buffer.  It is extremely slow.  Now, if grep found a potential match,
	it looks for the previous newline, and examines from there.

2015-10-13  Jim Meyering  <>

	maint: use single quote rather than UTF-8 multi-byte version
	* tests/backref-alt: Translate unnecessary non-ASCII in comment.

2015-10-13  Paul Eggert  <>

	dfa: make the executable a bit smaller
	* src/dfa.c (dfamust): Hoist MB_CUR_MAX calculation out of loops.

2015-10-13  Norihiro Tanaka  <>

	dfa: fix bug in alternate of sub-patterns that differ only in constraints
	Fix a bug where a line incorrectly matches alternates of sub-patterns
	that differ only in the constraints, e.g., the ERE '^a|a$'.
	Reported by Greg Boyd in:
	* src/dfa.c (dfamust): For a pattern with constraints, check that it is
	matched including the constraints, to judge whether it is exact.

	dfa: fix off-by-one error
	* src/dfa.c (dfamust): Fix off-by-one error in computing 'must' length,
	which caused the 'must' to be too short.  See:

2015-10-12  Jim Meyering  <>

	doc: NEWS: mention a bug fix
	* NEWS (Bug fixes): Describe it.
	This bug was introduced by commit v2.18-85-g2c94326
	and fixed by commit v2.21-51-g256a4b4.

2015-10-11  Paul Eggert  <>

	tests: add test case for Bug#21670
	* tests/options: Add test #4 to catch Bug#21670.
	Also, do not overescape # in shell strings.

2015-09-19  Paul Eggert  <>

	Add test for pop_fail_stack bug
	Problem reported by Hanno Böck in:
	If you use --with-included-regex the bug fix is in gnulib, here:
	If you use glibc, the bug fix has not been installed yet.
	* tests/ (XFAIL_TESTS): Add backref-alt if system matcher.
	(TESTS): Add backref-alt.
	* tests/backref-alt: New file.
	* tests/triple-backref: Remove unused var.
	Don't skip if tested with glibc, as now handles this.

	build: update gnulib submodule to latest

2015-08-19  Norihiro Tanaka  <>

	grep: avoid use of uninitialized variable
	EGexecute would use "backref" uninitialized.
	While that could have no bearing on correctness, it could
	impact performance, via an unnecessary use of regexp.
	* src/dfasearch.c (EGexecute): Initialize backref.
	Reported as
	Introduced by commit v2.21-55-gea0ebaa.

2015-08-12  Norihiro Tanaka  <>

	grep: remove fgrep code for case insensitive match
	The fgrep matcher is no longer called in case insensitive matching,
	so remove the code to support it.
	* src/kwsearch.c (mb_case_map_apply): Remove function.
	(Fexecute): Remove now-unused code.

2015-08-12  Paul Eggert  <>

	dfa: optimize [x-x]
	* src/dfa.c (parse_bracket_exp): Treat [x-x] as if it were [x].
	This also pacifies GCC, which otherwise complains about wc2
	being set but not used.

2015-08-12  Norihiro Tanaka  <>

	dfa: remove unused multibyte support
	Now regex should be used for range, collating element, equivalent class
	in non POSIX locales.  So remove code to support these features.
	* dfa.c (struct mb_char_classes): Remove members ch_classes,
	nch_classes, ranges, nranges, equivs, nequivs, coll_elems, ncoll_elems.
	All uses removed.
	(match_mb_charset): Remove function.

2015-08-01  Jim Meyering  <>

	tests: mb-non-UTF8-performance: use new function
	* tests/mb-non-UTF8-performance: Rewrite to use
	the user-time measuring function in init.cfg.

	tests: long-pattern-perf: measure user time, not elapsed
	Measuring user time makes this test less prone to false
	positive failure, and also lets us use a tighter bound.
	* tests/long-pattern-perf: Measure elapsed user time rather than
	wall-clock time, to permit a tighter bound on the ratio of
	N-to-10N timings.  Suggested by Giuseppe Ottaviano.
	Also, use regexps built from mostly 5-digit numbers, so that the 10:1
	ratio applies to lines of "seq" output as well as to total bytes.

	tests: new function to measure elapsed user time
	* tests/init.cfg (user_time_): New function.

2015-07-25  Norihiro Tanaka  <>

	dfa: remove word delimiter support for multibyte locales
	DFA supports word delimiter expressions, but it does not behave
	correctly for multibyte locales.  Even if it were to be fixed,
	the DFA matcher's performance would be no better than that of regex.
	Thus, this change removes DFA support for word delimiter expressions
	in multibyte locales.

	* src/dfa.c (dfa_supported): Return false also when a pattern uses any
	word delimiter expression in a multibyte locale.

2015-07-25  Norihiro Tanaka  <>

	dfa: avoid execution for a pattern including an unsupported expression
	If a pattern includes a construct unsupported by the DFA matcher,
	the DFA search would fail in most cases.  Make dfaexec immediately
	return for any such pattern.

	* src/dfa.c (struct dfa_state) [has_backref, has_mbcset]: Remove members
	and all uses.
	(dfaexec_main): Remove 'backref' parameter.  Update callers.
	(dfaexec_noop): New function.
	(dfa_supported): New function.
	(dfassbuild): Remove now-unused code.
	(dfacomp): When a pattern uses a DFA-unsupported construct, do not
	waste time performing any further analysis.

2015-07-19  Norihiro Tanaka  <>

	dfa: DEBUG: print detail of DFA states
	When compiled with -DDEBUG, grep outputs tokens etc.
	With this change, also print DFA states and transitions.
	This change is very useful when debugging those.

	* src/dfa.c (prtok) [DEBUG]: Change `%c' to `%02x' in printf format.
	(state_index) [DEBUG]: Print detail of new state.
	(dfastate) [DEBUG]: Print detail of DFA states.
	Reported as

2015-07-18  Norihiro Tanaka  <>

	tests: sjis-mb: accept two more locales
	* tests/sjis-mb: Accept the ja_JP.SJIS and ja_JP.PCK locales
	as well as ja_JP.SHIFT_JIS, so this test is less likely to
	be skipped unnecessarily.  Reported as

2015-07-18  Jim Meyering  <>

	tests: add a test for the performance fix
	* tests/long-pattern-perf: New file.
	* tests/ (TESTS): Add it.

2015-07-18  Norihiro Tanaka  <>

	dfa: speed up handling of long pattern
	DFA tries to find a long sequence of characters that must appear
	in any matching line.  However, when a pattern is long (length N),
	it is very slow, because it makes O(N^2) strstr calls.
	This change reduces that to O(N) by processing each sequence of
	adjacent "regular" characters as a group.

	Compare the run times of this command before and after this change:
	(on a i7-4770S CPU @ 3.10GHz using rawhide (~fedora 22) and compiled
	with gcc 6.0.0 20150627)
	  : | env time -f %e grep -f <(seq -s '' 9999)
	  Before: 0.85
	   After: 0.02

	* src/dfa.c (dfamust): Process each string of concatenated normal
	characters as a unit.
	* NEWS (Improvement): Mention it.
	Prompted by a bug report and patch by Ivan Yanikov

2015-07-17  Jim Meyering  <>

	tests: fix mis-applied patch.
	* tests/include-exclude: I applied "|sort" to the wrong creation
	of "out", and didn't push the same patch that I'd tested.

	tests: avoid FS-dependent false-positive failure
	* tests/include-exclude: Sort file name list, so that this test
	is not sensitive to the order in which those names are returned
	via readdir. I noticed the failure on a Fedora 21 system using ext4.
	Also fix a typo: s/framework_failure+/framework_failure_/

2015-07-13  Paul Eggert  <>

	grep: fix bug with --exclude-dir and command line
	Reported by Aron Griffis in:
	* NEWS: Document this.
	* src/grep.c (grepdirent): Don't check whether the file is skipped
	when on the command line, as that's the caller's responsibility.
	(main): Anchor the exclude patterns.
	* tests/include-exclude: Adjust test case to match fixed behavior.
	Add some more test cases.

	tests: fix $? typo in null-byte
	* tests/null-byte: Don't assume $? survives an invocation of 'test'.

2015-07-05  Jim Meyering  <>

	maint: dfa: used unsigned types where appropriate
	* src/dfa.c (case_folded_counterparts): Return unsigned int, not int.
	Change type of two locals to unsigned int, to reflect that their
	values are never negative.
	(parse_bracket_exp): Adjust type of result at each use, as well
	as that of related index variables.

2015-07-04  Norihiro Tanaka  <>

	dfa: build struct dfamust on demand
	If we won't use KWset, do not build a "struct dfamust".
	Now it is built only when needed.
	* src/dfa.c (struct dfa) [musts]: Remove member.
	(dfacomp): Don't build dfamust here.
	(dfamustfree): New function to free a struct dfamust.
	(dfamust): Make it a global function, and make it return a pointer
	to a malloc'd struct dfamust.
	(dfamusts): Remove it.
	* src/dfa.h (struct dfamust) [next]: Remove member.
	In the implementation preceding this patch, there was
	never more than one of these in a given "struct dfa".
	(dfamustfree, dfamust): Add prototypes.
	(dfamusts): Remove prototype.
	(dfaalloc): Declare with _GL_ATTRIBUTE_MALLOC.
	To make that symbol usable there, move the inclusion
	of "xalloc.h" from dfa.c to this file, dfa.h.
	* src/dfasearch.c (kwsmusts): Adapt to use the new interface.
	Update the comments to reflect reality.
	This addresses

2015-07-04  Paul Eggert  <>

	grep: use recent gnulib syntax bits
	* src/grep.c (Gcompile, Ecompile): Use plain RE_SYNTAX_GREP
	and RE_SYNTAX_EGREP, now that we assume a recent-enough gnulib.

	maint: ignore gendocs_template_min
	* doc/.gitignore: Add '/gendocs_template_min'.

	build: update gnulib submodule to latest

	dfa: '.' and '[^x]' now consistently match newline
	* src/dfa.c (parse_bracket_exp, lex, add_utf8_anychar)
	are about LF, not about eolbyte.  This patch does not affect
	'grep', but may affect other users of dfa.c.

	grep: -z '[^x]' now consistently matches newline
	Problem reported by Norihiro Tanaka in:
	* NEWS: Document this.
	* src/grep.c (Gcompile, Ecompile): Clear RE_HAT_LISTS_NOT_NEWLINE.
	* tests/utf8-bracket: Test this.

2015-07-03  Paul Eggert  <>

	grep: -z '.' now consistently matches newline
	Problem reported by Balazs Kezes in:
	* NEWS: Document this.
	* tests/utf8-bracket: New file, to test for this bug.
	* src/grep.c (Gcompile, Ecompile): Also specify RE_DOT_NEWLINE.
	* tests/ (TESTS): Add it.

	grep: simplify print_line_middle slightly
	* src/grep.c (print_line_middle): Simplify.

	grep: don't mishandle left context in -P
	* src/pcresearch.c (jit_exec): New arg SEARCH_OFFSET.
	Caller changed.
	(Pexecute): Pass the left context to pcre_exec, so that PCRE
	regular-expression matching can see it.
	* tests/pcre-context: New file, to test for this bug.
	* tests/ (TESTS): Add it.

2015-06-28  Jim Meyering  <>

	tests/case-fold-backref: factor test

2015-06-26  Paul Eggert  <>

	grep: don't hang on command-line fifo if -D skip
	* NEWS: Document this.
	* src/grep.c (skip_devices):
	New function, with code taken from grepdirent.
	(grepdirent): Use it.  Avoid an unnecessary initialization.
	(grepfile): If skipping devices, open files with O_NONBLOCK.
	Throw in O_NOCTTY while we're at it.
	(grepdesc): Skip devices here, too.  Not only does this fix the
	bug, it fixes an unlikely race condition if some other process
	renames a device between fstatat and openat.
	* tests/skip-device: Add a test for this bug.

	grep: minor tweaks
	* src/grep.c (main): Change recently-added static vars to be
	constants, which makes them sharable.  Prefer 'return' to 'exit'
	when returning/exiting from 'main'.  Move decl closer to first use
	and rename local from 'ok' (which was confusing) to 'status'.
	Prefer named constant STDOUT_FILENO to unnamed constant 1.

2015-06-26  Jim Meyering  <>

	maint: unify three argv-processing calls
	* src/grep.c (main): Unify three calls to grep_commandline_arg.

	maint: alphabetize anonymous enum member names

2015-05-30  Paul Eggert  <>

	test: tighten tests for bracket exprs
	* tests/posix-bracket: Test '[a-a[.-.]--]'.
	Also, test that failures are with status 1
	(nonmatching data), not status 2 (invalid expressions).

2015-04-26  Jim Meyering  <>

	maint: update bootstrap from gnulib
	* bootstrap: Update from gnulib.

	maint: reword a diagnostic not to trigger leading capital check
	* src/pcresearch.c: Reword diagnostic to avoid "make syntax-check"

	maint: sort test names in tests/ and add syntax-check rule
	* (sc_sorted_tests): New rule.
	* tests/ (TESTS): Alphabetize.

2015-04-25  Norihiro Tanaka  <>

	dfa: make find_pred return NULL for an invalid predicate
	This could never happen when invoked via grep, but could have triggered
	a bug if dfa.c's find_pred function were invoked by some other program.
	* src/dfa.c (find_pred): Return NULL for an invalid predicate.
	* tests/invalid-char-class: New file to test for this.
	* tests/ (TESTS): Add that new file name to the list.
	This addresses

2015-04-06  Paul Eggert  <>

	build: improve pkg-config doc and error handling
	Error-handling improvement suggested by Mike Frysinger in:
	* NEWS: Document pkg-config changes.
	* README-prereq: pkg-config is now a prereq when building from
	* m4/pcre.m4 (gl_FUNC_PCRE): Report an error if pcre is explicitly
	requested but not available.  Defer to user-supplied PCRE_CFLAGS

	build: remove typo and don't bother with /usr/include/pcre
	Problem reported by Holger Bruenjes.
	* m4/pcre.m4: Remove test for /usr/include/libpng (a typo).
	Come to think of it, don't bother worrying about
	/usr/include/pcre, as hosts with that problem can use pkg-config
	or configure with CFLAGS by hand.

	build: use pkg-config (if available) to configure libpcre
	Problem reported by Mike Frysinger in:
	* bootstrap.conf (bootstrap_post_import_hook):
	Copy pkg-config's pkg.m4.
	* m4/pcre.m4 (gl_FUNC_PCRE): Rewrite to use pkg-config if
	available, and to test that pcre_compile can be linked to.
	(grep_LDADD): Add PCRE_LIBS.
	* src/pcresearch.c: Simply include <pcre.h> if HAVE_LIBPCRE,
	since 'configure' arranges for the appropriate -I option now.

2015-03-11  Paul Eggert  <>

	grep: output "." file name in diagnostic
	This is bug C as reported by David Grayson in:
	This bug occurs only in obscure circumstances, and I didn't see
	how to write a reasonable test case for it.
	* src/grep.c (filename_prefix_len): Remove, replacing with ...
	(omit_dot_slash): New static var.  All uses of the former replaced
	with uses of the latter.
	(grepdirent): Don't add 2 if the filename is just ".".

	egrep, fgrep: just use what's in PATH
	* src/ Don't monkey with PATH; just use whatever 'grep'
	is in the path.  This is simpler, and lets the user specify
	default options with a script for only grep, with no need for
	egrep and fgrep scripts.
	Fixes: bug#19998

	doc: give a script wrapper example
	* doc/grep.texi (Environment Variables): Give an example of a
	wrapper script, as an alternative to using GREP_OPTIONS.
	Fixes: bug#19998

	doc: clarify how -a matches
	* doc/, doc/grep.texi (File and Directory Selection):
	Give an example of how non-text bytes affect pattern matching in
	binary files.
	Fixes: bug#20080

2015-02-23  Paul Eggert  <>

	Cover the non-INSTALL case
	* README: Mention what to do if there is no INSTALL file.
	Fixes: bug#19928

2015-02-11  Jim Meyering  <>

	maint: use ASAN-poisoning more carefully
	The ASAN-poisoning instituted by commit v2.21-14-g1555185 was
	incomplete, since the poisoned tail of the read buffer could well
	be the target of a legitimate follow-on read.  To accommodate that,
	we must unpoison each such region just before beginning fillbuf's
	read loop.
	* src/grep.c [HAVE_ASAN] (asan_poison): Define.
	(clear_asan_poison): Define.
	(fillbuf): Clear before reading, since we are likely to read
	into memory that was poisoned on the preceding iteration.
	* tests/two-files: New file, to test for this.
	* tests/ (TESTS): Add it.

2015-02-10  Paul Eggert  <>

	Grow the JIT stack if it becomes exhausted
	Problem reported by Oliver Freyermuth in:
	* NEWS: Document the fix.
	* tests/ (TESTS): Add pcre-jitstack.
	* tests/pcre-jitstack: New file.
	* src/pcresearch.c (NSUB): Move decl earlier, since it's needed
	earlier now.
	(jit_stack_size) [PCRE_STUDY_JIT_COMPILE]: New static var.
	(jit_exec): New function.
	(Pcompile): Initialize jit_stack_size.
	(Pexecute): Use new jit_exec function.  Report a useful diagnostic

2015-02-01  Jim Meyering  <>

	maint: reference CVE-2015-1345 from NEWS
	* NEWS: Mention the CVE that was addressed by v2.21-13-g83a95bd,
	"grep -F: fix a heap buffer (read) overrun".

2015-01-18  Jim Meyering  <>

	maint: convert "goto" to "continue" and remove now-spurious label
	* src/kwset.c (bmexec_trans): Using "goto big_advance" here is
	equivalent to using "continue".  Make that change and remove
	the now-unused label.

2015-01-10  Jim Meyering  <>

	tests: add support for ASAN memory poisoning
	This lets us reliably detect with ASAN some UMR bugs
	that would otherwise be detectable only some of the time
	with MSAN.  Use __asan_poison_memory_region to mark the unused
	portion of a read buffer as inaccessible.  Then, with ASAN,
	any attempt to access those bytes results in an ASAN abort.
	* src/system.h: Include "ignore-value.h".
	(__has_feature): Define.
	(HAVE_ASAN): Define when address sanitizer is enabled.
	[HAVE_ASAN]: Declare these two __asan_* symbols.
	[!HAVE_ASAN] (__asan_poison_memory_region): Define stub.
	[!HAVE_ASAN] (__asan_unpoison_memory_region): Likewise.
	* src/grep.c: Use __asan_poison_memory_region.

2015-01-09  Yuliy Pisetsky  <>

	grep -F: fix a heap buffer (read) overrun
	grep's read buffer is often filled to its full size, except when
	reading the final buffer of a file.  In that case, the number of
	bytes read may be far less than the size of the buffer.  However, for
	certain unusual pattern/text combinations, grep -F would mistakenly
	examine bytes in that uninitialized region of memory when searching
	for a match.  With carefully chosen inputs, one can cause grep -F to
	read beyond the end of that buffer altogether.  This problem arose via
	commit v2.18-90-g73893ff with the introduction of a more efficient
	heuristic using what is now the memchr_kwset function. The use of
	that function in bmexec_trans could leave TP much larger than EP,
	and the subsequent call to bm_delta2_search would mistakenly access
	beyond end of the main input read buffer.

	* src/kwset.c (bmexec_trans): When TP reaches or exceeds EP,
	do not call bm_delta2_search.
	* tests/kwset-abuse: New file.
	* tests/ (TESTS): Add it.
	* Update.
	* NEWS (Bug fixes): Mention it.

	Prior to this patch, this command would trigger a UMR:

	  printf %0360db 0 | valgrind src/grep -F $(printf %019dXb 0)

	  Use of uninitialised value of size 8
	     at 0x4142BE: bmexec_trans (kwset.c:657)
	     by 0x4143CA: bmexec (kwset.c:678)
	     by 0x414973: kwsexec (kwset.c:848)
	     by 0x414DC4: Fexecute (kwsearch.c:128)
	     by 0x404E2E: grepbuf (grep.c:1238)
	     by 0x4054BF: grep (grep.c:1417)
	     by 0x405CEB: grepdesc (grep.c:1645)
	     by 0x405EC1: grep_command_line_arg (grep.c:1692)
	     by 0x4077D4: main (grep.c:2570)

	See the accompanying test for how to trigger the heap buffer overrun.

	Thanks to Nima Aghdaii for testing and finding numerous
	ways to break early iterations of this patch.

2015-01-08  Jim Meyering  <>

	grep: avoid false-positive UMR
	For some inputs, valgrind would report an uninitialized
	memory read error, but it was harmless.
	* src/grep.c (fillbuf): Initialize those trailing bytes.

2015-01-01  Jim Meyering  <>

	gnulib: update to latest

	maint: update copyright year ranges to include 2015
	Run "make update-copyright".  Also, ...
	* grep.texi: Update manually, converting each "--" to "-".

2014-12-15  Paul Eggert  <>

	doc: document binary-data heuristic better
	Problem reported by Martin Hoch in:
	* doc/grep.texi (File and Directory Selection):
	Document what non-text bytes are.
	(Usage): Fix cross reference.

2014-12-12  Jim Meyering  <>

	maint: fix a new "make syntax-check" failure
	* tests/dfa-match-aux.c: s/can not/cannot/

2014-12-12  Norihiro Tanaka  <>

	build: avoid build failure with --enable-gcc-warnings and no PCRE
	* src/pcresearch.c [HAVE_LIBPCRE] (empty_match): Guard the declaration
	of this PCRE-only variable.

2014-12-07  Paul Eggert  <>

	tests: port fmbtest to CentOS 6 and earlier
	* tests/fmbtest: Port to platforms where the 'sed' pattern
	'[^0-9]' does not match every non-digit character.  Problem
	reported by Norihiro Tanaka in:

2014-12-06  Norihiro Tanaka  <>

	dfa: simplify dfaexec
	* src/dfa.c (dfaexec): Simplify by rearrangement of IF conditions.
	This commit induces no semantic change, and reverts part of commit

2014-12-06  Norihiro Tanaka  <>

	dfa: avoid invalid match or infinite loop in unused matching mode
	Neither grep nor gawk uses this DFA code in its matching mode,
	since each always calls dfacomp with a nonzero final argument.
	However, when used in that mode, it had bug:
	After failing to match in matching mode, it should return NULL,
	but instead would either report a false match or enter an
	infinite loop.

	* src/dfa.c (dfaexec_main): After failing to match in matching mode
	return NULL, rather than transitioning to the next state.
	* tests/dfa-match: Add a new test.
	* tests/dfa-match-aux.c: Add a new program to exercise this
	otherwise-unused part of dfa.c.
	* tests/ Add a rule to build new test.
	(check_PROGRAMS): Add dfa-match-aux.
	(AM_CPPFLAGS): Add -I$(top_srcdir)/src.
	(TESTS): Add dfa-match.
	* (exclude_file_name_regexp--sc_bindtextdomain):
	Exempt the new test file from some syntax-check rules.

2014-12-04  Santiago Ruano Rincón  <>

	doc: document grep-2.11 change in behavior of -r, --recursive
	* doc/grep.texi (--recursive, -r): Mention the new behavior
	of recursively searching "." when there is no FILE argument.
	* doc/ Likewise.
	That change first appeared in grep-2.11, released on 2012-03-02.

2014-11-24  Jim Meyering  <>

	maint: correct for four Author: name misspellings
	* .mailmap: Correct for misspelling in Norihiro Tanaka's last name
	as listed in four commit Author: fields: s/Norihirio/Norihiro/

2014-11-23  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.21
	* NEWS: Record release date.

2014-11-21  Jim Meyering  <>

	tests: sjis-mb: remove now-obsolete and failing sub-tests
	* tests/sjis-mb: Commit v2.18-123-geb3292b changed how grep
	handles patterns with encoding errors.  These SJIS tests are
	skipped so often that we didn't notice until now that there were
	two tests of that changed behavior, and that on any system with
	the ja_JP.SHIFT_JIS locale, they would always fail.  Remove those
	two tests, since this functionality is well tested separately,
	via tests/prefix-of-multibyte.

2014-11-20  Norihiro Tanaka  <>

	grep -F could erroneously fail to match in non-UTF8 multibyte locales
	This fixes a bug that can strike only when using a non-UTF8 multibyte
	locale like ja_JP.SHIFT_JIS.

	Consider this example: it would mistakenly fail to match before
	this patch:

	  printf '\203AA\n'|LC_ALL=ja_JP.SHIFT_JIS src/grep -F A

	When searching for a single byte that happens to be the latter
	byte of a multibyte character, and the target byte also follows
	that multibyte character, grep -F would advance an internal pointer
	by one byte too many, thus missing the target byte.  A test case
	for this bug is already included in tests/sjis-mb.

	* src/kwsearch.c (Fexecute): Skip one byte less, after matched middle of a
	multi-byte character.  Introduced by commit v2.18-119-gfb7d538.

2014-11-17  Jim Meyering  <>

	tests: big-match: disable OOM-provoking subtest
	* tests/big-match: Our application of this regexp '^.*x\(\)\1'
	to a file containing a single matching line of length 2GiB+2
	would cause inordinate memory consumption (over 100GB) via
	regexec.c, but no leak.  That would cause disruption on most
	systems, so remove this subtest.  Reported by Assaf Gordon.

2014-11-16  Norihiro Tanaka  <>

	dfa: avoid undefined behavior
	* src/dfa.c (dfassbuild): Don't call memcpy with a second
	argument of NULL, even when the size (3rd argument) is 0.

2014-11-14  Jim Meyering  <>

	gnulib: update to latest

2014-11-14  Norihiro Tanaka  <>

	grep -F -x -o PAT would print an extra newline for each match
	* src/kwsearch.c (Fexecute): Correctly compute the length of a match
	by subtracting 2 (not 1) when match_lines is set.  With -x, we augment
	the "line" by both prepending and appending an EOLBYTE to the search
	pattern.  Here, we must correct for that.  However, to compensate,
	when we are using -x (--line-regexp) and start_ptr is NULL, we have
	to add 1 to the length so that we still print the trailing EOLBYTE.
	Introduced by commit v2.18-85-g2c94326.
	* tests/match-lines: Add a new test.
	* tests/ (TESTS): Add it.
	* NEWS (Bug fixes): Mention it.

2014-11-11  Paul Eggert  <>

	tests: port to Darwin
	The 'sed' command 's/.//' does not delete all bytes in the C locale.
	Problem reported by Nelson H. F. Beebe.
	* tests/fmbtest: Don't assume that sed treats bytes with the
	top bit set as valid characters in the C locale, as this is not
	true for Darwin.  Use the cs_CZ.UTF-8 locale instead, and
	simplify the sed script.

	tests: fix recently-introduced stray output
	* tests/init.cfg (require_pcre_): Remove stray debugging output.

	build: port to GCC 4.6.4 + glibc 2.5
	On platforms this old, building with _FORTIFY_SOURCE equal to 2
	results in duplicate definitions of standard library functions.
	Problem reported by Nelson H. F. Beebe.
	By default, do not enable this unless GNULIB_PORTCHECK is defined.
	This better matches the original intent, which as I recall was to
	enable these extra checks only with --enable-gcc-warnings.

	tests: port to libpcre sans UTF-8 support
	Problem reported by Nelson H. F. Beebe.
	* tests/pcre-infloop, tests/pcre-invalid-utf8-input, tests/pcre-utf8:
	Skip the test unless PCRE works in an en_US.UTF-8 locale.

2014-11-09  Jim Meyering  <>

	tests: do not fail when the zh_CN.UTF-8 locale is not installed
	* tests/word-multibyte: This test would fail on a system with
	no zh_CN.UTF-8 locale. Use it only if it is installed.

	tests: avoid hex_printf_ portability problems
	* tests/init.cfg (hex_printf_): Spell out a-f and A-F, for
	non-C locales, ensure that the input to sed is newline-terminated,
	and quote the final octal format string.
	Suggestions from Paul Eggert.

2014-11-08  Jim Meyering  <>

	tests: avoid a multibyte tr portability problem
	* tests/init.cfg (tr): New wrapper function.
	See comments for details.  Reported by Norihiro Tanaka

	maint: remove spurious LC_ALL setting from one test
	* tests/word-multibyte: Remove unnecessary setting of LC_ALL.

	tests: fix typo in previous change
	* tests/init.cfg (hex_printf_): Fix typo s/A-f/A-F/.
	For the record, I introduced that error, not Norihiro.

2014-11-08  Norihiro Tanaka  <>

	tests: avoid awk+printf+\xHH portability trap
	* tests/init.cfg (hex_printf_): Rewrite in terms of printf and sed.
	Using awk's printf with \xHH in the format string was not portable
	to the awk of Solaris 10, AIX 7 or HP-UX 11.23, as reported in
	* tests/word-multibyte: Use printf rather than hex_printf_,
	and give the character we're printing a name: e_acute (rather
	than A-grave), since that is used in other tests.
	a trailing \n in the format string, adjust by removing it, and
	instead invoking echo.
	* tests/multibyte-white-space: Simply remove each trailing \n.
	They were not needed.

2014-11-07  Jim Meyering  <>

	tests: avoid printf+\xHH portability trap
	* tests/word-multibyte: Using the bourne shell's printf function
	with strings like "\xHH\xHH" happens to work for most interactive
	shells, but not for dash.  That is not portable.  Use our hex_printf_
	awk wrapper instead. Without this change, this test would fail on
	a Debian system for which /bin/sh is configured to be "dash".

	maint: move helper function, hex_printf to init.cfg
	* tests/init.cfg (hex_printf_): New function, from ...
	* tests/multibyte-white-space: ... here.  Reflect the
	s/hex_print/hex_printf_/ renaming.

2014-11-02  Paul Eggert  <>

	grep: port O_NOFOLLOW errno checking to NetBSD
	Problem reported by Assaf Gordon in:
	* NEWS: Document it.
	* src/grep.c (open_symlink_nofollow_error):
	New function, which does the right thing on NetBSD.
	(grepfile): Use it.

2014-10-31  Jim Meyering  <>

	build: generate man pages even when existing targets are read-only
	* doc/ (grep.1): Use mv -f to move temporary to target,
	in case the target is read-only.  Also, always make the generated
	files read-only.
	(egrep.1 fgrep.1): Likewise.
	This avoids a build failure reported by Eric Blake in

2014-10-30  Jim Meyering  <>

	tests: avoid false-positive failure due to some zh_CN.* locales
	On some systems, and for some zh_CN.* locales (e.g., OpenBSD5.5) the
	E-acute pair of bytes do not qualify as a word-constituent character.
	* tests/word-multibyte: Use zh_CN.UTF-8, rather than "zh_CN".
	Reported by Assaf Gordon and Bruce Dubbs in

2014-10-29  Jim Meyering  <>

	gnulib: update to latest; bootstrap, too
	* gnulib: Update to latest.
	* bootstrap: Copy latest from gnulib.

2014-10-28  Jim Meyering  <>

	tests: make new test script executable
	* tests/word-multibyte: Make this file executable.

2014-10-28  Norihiro Tanaka  <>

	dfa: make \w and \W work in multibyte locales
	Reported by Jaroslav Skarvada in:
	Now, \w and \W are supported in not only single byte locale but multibyte

	* src/dfa.c (PUSH_LEX_STATE, POP_LEX_STATE): Move definitions "up",
	so they are not within the function.
	(lex): Make \w and \W work in a multibyte locale, the same way
	we made \s and \S work.
	* tests/word-multibyte: New test for this change.
	* tests/ Add a rule to build new test.
	* NEWS (Bug fixes): Mention it.

2014-10-26  Norihiro Tanaka  <>

	dfa: avoid false match in a non-UTF8 multibyte locale
	This command should print nothing:

	  printf '\263\244\263\244\n' \
	    | LC_ALL=ja_JP.eucJP grep -E "$(printf '^x|\244\263')"

	Before this patch, it would print its sole input line.
	* src/dfa.c (struct dfa): Add new members: min_trcount,
	initstate_letter, initstate_others.
	(dfaanalyze): Build states with not only a newline context but others.
	(build_state): Don't release initial states.
	(skip_remains_mb): Add a parameter.
	Add a comment describing all parameters.
	(dfaexec_main): When there are multiple start states, we are about
	to transition from one state to another and the current byte is not
	the first byte of a multibyte character, first advance past the
	current multibyte character.
	* tests/euc-mb: Add a new test.
	* NEWS (Bug fixes): Mention it.
	This addresses

2014-10-25  Paul Eggert  <>

	tests: work around older libpcre bugs when testing -P and UTF-8
	* tests/pcre-invalid-utf8-input: Add require_timeout_ and
	require_compiled_in_MB_support.  Put a timeout of 3 seconds on
	grep, to avoid having this test case loop forever with older
	versions of libpcre, such as those found on RHEL 6.5.
	Reported by Jim Meyering in:

2014-10-24  Norihiro Tanaka  <>

	tests: add test for grep -P fix
	* tests/pcre-o: New test for this change.
	* tests/ (TESTS): Add it.

2014-10-24  Paul Eggert  <>

	grep: fix grep -P crash
	Reported by Shlomi Fish in:
	Commit 9fa500407137f49f6edc3c6b4ee6c7096f0190c5 (2014-09-16) is a
	hack that I put in to speed up 'grep -P'.  Unfortunately, not only
	is it violation of modularity, it's also a bug magnet, as we have
	found out with Bug#18738 and Bug#18806.  Remove the optimization
	instead of applying more bandaids.  Perhaps we can think of a
	better way of doing the optimization, or perhaps we can just live
	with a slower grep -P (as -P is inherently slower anyway...).
	* src/grep.c, src/grep.h (validated_boundary):
	Remove.  All uses removed.
	* src/pcresearch.c (Pexecute): Do not worry about validated_boundary.

2014-10-19  Norihiro Tanaka  <>

	dfa: remove two erroneous clauses from a now-unused function
	RE_DOT_NEWLINE and RE_DOT_NOT_NULL apply only to a dot that
	matches any character.  Do not consider them when matching
	with a bracket expression.

	* src/dfa.c (match_mb_charset): Remove tests for RE_DOT_NEWLINE

2014-10-19  Norihiro Tanaka  <>

	dfa: process all MBCSET constructs via glibc's matcher
	The DFA matcher does not support collating symbols or equivalence
	classes, so ensure that any MBCSET reference is handled by the glibc
	matcher.  dfa.c already handled this in one case, but not the other,
	so that a command like "printf '\0' |src/grep -aE '^\s?$'" would
	mistakenly end up using dfa.c's match_mb_charset function rather
	than glibc's matcher.

	* src/dfa.c (dfaexec_main): Move that code into the
	State_transition macro.  This renders the match_mb_charset
	unused by grep.
	* tests/multibyte-white-space: Add a test to exercise the
	just-rendered-inaccessible code path.

2014-10-15  Norihiro Tanaka  <>

	grep: initialize validation_boundary properly before use
	* src/grep.c (main): Initialize validation_boundary before pre-searching
	for an empty line.

2014-10-15  Paul Eggert  <>

	grep: fix off-by-one bug in -P optimization
	Reported by Norihiro Tanaka in:
	* src/pcresearch.c (Pexecute): Fix off-by-one bug with
	* tests/init.cfg (envvar_check_fail): Catch off-by-one bug.

2014-10-08  Norihiro Tanaka  <>

	dfa: fix a theoretical bug
	* src/dfa.c (dfaexec_main): After searching for a match from
	the initial state, set the previous state, S1, to 0.
	So far, we have found no case in which this fix makes a difference.

2014-10-07  Paul Eggert  <>

	doc: modernize and simplify man page
	* doc/ (Tx, Id): Remove.  All uses removed.
	(MTO, URL): New macros, used for email and URL.
	Use them when appropriate.
	In main text, omit chatty discussions of other implementations;
	the full manual suffices for this sort of thing.

	doc: clarify exit status
	Reported by Santiago Ruano Rincón in:
	* doc/ (EXIT STATUS):
	* doc/grep.texi (Exit Status): Clarify.

2014-10-07  Norihiro Tanaka  <>

	dfa: test for just-fixed bug
	* tests/mb-dot-newline: New file.
	* tests/ (TESTS): Add it.
	* NEWS (Bug fixes): Mention it.
	Bisection suggests that the bug was introduced by
	commit v2.18-123-geb3292b.  Also see;bug=18580

2014-10-05  Norihiro Tanaka  <>

	dfa: factor out a new nontrivial block of duplicated code
	* src/dfa.c (State_transition): New macro.
	(dfaexec_main): Use it twice.

	dfa: check end of input buffer after transition in non-UTF8 multibyte locale
	* src/dfa.c (dfaexec_main): Check for end of input buffer after each
	transition in a non-UTF8 multibyte locale.
	* tests/mb-non-UTF8-overrun: New test.
	* tests/ (TESTS): Add it.
	* src/grep.c (main): With this fix, we no longer need the fourth
	byte of "eolbytes".

2014-10-04  Jim Meyering  <>

	grep: avoid stack buffer read-underrun and overrun
	Testing binaries built with -fsanitize=address caused aborts due
	to stack underrun and overrun.
	* src/grep.c (main): Allocate a larger buffer for eolbytes:
	one byte before the beginning and one more after the end.
	For details, see

2014-10-04  Norihiro Tanaka  <>

	grep: fix subscript error when testing whether empty lines match
	src/grep.c (grep): When testing whether an empty line matches,
	make the input buffer one byte longer, as dfaexec uses that
	for a sentinel.

2014-09-27  Paul Eggert  <>

	dfa: minor tweaks, mostly to remove __attribute__ ((noinline))
	That attribute isn't portable, and I found a way to get similar
	performance with standard C features.
	* NEWS: Document the recently-installed performance improvement.
	* src/dfa.c (struct dfa): New member dfaexec.
	(dfaexec_main): Remove unnecessary 'const'.
	(dfaexec_mb, dfaexec_sb): Remove __attribute__ ((noinline));
	no longer needed.
	(dfaexec): Use new dfaexec member.
	(dfainit, dfaoptimize, dfassbuild): Initialize it.

2014-09-27  Norihiro Tanaka  <>

	dfa: separate dfaexec function to help optimization by compiler
	* src/dfa.c (dfaexec_main): Rename from dfaexec, add inline attribute.
	(dfaexec_mb): New function.  Run it when d->multibyte is true.   For this
	function inlination must be avoided.
	(dfaexec_sb): New function.  Run it when d->multibyte is false.  For this
	function inlination must be avoided.
	(dfaexec): Call dfaexec_mb or dfaexec_sb accoding to d->multibyte.

2014-09-27  Norihiro Tanaka  <>

	dfa: speed-up at initial state
	DFA state is always 0 until have found potential match.  So we improve
	matching there by continuing to use the transition table.

	* src/dfa.c (skip_remains_mb): New function.
	(dfaexec): Speed-up at initial state.

2014-09-27  Paul Eggert  <>

	maint: generalize the -Wcast-align fix
	* src/grep.c (CAST_ALIGNED): New macro.
	(skip_easy_bytes): Use it.

2014-09-27  Jim Meyering  <>

	maint: suppress a false-positive -Wcast-align warning
	Building with --enable-gcc-warnings and gcc-4.9.1 would provoke this:
	  grep.c:499:12: error: cast from 'const char *' to 'const uword *'\
	      (aka 'const unsigned long *') increases required alignment from\
	      1 to 8 [-Werror,-Wcast-align]
	    for (s = (uword const *) p; ! (*s & hibyte_mask); s++)
	* src/grep.c (skip_easy_bytes): Use a pragma to suppress
	gcc's false-positive cast-alignment warning.

2014-09-26  Paul Eggert  <>

	grep: don't check extensively for invalid prefix bytes unless -P
	Problem reported by Jim Meyering in:
	* src/grep.c (grep): After the first buffer is checked, leave the
	file-type checker in TEXTBIN_UNKNOWN state only when -P is used.
	Only the -P matcher has performance problems with checking binary
	data that make it worthwhile to check every prefix input byte so
	the -P matcher's TEXTBIN_UNKNOWN optimizations can come into play.
	Other matchers can simply check the data directly, and using
	TEXTBIN_UNKNOWN with them slows 'grep' down for no benefit.

	grep: scan for valid multibyte strings more quickly
	Scan valid multibyte strings more quickly in the common case of
	encodings that are upward compatible with ASCII, such as UTF-8.
	You'd think there'd be a fast standard way to do this nowadays,
	but nooooo....
	Problem reported by Jim Meyering in:
	* src/grep.c (HIBYTE): New constant.
	(easy_encoding): New static var.
	(init_easy_encoding, skip_easy_bytes): New functions.
	(uword): New type.
	(buffer_textbin): Skip easy bytes quickly.
	Don't bother with mb_clen here, since skip_easy_bytes typically
	captures the easy cases; just use mbrlen directly.
	(buffer_textbin, file_textbin): First arg is no longer a const
	pointer, since the byte past the end is now an overwritten sentinel.
	(fillbuf): Make room for a uword after the buffer, for skip_easy_bytes.
	(main): Call init_easy_encoding.

2014-09-17  Paul Eggert  <>

	grep: speed up processing of holes before EOF on Solaris
	* src/grep.c (fillbuf): If SEEK_DATA fails with errno == ENXIO,
	skip over the hole at EOF.

	grep: port to platforms lacking SEEK_DATA
	Reported by Norihiro Tanaka in:
	* src/grep.c (SEEK_DATA): Default to SEEK_SET if not defined.
	(SEEK_HOLE): Move to top level, and default it to SEEK_SET.
	(file_textbin): Adjust to new default.
	(fillbuf): Don't bother with SEEK_DATA if it defaults to SEEK_SET.

	grep: skip past holes efficiently
	Take advantage of the relaxed rules for treating non-text bytes in
	binary data, by efficiently skipping past holes on platforms
	supporting lseek's SEEK_DATA flag.
	On one test on a circa-2008 Sun Fire V40z running Solaris 11.2,
	'grep x' took 0.009 real-time seconds to scan a holey file of size
	9,223,372,036,854,775,802 bytes, for a nominal scan rate of 1 ZB/s.
	grep 2.20's scan rate on this platform was 843 MB/s, so this is a
	speedup by a factor of 1.2 trillion.  The speedup factor is not
	as great on GNU/Linux hosts, due to what appear to be SEEK_DATA
	inefficiencies, but presumably this will be cleared up in time.
	* NEWS: Document this.
	* src/grep.c, src/grep.h (eolbyte): Now char, not unsigned char.
	This is for compatibility with the rest of the code.
	The old (performance?) reasons for 'unsigned char' are now moot.
	* src/grep.c (skip_nuls, skip_empty_lines, seek_data_failed):
	New static vars.
	(totalnl): Move up, since it's about input, not output, and
	fillbuf now uses it.
	(add_count): Move up, since fillbuf now uses it.
	(all_zeros): New function.
	(fillbuf): Use SEEK_DATA to skip past holes efficiently,
	on systems that support this.
	(grep, main): Set the new static vars.

	grep: improve -P performance in typical cases
	* src/grep.c, src/grep.h (enum textbin): Move to grep.h.
	(input_textbin, validated_boundary): New vars.
	* src/grep.c (grepbuf, grep): Initialize them.
	* src/pcresearch.c (Pexecute): Do a multiline search
	when the input is known to be free of encoding errors.
	Quickly discard bytes that are obviously encoding errors.
	Quickly match empty strings.

	grep: minor -P speedup with jit_stack
	* src/pcresearch.c (jit_stack): No longer static.

	grep: non-text bytes in binary data may be treated as line ends
	* NEWS, doc/grep.texi (File and Directory Selection):
	Document this change.
	* src/grep.c (zap_nuls): New function.
	(grep): Use it.
	* tests/null-byte: Relax to allow new behavior.

	grep: -z no longer considers '\200' to be binary data
	This avoids a problem when using grep -z in a Windows-1252 locale.
	Plus, it lets 'grep -z' run a bit faster.
	* NEWS: Document this.
	* src/grep.c (buffer_textbin): Don't look for '\200' if -z.
	* tests/pcre-z: Test for new behavior.

	grep: refactor binary-vs-unknown-vs-text flags for clarity
	* src/grep.c (enum textbin): New enum.
	(textbin_is_binary): New function.
	(buffer_textbin, file_textbin, grep): Use them, for clarity.

2014-09-16  Paul Eggert  <>

	grep: fix -P speedup bug with empty match
	* src/pcresearch.c (NSUB): New top-level constant, replacing
	'nsub' within Pexecute.
	(Pcompile, Pexecute): Use it.
	(Pexecute): Don't assume sub[1] is zero after a PCRE_ERROR_BADUTF8
	match failure.
	* tests/pcre-invalid-utf8-input: Test for this bug.

	grep: port -P speedup to hosts lacking PCRE_STUDY_JIT_COMPILE
	* src/pcresearch.c (Pcompile): Do not assume that
	(empty_match): Define on all platforms.

	grep: use mbclen cache in one more place
	* src/grep.c (fgrep_to_grep_pattern): Use mb_clen here, too.

	grep: avoid false alarms for mb_clen and to_uchar
	* (_gl_TS_unmarked_extern_functions): New var,
	to bypass the tight_scope false alarms on mb_clen and to_uchar.

	grep: use mbclen cache more effectively
	* src/grep.c (buffer_textbin, contains_encoding_error):
	Use mb_clen for speed.
	(buffer_textbin): Bypass mb_clen in unibyte locales.
	(main): Always initialize the cache, since it's sometimes used in
	unibyte locales now.  Initialize it before contains_encoding_error
	might be called.
	* src/search.h (SEARCH_INLINE): New macro.
	(mbclen_cache): Now extern decl.
	(mb_clen): New inline function.
	* src/searchutils.c (SEARCH_INLINE, SYSTEM_INLINE): Define.
	(mbclen_cache): Now extern.
	(build_mbclen_cache): Put 1 into the cache when mbrlen returns 0.
	(mb_goback): Use mb_len for speed, and rely on it returning nonzero.
	* src/system.h (SYSTEM_INLINE): New macro.
	(to_uchar): Use it.

	grep: improve performance for older glibc
	glibc has a bug where mbrlen and mbrtowc mishandle length-0 inputs.
	Working around it in gnulib slows grep down, so disable the tests for it
	and make sure grep works even if the bug is present.
	* bootstrap.conf (avoided_gnulib_modules): Add mbrtowc-tests.
	* (gl_cv_func_mbrtowc_empty_input): Assume yes.
	* src/searchutils.c (mb_next_wc): Don't invoke mbrtowc on empty input.

	grep: treat a file as binary if its prefix contains encoding errors
	* NEWS:
	* doc/grep.texi (File and Directory Selection):
	Document this.
	* src/grep.c (buffer_encoding, buffer_textbin): New functions.
	(file_textbin): Rename from file_is_binary.  Now returns 3-way value.
	All callers changed.
	(file_textbin, grep): Check the input more carefully for text vs
	binary data.
	(contains_encoding_error): Remove; use replaced by buffer_encoding.
	* tests/backref-multibyte-slow:
	* tests/high-bit-range:
	* tests/invalid-multibyte-infloop:
	Use -a, since the input is now considered to be binary.
	* tests/invalid-multibyte-infloop: Add a check for new behavior.

	grep: use bool for boolean in grep.c
	* src/grep.c (show_version, suppress_errors, only_matching)
	(align_tabs, match_icase, match_words, match_lines, errseen)
	(write_error_seen, is_device_mode, usable_st_size)
	(file_is_binary, skipped_file, reset, fillbuf, out_quiet)
	(out_line, out_byte, count_matches, no_filenames, line_buffered)
	(done_on_match, exit_on_match, print_line_head, prline, grep)
	(grepdirent, grepfile, grepdesc, grep_command_line_arg)
	(get_nondigit_option, main): Use bool for boolean.
	(print_line_head, prline): Use char for byte.
	* src/grep.h: Include <stdbool.h>, and adjust decls to match
	changes in grep.c.

	grep: speed up -P on files containing many multibyte errors
	* src/pcresearch.c (empty_match): New var.
	(Pcompile): Set it.
	(Pexecute): Use it.

	grep: remove/refactor unnecessary code about line splitting
	* src/grep.c (do_execute): Remove.  Caller now uses 'execute'.
	* src/pcresearch.c (Pexecute): Improve comment about this.

2014-09-12  Paul Eggert  <>

	grep: diagnose -P in non-UTF-8 multibyte locale
	* src/pcresearch.c (Pcompile):
	libpcre supports only unibyte and UTF-8 locales,
	so report an error and exit if used in other locales.
	* NEWS: Mention this.
	* tests/euc-mb: Test this.

2014-09-12  Jim Meyering  <>

	doc: move NEWS note about GREP_OPTIONS into proper section
	* NEWS (Changes in behavior): Move the note about GREP_OPTIONS
	from the 2.20 section into the section for the upcoming release.

2014-09-12  Paul Eggert  <>

	grep: make GREP_OPTIONS obsolescent
	* NEWS:
	* doc/grep.texi (Environment Variables):
	Document that GREP_OPTIONS is obsolescent now.
	* src/grep.c (main): Warn if GREP_OPTIONS is used.
	* tests/r-dot, tests/skip-device: Don't use GREP_OPTIONS.

2014-09-11  Paul Eggert  <>

	doc: bug tracker has moved to
	* doc/
	* doc/grep.texi (Reporting Bugs): Document this.

	grep: fix false matches with -P '...$' and invalid UTF-8
	* tests/pcre-invalid-utf8-input: Add a test for that.

	grep: fix false matches with -P '...$' and invalid UTF-8
	* src/pcresearch.c (Pexecute): Use PCRE_NOTEOL when matching
	initial substrings of a line.

2014-09-10  Jim Meyering  <>

	tests: add expect-to-fail test for a glibc regexp bug
	* tests/triple-backref: New file.
	* tests/ (TESTS): Add it.
	(XFAIL_TESTS): List it as a known, always-failing test.
	Based on the bug report from Paul Eggert:

	maint: avoid distcheck failure
	* (EXTRA_DIST): Add .mailmap.

2014-09-10  Paul Eggert  <>

	grep: port recent fix to older pcre version
	* src/pcresearch.c (Pexecute): Don't assume that a pcre_exec
	that returns PCRE_ERROR_NOMATCH leaves its sub argument alone.
	This assumption is false for libpcre-3 version 8.31-2ubuntu2.

2014-09-09  Paul Eggert  <>

	grep: -P now treats invalid UTF-8 input as non-matching
	Problem reported by Santiago Vila in:
	* NEWS: Mention this.
	* src/pcresearch.c (Pexecute): Treat UTF-8 encoding errors
	as non-matching data, instead of exiting 'grep'.
	* tests/pcre-infloop: grep now exits with status 1, not 2.
	* tests/pcre-invalid-utf8-input: grep now exits with status 0, not 2.

2014-08-14  Paul Eggert  <>

	grep: fix integer-width bugs in undossify_input etc.
	undossify_input bug reported by Vincent Lefevre in:
	* src/dosbuf.c (undossify_input): Return size_t, not int.
	* src/grep.c (fillbuf): Work portably even if safe_read returns a
	value greater than SSIZE_MAX, e.g., if there's an I/O error.

2014-08-03  Paul Eggert  <>

	doc: document LANGUAGE
	Reported by Benno Schulenberg in:
	* doc/grep.texi (Environment Variables): Document LANGUAGE.

	doc: prefer @env to @code
	Reported by Benno Schulenberg in:
	* doc/grep.texi: Avoid @code in favor of @env, or of nothing at all.

2014-07-11  Paul Eggert  <>

	doc: Document -r vs --exclude more carefully.
	Problem reported by Hugues Andreux in:
	* doc/grep.texi (File and Directory Selection): Be more careful
	about documenting the interaction between recursive searching,
	--include, --exclude, and --exclude-dir.

2014-06-27  Jim Meyering  <>

	maint: split long lines, and enforce the 80-column limit
	* (sc_long_lines): New rule, from coreutils; exempt tests/*
	* src/grep.c (usage): Tweak -F wording to shorten a line.
	Correct grammar in a comment.
	Split the --exclude-file=... description to fit within 80 columns.
	Use emit_bug_reporting_address, eliminating another long line.
	* src/dfa.c: Split long lines.  No semantic change.
	* doc/grep.texi: Likewise.
	* tests/include-exclude: Split a long line.
	* tests/backref: Split long lines.
	* tests/empty: Likewise.
	* tests/fmbtest: Likewise.

	doc: update HACKING
	* HACKING: Update from coreutils.

	maint: generate distributed THANKS from VC'd
	* (THANKS): New rule.
	* New file.
	* THANKS: Remove.  Now it's generated from the combination of and git logs.
	* .mailmap: New file.
	* (sc_THANKS_in_duplicates): New syntax-check rule, from
	* .gitignore: Add THANKS.
	* thanks-gen: New file, from coreutils.

2014-06-27  Paul Eggert  <>

	grep: with -E, unmatched ')' matches itself
	Problem reported by Nathan Weeks in:
	* src/grep.c (Ecompile): Also specify RE_UNMATCHED_RIGHT_PAREN_ORD.
	* doc/grep.texi (Fundamental Structure), NEWS: Document this.
	* tests/ere.tests: Add a couple of tests for this.
	* tests/spencer1.tests: Fix exit status.

2014-06-17  Paul Eggert  <>

	build: avoid -Wstack-protector
	This allows the use of --enable-gcc-warnings on Gentoo and Ubuntu.
	* (WERROR_CFLAGS): Avoid -Wstack-protector.

	This can be worked around, but the cure is worse than the disease.

2014-06-17  Paul Eggert  <>

	build: don't make output files read-only
	This led to problems, such as the prompt "mv: try to overwrite
	'egrep', overriding mode 0555 (r-xr-xr-x)? " during a build.
	It can be worked around, but the cure is worse than the disease;
	making output files read-only is more trouble than it's worth.
	* doc/ (grep.1, egrep.1, fgrep.1):
	* lib/ (colorize.c):
	* src/ (egrep fgrep):
	Don't make output files read-only.  Prefer separate commands to
	'&&' when either will do.

2014-06-08  Paul Eggert  <>

	maint: remove grep.spec
	* grep.spec: Remove; obsolete and evidently not used.

2014-06-07  Paul Eggert  <>

	doc: use gnulib fdl module
	* bootstrap.conf (gnulib_modules): Add fdl.
	* doc/fdl.texi: Remove, as this now comes from gnulib.
	* doc/.gitignore: Update to match current sources.

2014-06-06  Jim Meyering  <>

	build: improve rule to generate egrep+fgrep scripts
	* src/ (egrep fgrep): chmod a=rx generated files,
	and remove $@-t before attempting to redirect to it, in case it
	is read-only.

	build: don't redirect directly to $@
	* lib/ (colorize.c): Don't redirect directly to target, $@.
	Otherwise, we could create a corrupt colorize.c file with a
	timestamp that indicates it is up to date.
	Also, make the generated file read-only.

2014-06-05  Paul Eggert  <>

	grep: undo part of previous change
	* src/dfa.c (enlist): Undo part of previous change that doesn't
	look correct and doesn't help performance much anyway.

	grep: use system strstr if available and fast
	Problem reported by Norihiro Tanaka in:
	* NEWS: Document this.
	* bootstrap.conf (gnulib_modules): Add strstr.
	* src/dfa.c (istrstr): Remove.
	(enlist): Use strstr instead.  Wait until we need memory before
	allocating it; this can save an unnecessary allocate and free.

	build: update gnulib submodule to latest

2014-06-03  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.20
	* NEWS: Record release date.

2014-05-30  Jim Meyering  <>

	grep: fix --max-count=N (-m N) to stop reading after Nth match
	With --max-count=N (-m N), grep is supposed to stop reading input
	after it has found the Nth match.  However, a recent context-
	related change made it so grep would always read to end of file.
	* src/grep.c (prtext): Don't let a negative "out_after" value
	make "pending" line count negative.
	* tests/max-count-overread: New test, for this.
	* tests/ (TESTS): Add it.
	* NEWS (Bug fixes): Mention it.
	* THANKS: Add names of two recent bug reporters.
	This bug was introduced by commit v2.18-139-g5122195.
	Reported by Marc Aldorasi in

2014-05-29  Jim Meyering  <>

	dfa: fix off-by-one under-allocation from recent change
	Commit v2.19-10-gc32ff67 mistakenly made this change:
	  -realloc_trans_if_necessary (d, 1);
	  +realloc_trans_if_necessary (d, 0);
	which led to a heap buffer overflow.
	* src/dfa.c (dfaexec): Allocate space for one state, as before.

2014-05-28  Norihiro Tanaka  <>

	dfa: fix bug with regex containing multiple begin/end-line constraints
	grep -E 'a(b$|c$)' would mistakenly match "aa".
	* src/dfa.c (dfamust): When resetting 'is' in OR, also reset
	'begline' and 'endline' of 'must'.
	* NEWS (Bug fixes): Mention it.
	This bug was introduced via commit v2.18-85-g2c94326.
	Reported by Péter Radics in <>.

2014-05-26  Norihiro Tanaka  <>

	dfa: simplify building initial state
	build_state_zero doesn't need the struct dfa to be initialized,
	so remove the initialization and simplify.
	* src/dfa.c (build_state_zero): Remove.
	(dfaexec): Call realloc_trans_if_necessary and build_state directly.

	dfa: revert "grep: do not count newline before the start of buffer"
	This reverts commit 5dc3af2806d21455b818be3f9da26c372e4a7f8d.
	The previous change renders that commit unnecessary.

	dfa: do not clear the first state of a transition table
	If number of DFA states reaches 1024, build_state clears transition
	tables to save memory.  However, the initial state is always used,
	so clearing it just wastes time.
	* src/dfa.c (build_state): Do not clear the initial state's
	transition and failure tables.

	grep: remove unnecessary argument
	* src/grep.c (do_execute): Remove argument 'start_ptr'.  It's always null.
	All uses changed.

2014-05-24  Paul Eggert  <>

	grep: --exclude-dir=FOO/ now ignores the trailing slash
	Problem reported by Khaled Ziyaeen; see:
	* NEWS, doc/grep.texi (File and Directory Selection): Document this.
	* src/grep.c (main): Implement this.
	* tests/include-exclude: Test this.

	dist: don't distribute lib/colorize.c
	'configure' creates this file, so it shouldn't be distributed; see:
	* (COLORIZE_SOURCE): New macro.
	Don't use AC_CONFIG_LINKS for lib/colorize.c.
	* lib/ (nodist_libgreputils_a_SOURCES): New macro.
	(libgreputils_a_SOURCES): Remove colorize.c.
	(CLEANFILES): Add colorize.c
	(colorize.c): New rule.

2014-05-23  behoffski  <>

	maint: uncapitalize first letter of two dfaerror message strings
	* dfa.c (lex): Make two message strings consistent with all of
	the others: do not capitalize the first letter of the first word.

2014-05-23  Jim Meyering  <>

	maint: revert "grep: port mb_next_wc to RHEL 6.5 x86-64"
	This reverts commit v2.18-148-ga6ae68d.
	Now that we have gnulib change v0.1-131-g2a045bc, "mbrlen, mbrtowc:
	fix bug with empty input", this work-around is no longer needed.

	gnulib: update, for mbrlen/mbrtowc empty input bug fix

2014-05-22  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.19
	* NEWS: Record release date.

2014-05-21  Jim Meyering  <>

	maint: avoid new false-positive syntax-check failure
	* (exclude_file_name_regexp--sc_prohibit_doubled_word):
	Exempt new test file that contains legitimate use of "in in".

2014-05-17  Norihiro Tanaka  <>

	tests: add test case for newline-count fix
	* tests/count-newline: New test.
	* tests/ (TESTS): Add it.

2014-05-16  Norihiro Tanaka  <>

	grep: do not count newline before the start of buffer
	* src/dfa.c (build_state): When checking whether the previous
	character was a newline, do not count any newline before the
	start of the buffer.

2014-05-15  Paul Eggert  <>

	grep: port mb_next_wc to RHEL 6.5 x86-64
	* src/searchutils.c (mb_next_wc): Work around glibc bug 16950; see:
	This bug was masked in the other GNU/Linux tests I made.  It was
	exposed on RHEL 6.5 x86-64, where the compiler (GCC Red Hat 4.4.7-4)
	happened to use temporaries in a different way.
	Also see recent changes to the Gnulib documentation in this area:

	tests: port mb-non-UTF8-performance to RHEL 6.5
	* tests/mb-non-UTF8-performance (timeout): Use an integer,
	as 'timeout 1.234' doesn't work in EUC locales.

2014-05-12  Paul Eggert  <>

	egrep, fgrep: port to Solaris 10 /bin/sh
	This old shell doesn't grok ${0%/*}; see:
	* src/ (egrep fgrep): Don't assume the shell does substrings.
	* src/ (dir): New var, so that the substring calculation is
	done only once (which is a small win even with newer shells),
	and so that the calculation is easier to edit on older shells.

2014-05-10  Jim Meyering  <>

	maint: NEWS: adjust wording to reflect move
	* NEWS (Improvements): Correct direction-relative wording,
	now that the referent is below, not above.

	maint: NEWS: move "Improvements" to the top
	* NEWS: Move the small "Improvements" section to precede
	the longer "Bug fixes" one.

	gnulib: update submodule to latest, and bootstrap
	* gnulib: Update submodule.
	* bootstrap: Update from gnulib.

2014-05-10  Paul Eggert  <>

	dfa: omit double includes
	* src/dfa.c: Don't include stddef.h or stdbool.h, as dfa.h includes
	them already, and it's the same module as we are.
	Suggested by Aharon Robbins in:

	dfa: fix bug with \< etc in multibyte locales
	Problem reported by Stephane Chazelas in:
	* NEWS: Document the fix.
	* src/dfa.c (dfaoptimize): Remove any superset if changing from
	UTF-8 to unibyte, and if the pattern has no backreferences.
	(dfassbuild): In multibyte locales, treat \< \> \b \B as
	backreferences in the DFA, since the DFA relies on unibyte
	tests to check them.
	(dfacomp): Optimize after building the superset, so that
	dfassbuild can depend on d->multibyte.  A downside is that
	dfaoptimize must remove supersets that are likely slower than the
	DFA after optimization, but that's been done in the
	above-described change.
	* tests/ (XFAIL_TESTS): Remove word-delim-multibyte,
	since the test works now.

	tests: add test case for -C 0 change
	* tests/context-0: New test.
	* tests/ (TESTS): Add it.

	grep: -A 0, -B 0, -C 0 now output a separator
	Problem reported by Dan Jacobson in:
	* NEWS:
	* doc/grep.texi (Context Line Control): Document this.
	* src/grep.c (prtext): Output a separator even if context is zero.
	(main): Default context is now -1, not 0.

2014-05-09  Paul Eggert  <>

	grep: minor improvements to retry-DFA-superset patch
	* src/dfasearch.c (EGexecute): Avoid unnecessary test in a context
	where memrchr cannot return a null pointer.

2014-05-09  Norihiro Tanaka  <>

	grep: retry DFA superset after matching multiple lines
	* src/dfasearch.c (EGexecute): Without this patch, the code reverts
	to KWset when the DFA superset matches multiple lines.
	However, if the DFA superset matches multiple lines, it most likely
	also matches a single line, and reverting to KWset means dfafast
	won't work effectively.  Change the code so that it retries the DFA
	superset immediately after it matches multipline lines.  On my platform
	this improves the performance of "LC_ALL=C grep '\(ab\)cd\1d' k" from
	3.48 to 2.14 seconds realtime, where k contains the output of
	"yes abcdabc | head -50000000".

	dfa: fix inconsistency in multibyte locales
	* src/dfa.c (dfaexec): Use the same exit condition in multibyte
	locales as in unibyte.

2014-05-08  Jim Meyering  <>

	maint: mark some breakless cases with /* fallthrough */ comment
	* src/dfa.c (addtok_mb, dfaanalyze): Add comment so that it is
	clear that the "break" statement is deliberately omitted.

2014-05-08  Paul Eggert  <>

	dfa: assume C89 for CHAR_BIT
	* src/dfa.c (CHARBITS): Remove.  All uses replaced by CHAR_BIT.
	(NOTCHAR): Now an enum, since it need not be a macro.

	dfa: don't assume unsigned int is exactly 32 bits wide
	Sun C 5.12 (sparc) warns of the potential unportability.
	* src/dfa.c (charclass_word): New type, for clarity.
	All relevant uses of 'unsigned' changed.
	(CHARCLASS_WORD_BITS): Rename from INTBITS.  All uses changed.
	Now an enum, since it needn't be a macro.
	(CHARCLASS_WORDS): Rename from CHARCLASS_INTS.  All uses changed.
	(setbit, clrbit): Cast 1 to charclass_word, for clarity.
	(notset, add_utf8_anychar, dfastats):
	Don't assume unsigned int is exactly 32 bits wide.
	(dfastate): Don't rely on implementation-defined conversion of
	greater-than-INT_MAX unsigned to int.  Change bit test to resemble
	tstbit more.

	maint: fix indenting to pacify 'prohibit_tab_based_indentation'
	* src/dfa.c: Use spaces and not tabs to indent some lines.

	grep: simplify and clarify invert-related code
	* src/grep.c (out_invert, prtext): Use bool for booleans.
	(prline): Remove unnecessary '!!' on a value that is always 0 or 1.
	(prtext): Remove last arg NLINESP; use !out_invert instead.  All uses
	changed.  Move decls to nearer uses, since we can assume C99 here.
	Update 'outleft' and 'after_last_match' here; it's simpler.
	(grepbuf): Compute return value by subtracting new from old 'outleft',
	rather than by keeping a separate running total.  Avoid code duplication
	by arranging for prtext to be called from one place, not three.

2014-05-08  Norihiro Tanaka  <>

	grep: improve performance of -v when combined with -L, -l or -q
	Problem reported by Jörn Hees in:
	* src/grep.c (grepbuf, grep): When -v is combined with -L, -l, or -q,
	don't read data unnecessarily after a non-match is found.

2014-05-06  Paul Eggert  <>

	doc: mention performance changes
	* NEWS: Discuss recent performance improvements and downgrades.

	dfa: clarify use of "if"
	The phrase "Y is true if X" is logically equivalent to "X implies Y",
	but often "X if and only if Y" was intended.
	* src/dfa.c, src/dfa.h: Reword to avoid the incorrect use of "if".

	dfa: minor performance improvement for previous change
	* src/dfa.c (struct dfa): New member 'fast'.  Remove 'has_backref'.
	All uses changed.

2014-05-06  Norihiro Tanaka  <>

	dfa: speed up 'dfaisfast'
	* src/dfa.c (struct dfa): New member 'has_backref'.
	(addtok_mb): Set it.
	(dfaisfast): Use it.

2014-05-05  Paul Eggert  <>

	grep: fix -w match next to a multibyte letter
	* NEWS: Document this.
	* src/dfasearch.c, src/kwsearch.c (WCHAR): Remove.
	(wordchar): New static function.
	* src/dfasearch.c (EGexecute):
	* src/kwsearch.c (Fexecute): Use the new functions, so that the
	code works correctly if a multibyte character adjacent to the
	match has two or more bytes.
	* src/search.h, src/searchutils.c (mb_prev_wc, mb_next_wc):
	New functions.
	* tests/word-delim-multibyte: Add a test for grep -w (which now
	passes), and a test for \> (which still fails).  The \< test also
	still fails.

	grep: improve internal API for multibyte boundary
	* src/search.h, src/searchutils.c (mb_goback): Rename from
	is_mb_middle.  Omit last arg.  Return number of bytes to go back,
	not just a boolean.  All uses changed.
	* src/dfasearch.c (EGexecute):
	* src/kwsearch.c (Fexecute): Adjust to API change.
	* src/kwsearch.c (Fexecute): Eliminate common subexpression.

	grep: fix encoding-error incompatibilities among regex, DFA, KWset
	This follows up to and fixes a different
	set of incompatibilities, namely between the regex matcher and the
	other matchers, when the pattern contains encoding errors.
	The GNU regex matcher is not consistent in this area: sometimes
	an encoding error matches only itself, and sometimes it
	matches part of a multibyte character.  There is no documentation
	for grep's behavior in this area and users don't seem to care,
	and it's simpler to defer to the regex matcher for problematic
	cases like these.
	* NEWS: Document this.
	* src/dfa.c (ctok): Remove.  All uses removed.
	(parse_bracket_exp, atom): Use BACKREF if a pattern contains
	an encoding error, so that the matcher will revert to regex.
	* src/dfasearch.c, src/grep.c, src/pcresearch.c, src/searchutils.c:
	Don't include dfa.h, since search.h now does that for us.
	* src/dfasearch.c (EGexecute):
	* src/kwsearch.c (Fexecute): In a UTF-8 locale, there's no need to
	worry about matching part of a multibyte character.
	* src/grep.c (contains_encoding_error): New static function.
	(main): Use it, so that grep -F is consistent with plain fgrep
	when the pattern contains an encoding error.
	* src/search.h: Include dfa.h, so that kwsearch.c can call using_utf8.
	* src/searchutils.c (is_mb_middle): Remove UTF-8-specific code.
	Callers now ensure that we are in a non-UTF-8 locale.
	The code was clearly wrong, anyway.
	* tests/fgrep-infloop, tests/invalid-multibyte-infloop:
	* tests/prefix-of-multibyte:
	Do not require that grep have a particular behavor for this test.
	It's OK to match (exit status 0), not match (exit status 1), or
	report an error (exit status 2), since the pattern contains an
	encoding error and grep's behavior is not specified for such
	patterns.  Test only that KWset, DFA, and regex agree.
	* tests/prefix-of-multibyte: Add tests for ABCABC and __..._ABCABC___.

2014-05-04  Paul Eggert  <>

	dfa: minor simplification
	* src/dfa.c (parse_bracket_exp): Use enum, not macro, and move var
	to just the scope it's needed.

	grep: simplify and fix problems with KWset-DFA agreement patch
	* src/dfa.c (dfambcache, parse_bracket_exp): Simplify.
	(mbs_to_wchar, wctok, FETCH_WC, match_anychar, match_mb_charset)
	(check_matching_with_multibyte_ops, transit_state_consume_1char)
	(transit_state, dfaexec): Use wint_t, not wchar_t, so that
	WEOF is treated correctly on platforms where WEOF is not a valid
	wchar_t value.
	(ctok, lex): Use int, not unsigned int, for characters,
	so that EOF is treated more naturally.
	(parse_bracket_exp): Use NOTCHAR to mark uninitialized char, since
	FETCH_WC can now set the char to EOF.
	(lex): Remove unnecessary test for EOF.
	(parse_bracket_exp, atom): Swap then and else parts, to put
	the small one first; this is more readable here.
	* src/searchutils.c (is_mb_middle): Simplify.

	tests: improve coverage for prefix-of-multibyte
	* tests/prefix-of-multibyte: Also test the regex version.

2014-05-04  Norihiro Tanaka  <>

	grep: make KWset and DFA agree about invalid sequences in patterns
	* src/dfa.c (dfambcache): Don't cache invalid sequences, because they can't be
	represented by wide characters.
	(dfambcache, mbs_to_wchar): Return WEOF for invalid sequences.
	(ctok): New global variable.
	(parse_bracket_exp, atom, match_anychar, match_mb_charset): Don't allow WEOF.
	(lex): Set 'ctok'.
	* src/kwsearch.c (Fexecute):
	* src/searchutils.c (is_mb_middle): Don't check here.
	* tests/invalid-multibyte-infloop: Adjust to fixed behavior.
	* tests/prefix-of-multibyte: Add test cases for this bug.

2014-05-03  Jim Meyering  <>

	maint: make ChangeLog generation more robust
	* (gen-ChangeLog): Sync changes from GNU coreutils,
	to ensure exit status is propagated, and to support an optional
	git-log-fix file.

2014-05-03  Paul Eggert  <>

	grep: clarify EGexecute slightly
	* src/dfasearch.c (EGexecute): Change if-then-else to !if-else-then.

2014-05-03  Norihiro Tanaka  <>

	grep: fix the bug in previous patch.
	* src/dfasearch.c (EGexecute): Do it.

2014-04-30  Paul Eggert  <>

	grep: simplify EGexecute further
	* src/dfa.c, src/dfa.h (dfasuperset): Arg is now const pointer.
	Now pure.
	* src/dfasearch.c (EGexecute): Coalesce some duplicate code.
	Don't worry about memrchr returning NULL when that's impossible.

2014-04-30  Norihiro Tanaka  <>

	grep: adjust timing back to kwset when dfaisfast is true
	* src/dfasearch.c (EGexecute): If DFA fails after kwset succeeds,
	the code doesn't return to kwset until it reaches the end of the buffer
	or finds a match.  Because of this, although some cases speed up,
	others slow down.

	Adjust the heuristic for switching to the DFA, so that it
	is more likely to switch at the right times.

2014-04-30  Norihiro Tanaka  <>

	grep: simplify superset
	* src/dfa.h (dfahint): Remove decl.
	(dfasuperset): New decl.
	* src/dfa.c (dfahint): Remove.
	(dfassbuild): Rename from dfasuperset.
	(dfasuperset): New function.  It returns the superset of D.
	* src/dfasearch.c: Use dfasuperset instead of dfahint, and simplify.

	dfa: optimize memory allocation
	* src/dfa.c (epsclosure): get the value of 'visited' from the argument.
	(dfaanalyze): Define and allocate variable 'visited'.
	(dfastate): Use not 'insert' but 'merge' to insert positions for
	state 0 of DFA.

2014-04-29  Norihiro Tanaka  <>

	kwset: improve performance by inlining tr
	Without this change, older versions of GCC won't inline 'tr', and this
	can hurt performance significantly.  See:
	* src/kwset.c (tr): Make it inline.

2014-04-27  Jim Meyering  <>

	gnulib: update to latest
	* gnulib: This fixes a bug whereby running bootstrap
	would remove our build-aux/git-log-fix file.

2014-04-27  Paul Eggert  <>

	kwset: improve performance by inlining more
	Problem reported by Norihiro Tanaka in <>.
	* src/kwset.c (bmexec_trans): Rename from bmexec, and make it inline.
	(bmexec): New implementation, which calls bmexec_trans.  This helps
	GCC inline more aggressively with the default optimization, and
	improves performance 25% with the reported benchmark on my host.

2014-04-26  Paul Eggert  <>

	kwset: speed up by using memchr2
	Idea suggested by Eric Blake in:
	* bootstrap.conf (gnulib_modules): Add memchr2.
	* src/kwset.c: Include stdint.h, for uintptr_t.  Include memchr2.h.
	(struct kwset): New members gc1, gc2, gc1help.
	(tr): Move earlier, so it can be used earlier.
	(kwsprep): Initialize struct kwset's new members.
	(memchr_kwset): Rename from memchr_trans.  Combine C and TRANS args into
	new arg KWSET.  All uses changed.  Use memchr2 when appropriate.
	(bmexec): Use new members instead of recomputing their values.
	Increase advance_heuristic; it's just a guess, but memchr2 probably
	makes it reasonable to increase it.

	kwset: improve performance when large Boyer-Moore key doesn't match
	* src/kwset.c (bmexec): As a heuristic, prefer memchr to seeking
	by delta1 only when the latter doesn't advance much.

	dfa: fix index bug in previous patch, and simplify
	* src/dfa.c, src/dfa.h (dfaisfast): Arg is const pointer.
	* src/dfa.c (dfaisfast): Simplify, since supersets never contain BACKREF.
	* src/dfa.h (dfaisfast): Declare to be pure.
	* src/dfasearch.c (EGexecute): Fix typo that could cause buffer
	read overrun when !dfafast.  Hoist duplicate computation out
	of an if's then and else parts.

2014-04-26  Norihiro Tanaka  <>

	grep: speed up for a case to repeat failure in DFA after success in kwset
	A DFA is typically much faster if it is unibyte and does not set BACKREF.
	Skip kwset if the DFA is fast.  For example:

	    yes abcdabc | head -50000000 >k
	    env LC_ALL=C time -p src/grep -i '' k

	This improved real-time from 4.86 to 1.34 s.

	* src/dfa.c, src/dfa.h (dfaisfast): New function.
	* src/dfasearch.c (EGexecute): Use it.

2014-04-24  Paul Eggert  <>

	dfa: fix recently-introduced memory leak
	Problem reported by Aharon Robbins in:
	* src/dfa.c (dfasuperset): free after dfafree.

	misc: fix doc and test bugs re grep -z
	Problem reported by Stephane Chazelas in:
	* doc/grep.texi (Usage): Remove incorrect example with -P.
	* tests/pcre: Improve test so that it actually tests whether \s
	matches a newline.

	dfa: minor simplification of dfaexec
	* src/dfa.c (dfaexec): Streamline updating of returned values.
	Don't bother to check d->multibyte before updating mbp.
	Avoid duplicate p > end test.

2014-04-24  Paul Eggert  <>

	dfa: simplify and be more consistent about MB_CUR_MAX
	* src/dfa.c (struct dfa): New member 'multibyte',
	replacing 'mb_cur_max'.  All uses changed.  Use this new member
	consistently, instead of sometimes referring to MB_CUR_MAX directly.

	dfa: fix comment
	* src/dfa.c (maybe_realloc): Fix comment to match behavior better.

2014-04-24  Norihiro Tanaka  <>

	grep: skip checking of multibyte character boundary, reaching at eolbyte
	* src/dfa.c (dfaexec): Skip checking of multibyte character boundary,
	reaching at eolbyte.

2014-04-24  Paul Eggert  <>

	dfa: fix incorrect comment that led to heap overrun
	* dfa.c (maybe_realloc): Fix comment to match behavior.

	dfa: minor tuneup of dfamust memory savings patch
	* src/dfa.c (allocmust): Use xmalloc, not xzalloc.
	Initialize the must completely, so that the caller need not
	invoke resetmust.  All callers changed.
	(dfamust): Omit asserts that aren't needed on typical machines
	where dereferencing NULL dumps core.  Don't leak memory if the
	pattern contains a NUL byte.

2014-04-24  Norihiro Tanaka  <>

	grep: avoid wasting memory for large patterns in dfamust
	* src/dfa.c (struct must): New member 'prev'.  It points to the
	previous must.
	(allocmust): New function.
	(freemust): New function.
	(dfamust): Use it.

2014-04-24  Jim Meyering  <>

	grep: fix new heap write buffer overrun
	* src/dfa.c (parse_bracket_exp): Fix off-by-one allocation error.
	Exposed by running the tests with an ASAN-enabled binary (i.e.,
	created using gcc's -fsanitize=address option).  Introduced by
	commit v2.18-70-gd3d9612, "dfa: simplify range char allocation".

2014-04-24  Paul Eggert  <>

	build: suppress unsafe-loop-optimizations warnings
	I ran into one of these while trying out GCC 4.9.0's new
	-fsanitize=undefined option.  The warning told me that GCC didn't
	do an unsafe optimization, but in 'grep' this is not typically a
	symptom of a programming error.
	* (WERROR_CFLAGS): Suppress -Wunsafe-loop-optimizations.

2014-04-23  Paul Eggert  <>

	dfa: fix memory leak reintroduced by previous patch
	Reported by Norihiro Tanaka in <>.
	* src/dfa.c (dfaexec): Allocate mb_match_lens and mb_follows only
	if not already allocated.
	(free_mbdata): Null out mb_match_lens to mark it as being freed.

2014-04-23  Jim Meyering  <>

	tests: use consistent spelling for locale name, en_US.UTF-8
	* tests/pcre-infloop: Spell locale name, en_US.UTF-8, consistently,
	converting this one use from "en_US.utf8", which would provoke a
	test failure on OS/X.

2014-04-23  Paul Eggert  <>

	dfa: omit static variables that limited dfaexec to one struct dfa
	Problem reported by Aharon Robbins in:
	* src/dfa.c (struct dfa): New member mbs.
	mb_follows is now a position_set, not a pointer to one;
	this simplifies memory allocation.  All uses changed.
	(mbs_to_wchar): Put DFA arg at the end, in place of the mbstate_t *arg,
	since the DFA now contains an mbstate_t.  All uses changed.
	(mbs): Remove static variable.
	(dfaexec): Remove static bool that attempted to optimize memory
	allocation, as this wasn't correct for Gawk.  Perhaps we can think
	of a better way to optimize memory.

2014-04-22  Paul Eggert  <>

	kwset: simplify and speed up Boyer-Moore unibyte -i in some cases
	This improves the performance of, for example,
	yes jjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjj | head -10000000 | grep -i jk
	in a unibyte locale.
	* src/kwset.c (memchr_trans): New function.
	(bmexec): Use it.  Simplify the code and remove some of the
	confusing gotos and breaks and labels.  Do not treat glibc memchr
	as a special case; if non-glibc memchr is slow, that is lower
	priority and I suppose we can try to work around the problem in

2014-04-22  Norihiro Tanaka  <>

	grep: speed-up by using memchr() in Boyer-Moore searching
	memchr() of glibc is faster than seeking by delta1 on some platforms.
	When there is no chance to match for a while, use it on them.
	* src/kwset.c (bmexec): Use memchr() in Boyer-Moore searching.

2014-04-22  Paul Eggert  <>

	kwset: simplify Boyer-Moore with unibyte -i
	This change doesn't significantly affect performance on my platform,
	and should make the code easier to maintain.
	Remove these macros, in favor of ...
	(tr, bm_delta2_search): New functions.  All uses changed.
	The latter function is inline because this improves code size and
	runtime CPU slightly on x86-64 with gcc -O2 (GCC 4.9.0).
	(bmexec): Prefer tr when that's simpler.

2014-04-22  Norihiro Tanaka  <>

	grep: may also use Boyer-Moore algorithm for case-insensitive matching
	* src/kwset.c (BM_DELTA2_SEARCH, LAST_SHIFT, TRANS): New macro.
	(bmexec): Use character translation table.
	(kwsexec): Call bmexec for case-insensitive matching.
	(kwsprep): Change the `if' condition.

2014-04-21  Paul Eggert  <>

	grep: -P now rejects invalid input sequences in UTF-8 locales
	See <> and <>.
	* NEWS: Document this.
	* src/pcresearch.c (Pexecute): Do not use PCRE_NO_UTF8_CHECK,
	as this leads to undefined behavior when the input is not UTF-8.
	* tests/pcre-infloop, tests/pcre-invalid-utf8-input:
	Exit status is now 2, not 1, when grep -P is given invalid UTF-8
	data in a UTF-8 locale.

	dfa: minor improvements to previous patch
	* src/dfa.c (dfamust): Use &=, not if-then.
	* src/dfa.h (struct dfamust):
	* src/dfasearch.c (begline, hwsmusts): Use bool for boolean.
	* src/dfasearch.c (kwsmusts):
	* src/kwsearch.c (Fcompile): Prefer decls after statements.
	* src/dfasearch.c (kwsmusts): Avoid conditional branch.
	* src/kwsearch.c (Fcompile): Unify the two calls to kwsincr.

2014-04-21  Norihiro Tanaka  <>

	grep: speed-up for exact matching with begline and endline constraints.
	dfamust turns on the flag when a state exactly matches the proposed one.
	However, when the state has begline and/or endline constraints, turns
	off it.

	This patch enables to match a state exactly, even if the state has
	begline and/or endline constraints.  If a exact string has one of their
	constrations, the string adding eolbyte to a head and/or foot is pushed
	to kwsincr().  In addition, if it has begline constration, start
	searching from just before the position of the text.

	* src/dfa.c (variable must): New members `begline' and `endline'.
	(dfamust): Consideration of begline and endline constrations.
	* src/dfa.h (struct dfamust): New members `begline' and `endline'.
	* src/dfasearch.c (kwsmusts): If a exact string has begline constration,
	start searching from just before the position of the text.
	(EGexecute): Same as above.
	* src/kwsearch.c (Fexecute): Same as above.

2014-04-20  Paul Eggert  <>

	dfa: fix bug that caused NUL to be mishandled in patterns
	This bug was introduced in the early-2012 patches that fixed some
	context-handling bugs.  Bisecting found commit
	d8951d3f4e1bbd564809aa8e713d8333bda2f802 (2012-02-05 18:00:43 +0100),
	but it apears the underlying problem was introduced in commit
	8b47c4cf6556933f59226c234b0fe984f6c77dc7 (2012-01-03 11:22:09 +0100).
	* NEWS: Mention bug fix.
	* src/dfa.c (char_context): Consider NUL to be a newline only if -z.
	* tests/ (TESTS): Add null-byte.
	* tests/null-byte: New file.

2014-04-19  Jim Meyering  <>

	build: reenable some compiler warning options

2014-04-18  Paul Eggert  <>

	dfa: fix pointer type conversion bug
	The code converted between size_t * and ptrdiff_t *, which wasn't
	diagnosed by modern x86-64 GCC but isn't portable.  Problem
	reported by Norihiro Tanaka in <>.
	* (WERROR_CFLAGS): Don't add -Wno-pointer-sign.
	We want GCC to diagnose pointer signedness problems, as they
	violate the C standard and other compilers no doubt complain too.
	* src/dfa.c (struct dfa): Change type of salloc to size_t.
	(realloc_trans_if_necessary): Convert signed value to size_t before
	passing its address to x2nrealloc.  Changing the type of tralloc
	to size_t might have led to problems elsewhere.

2014-04-18  Jim Meyering  <>

	maint: Revert "dfa: avoid new NULL dereference"
	This reverts commit 5190041fe515743ef4545abf287d243bc025c701.
	It was only a bug if one neglected to update to the latest gnulib.
	With the newer xn2realloc, there is no problem.

	dfa: avoid new NULL dereference
	* src/dfa.c (dfa_charclass_index): Restore a "+ 1" mistakenly omitted
	during recent improvements.  Introduced in v2.18-66-g6a60fd5.

2014-04-17  Paul Eggert  <>

	dfa: minor cleanup
	* src/dfa.c (MAX): Remove; no longer used.

2014-04-17  Norihiro Tanaka  <>

	dfa: speed up by checking multibyte characters on demand
	If dfaexec() runs in non-UTF8 locales, length and wide character
	representation are checked for all characters of a line in a input
	string.  However, if matched early in the line, results for remaining
	characters are wasted.

	This patch checks multibyte characters on demand.  It should work
	faster for early matches, and reduces memory requirements.

	* src/dfa.c (struct dfa): Remove members mblen_buf, nmblen_buf,
	inputwcs, ninputwcs.  All uses removed.
	(buf_begin, buf_end, prepare_wc_buf): Remove.  All uses removed.
	(SKIP_REMAINS_MB_IF_INITIAL_STATE): Remove.  This is now expanded
	when used.
	(match_anychar, match_mb_charset, check_matching_with_multibyte_ops):
	New arg wc, mbclen.  Remove arg idx.  All uses changed.
	(transit_state_consume_1char): New arg wc.  All uses changed.
	(transit_state): New arg 'end'.  All uses changed.

2014-04-17  Paul Eggert  <>

	dfa: trans reallocation microoptimization
	* src/dfa.c (realloc_trans_if_necessary):
	Help the compiler avoid unnecessary reloads.

	dfa: simplify dfmust initialization
	* src/dfa.c (dfamust): Don't initialize musts twice.
	Use zcalloc, not xmalloc followed by zeroing.
	Make result a const pointer.

	dfa: simplify freelist
	* src/dfa.c (freelist): Don't null out array while freeing its
	pointers; the caller can do that if needed.
	(resetmust): Null out zeroth entry of array.

	dfa: avoid duplicate strlen when allocating memory
	* src/dfa.c (dfamust): Use xstrdup, not strlen (twice) + xmemdup.

	dfa: simplify memory allocation
	* src/dfa.c (icatalloc, freelist, enlist, comsubs, addlists, inboth)
	(dfamust): Don't worry about null arguments or results,
	as memory allocators no longer can return null pointers.
	(dfamust): Invoke malloc just once when building a concatenated string.

	dfa: simplify position set and element count allocation
	* src/dfa.c (dfaanalyze): Allocation position set info all at one go,
	and similarly for element count info.

	dfa: simplify multibyte_prop allocation
	* src/dfa.c (struct dfa): Simplify by removing nmultibyte_prop;
	it should always be the same as talloc.  All uses changed.

	dfa: simplify range char allocation
	* src/dfa.c (struct dfa): Simplify by allocating one array of ranges
	rather than one for range starts and another for range ends.
	All uses changed.

	dfa: simplify transition table allocation
	* src/dfa.c (struct dfa): Remove member 'realtrans', as it can
	be computed from 'trans'.  All uses changed.
	(realloc_trans_if_necessary): Move earlier, to avoid a forward decl.
	Use x2nrealloc to compute new size, rather than doing it by hand,
	which omits a check for unlikely overflow.
	(realloc_trans_if_necessary, dfafree): Adjust to the fact that
	d->trans now might be either NULL, or 1 + the pointer to free.
	(build_state, build_state_zero): Use realloc_trans_if_necessary
	instead of duplicating its code.

	dfa: better size-overflow check
	* src/dfa.c (dfasuperset): Let xnmalloc do the multiplication,
	to check for size arithmetic overflow better.

	dfa: avoid unnecessary work and other initialization
	* src/dfa.c (dfaanalyze, dfainit):
	Don't bother allocating when x2nrealloc will do it for us.
	(dfastate): Allocate grps and labels on the stack, as their
	size is known at compile time.
	(build_state): Use xmalloc, not xnmalloc, since the multiplication
	can be done at compile-time.

	dfa: clarify memory allocation and port to IRIX
	This change was prompted by a porting problem:
	IRIX defines its own MALLOC macro, which clashes with ours.
	More generally, the MALLOC etc. macros are confusing, as they
	look like functions but do not have C-function semantics.
	A functional style makes the code easier to read, and though
	it lengthens the code a bit here it'll make other
	simplifications easier.
	All uses replaced by xnmalloc etc.
	(REALLOC_IF_NECESSARY): Remove; all uses replaced by ....
	(maybe_realloc): New function.
	(copy, merge): Free and allocate rather than realloc, as we
	needn't save the contents.

2014-04-14  Jim Meyering  <>

	tests: detect an infloop-inducing bug in grep -P (pcre-8.35)
	* tests/pcre-infloop: New test.
	* tests/ (TESTS): Add it.

2014-04-12  Paul Eggert  <>

	build: update gnulib submodule to latest

2014-04-11  Paul Eggert  <>

	grep: improvements for the open-CSET patch
	* src/dfa.c (dfamust): Simplify by removing some duplicate code.
	Optimize patterns like [aaa] even when not case-folding.
	Avoid an unnecessary copy of the charclass.

2014-04-11  Norihiro Tanaka  <>

	grep: open CSET and transform into uppercase when MB_CUR_MAX == 1
	In unibyte locales with -i, kwset matching isn't helpful, because
	dfamust doesn't extract the CSET entries.  Fix dmamust so that it
	does that, and makes it possible to take out a longer fixed string
	from tokens.
	* src/dfa.c (dfamust): open CSET and transform into uppercase
	when MB_CUR_MAX == 1.

2014-04-11  Paul Eggert  <>

	grep: cleanup for HAS_DOS_FILE_CONTENTS issue
	While cleaning up the empty-string fix, I noticed that one part of
	the code worried about CRLF in pattern files whereas another part
	did not.  Fix this by using the same approach in both places,
	and make the CRLF code more modular in the process.
	* src/dosbuf.c (dos_binary, dos_unix_byte_offsets): New functions.
	(undossify_input, dossified_pos): Do nothing if ! O_BINARY.
	* src/grep.c: Always include dosbuf.c so that the code is
	checked statically even on non-DOS hosts.
	(dos_binary, dos_unix_byte_offsets): New decls.
	(undossify_input): Declare unconditionally.
	* src/grep.c (fillbuf, print_line_head, main):
	* src/kwsearch.c (Fcompile):
	Simplify by not worrying about HAVE_DOS_FILE_CONTENTS.
	* src/grep.c (main): fopen with "rt" if O_TEXT; this is simpler
	than worrying about HAVE_DOS_FILE_CONTENTS elsewhere.
	* src/system.h (HAVE_DOS_FILE_CONTENTS): Remove.

	grep: cleanup for empty-string fix
	* NEWS: Document it.
	* src/dfasearch.c (GEAcompile):
	* src/kwsearch.c (Fcompile):
	Use C99-style decls to simplify.  Avoid duplicate code.
	* tests/empty-line: Add some more tests like this.

2014-04-11  Norihiro Tanaka  <>

	grep: no match for the empty string included in multiple patterns
	* src/dfasearch.c (EGAcompile): Fix it.
	* src/kwsearch.c (Fcompile): Fix it.

2014-04-08  Paul Eggert  <>

	grep: remove bool_bf
	The extra complexity of this microoptimization wasn't ever much help,
	and currently it generated bigger code with gcc -O2 (x86-64).
	* src/dfa.c (bool_bf): Remove.  All uses replaced by plain 'bool',
	without a bitfield.

2014-04-08  Jim Meyering  <>

	maint: avoid sc_po_check syntax-check failure (kwset.c)
	* po/ Remove kwset.c from this list, since it
	no longer contains a translatable diagnostic.

2014-04-08  Paul Eggert  <>

	grep: port better to hosts with nonstandard nl_langinfo
	On some hosts, nl_langinfo returns strings other than "UTF-8" when
	UTF-8 is used, and (worse) return "UTF-8" even if the encoding is
	single-byte.  Work around these problems by trying a sample
	character instead.
	* src/dfa.c, src/pcresearch.c, src/searchutils.c:
	Don't include <langinfo.h>.
	* src/dfa.c (using_utf8): Test for UTF-8 by trying a character
	rather than by invoking nl_langinfo (CODESET); this is more
	portable in practice, and removes a dependency on
	* src/pcresearch.c: Include dfa.h, for using_utf8.
	(Pcompile): Use using_utf8 rather than nl_langinfo.

2014-04-07  Paul Eggert  <>

	grep: prefer bool in DFA internals
	* src/dfa.c (bool_bf): New type.
	(dfa_state): Use it, as this seems to generate slightly better
	code with GCC.
	(struct mb_char_classes, struct dfa, equal, case_fold, dfasyntax)
	(laststart, parse_bracket_exp, lex, dfaparse, dfaanalyze, dfastate)
	(match_mb_charset, dfamust):
	Use bool for boolean.
	(using_utf8) [!HAVE_LANGINFO_CODESET]: Tune.
	(dfaanalyze): Prefer & to && and | to || on booleans; it's simpler here.
	(dfastate): Simplify charclass nonzero testing.  Redo has_mbcset
	test so that the compiler's more likely to optimize it.

2014-04-07  Norihiro Tanaka  <>

	grep: prefer regex to DFA for ANYCHAR in multibyte locales
	* src/dfa.c (dfa_state): New member has_mbcset.
	Rename backref to has_backref, and make it of type bool too.
	All uses changed.
	(state_index, dfastate): Initialize new member.
	(dfaexec): Prefer regex to DFA for ANYCHAR in multibyte locales.

2014-04-07  Paul Eggert  <>

	grep: remove trival_case_ignore
	This optimization is no longer needed, given the other
	optimizations recently installed.  Derived from a patch by
	Norihiro Tanaka; see <>.
	* bootstrap.conf (gnulib_modules): Remove assert-h.
	* src/dfa.c (CASE_FOLDED_BUFSIZE): Move here from dfa.h.
	Remove now-unnecessary static assert.
	(case_folded_counterparts): Now static.
	* src/dfa.h (CASE_FOLDED_BUFSIZE, case_folded_counterparts):
	Remove decls; no longer public.
	* src/dfasearch.c (kwsmusts): Use kwset even if fill MB_CUR_MAX > 1
	and case-insensitive.
	* src/grep.c (MBRTOWC, WCRTOMB): Remove.
	(fgrep_to_grep_pattern): Use mbrtowc, not MBRTOWC.
	(trivial_case_ignore): Remove; this optimization is no longer needed.
	All uses removed.

	grep: simplify memory allocation in kwset
	* src/kwset.c: Include kwset.h first, to check its prereqs.
	Include xalloc.h, for xmalloc.
	(kwsalloc): Use xmalloc, not malloc, so that the caller need not
	worry about memory allocation failure.
	(kwsalloc, kwsincr, kwsprep): Do not worry about obstack_alloc
	returning NULL, as that's not possible.
	(kwsalloc, kwsincr, kwsprep, bmexec, cwexec, kwsexec, kwsfree):
	Omit unnecessary conversion between struct kwset * and kwset_t.
	(kwsincr, kwsprep): Return void since memory-allocation failure is
	not possible now.  All uses changed.
	* src/kwset.h: Include <stddef.h>, for size_t, so that this
	include file doesn't require other files to be included first.

	grep: minor cleanups for Galil speedups
	* src/kwset.c: Update citations.
	Include stdbool.h.
	(kwsincr, kwsprep): Clarify by using C99 decls after statements.
	(kwsprep): Clarify by using MIN.  Avoid a couple of buffer copies
	when !TRANS.
	(bmexec): Use bool for boolean.  Prefer "continue;" to ";".

2014-04-07  Norihiro Tanaka  <>

	grep: use the Galil rule for Boyer-Moore algorithm in KWSet
	The Boyer-Moore algorithm is O(m*n), which means it may be much
	slower than the DFA.  Its Galil rule variant is O(n) and increases
	efficiency in the typical case; it skips sections that are known
	to match and does not compare more than once for a position in the text.
	To use the Galil rule, look for the delta2 shift at each position
	from the trie instead of the 'mind2' value.
	* src/kwset.c (struct kwset): Replace member 'mind2' with 'shift'.
	(kwsprep): Look for the delta2 shift.
	(bmexec): Use it.

2014-04-06  Paul Eggert  <>

	grep: cleanup DFA superset optimization
	* src/dfa.c (dfa_charclass_index): New function, with body of
	old dfa_charclass but with an extra parameter D.
	(charclass_index): Reimplement in terms of dfa_charclass_index.
	(dfahint): Clarify.
	(dfasuperset): Do not assign to 'dfa' static variable.  Instead,
	use a local, and use the new dfa_charclass_index function.  This
	doesn't fix any bugs, but it's clearer.  Initialize a few more
	members, to simplify dfafree.  Copy the charclasses with
	just one memcpy call.  Don't assign nonnull to D->superset until
	it's known to be valid; that's simpler.
	(dfafree, dfaalloc): Simplify based on dfasuperset initializations.
	* src/dfa.h (dfahint): Add comment.
	* src/dfasearch.c (EGexecute): Simplify use of memchr.
	Simplify by using memrchr.  Fix typo that could cause a buffer
	read overrun.

2014-04-06  Norihiro Tanaka  <>

	grep: optimization with the superset of DFA
	The superset of a DFA is like the DFA, except that for speed
	ANYCHAR, MBCSET and BACKREF are replaced by (CSET full bits) STAR,
	and mb_cur_max is 1.  For example, for 'a\(b\)c\1':
	  original: a b CAT c CAT BACKREF CAT
	  superset: a b CAT c CAT CSET STAR CAT  (The CSET has all bits set.)
	If a string matches a DFA, it matches the DFA's superset.
	Using the superset to filter can dramatically improve performance,
	over 200x in some cases.  See <>.
	* src/dfa.c (struct dfa): New member 'superset'.
	(dfahint, dfasuperset): New functions.
	(dfacomp): Create and analyze the superset.
	(dfafree): Free only non-NULL items.
	(dfaalloc): Initialize superset member.
	(dfaoptimize): If succeed in optimization for UTF-8 locale, don't use
	the superset.
	* src/dfa.h (dfahint): New decl.
	* src/dfasearch.c (EGexecute): Use dfahint.

2014-04-06  Jim Meyering  <>

	build: avoid OS X 10.8.5 build failure due to lack of static_assert
	* bootstrap.conf (gnulib_modules): Add assert-h, to accommodate the
	new use of static_assert on systems lacking support for that construct.
	Without this change, compilation of dfa.c failed on OS X 10.8.5 with
	gcc-4.9.0 20140324.  We should be using gnulib's assert-h module,
	regardless, for its nominal improved portability, since grep includes
	assert.h and uses assert.

2014-04-05  Norihiro Tanaka  <>

	grep: fix performance bug with regex in line-by-line mode
	* src/dfasearch.c (EGexecute): Match line-by-line with regex.

2014-04-05  Paul Eggert  <>

	grep: minor improvements to previous patch
	* src/dfa.c (MAX): New macro.
	(match_anychar, match_mb_charset, transit_state_consume_1char):
	Use it to simplify assignments.
	(SKIP_REMAINS_MB_IF_INITIAL_STATE): Prefer != 0 for unsigned.
	(free_mbdata): Omit an unnecessary 'free'.

2014-04-05  Norihiro Tanaka  <>

	grep: reuse multibyte DFA buffers in non-UTF8 locales
	* src/dfa.c (struct dfa): New members 'mblen_buf', 'nmblen_buf',
	'inputwcs', 'ninputwcs', 'mb_follows' and 'mb_match_lens'.
	(mblen_buf, inputwcs): Remove static vars.
	(SKIP_REMAINS_MB_IF_INITIAL_STATE, match_anychar, match_mb_charset)
	(transit_state_consume_1char, transit_state, prepare_wc_buf):
	Use new members instead of global variables.
	(check_matching_with_multibyte_ops): Use new members
	instead of new allocation.
	(dfaexec): Initialize new members.
	(free_mbdata): Free new members.

2014-04-05  Paul Eggert  <>

	grep: simplify dfa.c by having it not include mbsupport.h directly
	* src/mbsupport.h: Remove.
	* src/ (noinst_HEADERS): Remove mbsupport.h.
	* src/dfa.c, src/grep.c, src/search.h: Don't include mbsupport.h.
	* src/dfa.c: Include wchar.h and wctype.h unconditionally, as
	this simplifies the use of dfa.c in grep, and it does no harm
	in gawk.
	(setlocale, static_assert): Remove gawk-specific hacks, as
	gawk now does these itself.
	(struct dfa, dfambcache, mbs_to_wchar)
	(is_valid_unibyte_character, setbit_wc, using_utf8, FETCH_WC)
	(addtok_wc, add_utf8_anychar, atom, state_index, epsclosure)
	(dfaanalyze, dfastate, prepare_wc_buf, dfaoptimize, dfafree, dfamust):
	* src/dfasearch.c (EGexecute):
	* src/grep.c (main):
	* src/searchutils.c (mbtoupper):

2014-04-01  Norihiro Tanaka  <>

	dfa: avoid re-building a state built previously
	* src/dfa.c (dfaexec): Avoid to re-build a state built previously.

2014-03-28  Paul Eggert  <>

	dfa: improve port to freestanding DJGPP
	Suggested by Aharon Robbins (Bug#17056).
	* src/dfa.c (setlocale) [!LC_ALL]: Return NULL, not "C",
	reverting part of a recent change.
	(using_simple_locale): Return true if setlocale returns null.

2014-03-28  Jim Meyering  <>

	tests: placate "make syntax-check" re compare arg ordering
	* tests/euc-mb: Reverse order of arguments to compare.
	Be consistent in ordering compare arguments: expected followed
	by actual.

2014-03-28  Paul Eggert  <>

	dfa: avoid an indirection and port wint_t usage
	* src/dfa.c (struct dfa): Put mbrtowc_cache directly into struct dfa
	rather than having a pointer; this saves a malloc and an indirection.
	All uses changed.
	(dfambcache): Port to hosts where wint_t * can't be cast to wchar_t *.

2014-03-28  Norihiro Tanaka  <>

	grep: take mbrtowc_cache into new member of struct dfa
	When struct dfa more than one are used at the same time, mbrtowc cache
	may be conflict.  So, take mbrtowc_cache into new member of struct dfa,
	and define each mbrtowc cache for them.

	* src/dfa.c (struct dfa): New member `mbrtowc_cache'.
	(dfambcache): Rename from build_mbrtowc_cache.  Add dependency on struct dfa.
	(mbs_to_wchar): Add dependency on struct dfa.
	(FETCH_WC): Use it.
	(prepare_wc_buf): Use it.  Add dependency on struct dfa.
	(dfacomp): Call it.
	(dfafree): Release it.

2014-03-28  Paul Eggert  <>

	dfa: cache results of mbrtowc for speed
	Idea suggested by Norihiro Tanaka in Bug#16842.
	* src/dfa.c (mbrtowc_cache): New static var.
	(build_mbrtowc_cache, mbs_to_wchar): New functions.
	(FETCH_WC) [MBS_SUPPORT]: Speed up by using mbs_to_wchar
	instead of mbrtowc and wctob.
	(FETCH_WC) [!MBS_SUPPORT]: Rewrite in terms of old FETCH macro.
	(FETCH): Remove; no longer used.
	(lex): Simplify by avoiding the need for FETCH.
	(prepare_wc_buf) [MBS_SUPPORT]: Speed up by using mbs_to_wchar.
	Simplify the loop.
	(dfacomp): Initialize the cache.

2014-03-27  Norihiro Tanaka  <>

	grep: perform the kwset-helping DFA match in narrower range
	When kwsexec gives us the offset of a potential match, we compute
	line begin/end and then run the DFA matcher to see if there really
	is a match on that line.  When the beginning of the line, BEG, is
	not on a multibyte character boundary, advance BEG until it on such
	a boundary, before running the DFA search.
	* src/dfasearch.c (EGexecute): As above.  Add a comment.
	* tests/euc-mb: Add a test case that exercises this code.
	This addresses

2014-03-26  Jim Meyering  <>

	maint: fix "make dist"
	* src/ (egrep fgrep): Specify via
	$(srcdir)/, so non-srcdir builds work once again.

2014-03-26  Paul Eggert  <>

	dfa: improve port to freestanding DJGPP
	* src/dfa.c (setlocale) [!LC_ALL]: Return "C", not NULL (Bug#17056).
	(using_simple_locale): Store setlocale result in a ptr-to-const.

	egrep, fgrep: improve diagnostics from shell scripts
	This should fix Bug#17098.
	* src/ (EXTRA_DIST): Add
	(egrep fgrep): Depend on and Makefile.
	Build from new file, as this makes the build process
	easier to follow.  Arrange for $0 to look nicer in subgrep.
	* src/ New file.

2014-03-23  Paul Eggert  <>

	dfa: avoid undefined behavior
	* src/dfa.c (FETCH_WC, addtok_wc): Don't rely on undefined behavior
	when converting an out-of-range value to 'int'.
	(FETCH_WC, prepare_wc_buf): Don't rely on conversion state after
	mbrtowc returns a special value, as it's undefined for (size_t) -1.
	(prepare_wc_buf): Simplify test for valid character.

	grep: fix and simplify grep -iF optimization
	* src/grep.c (check_any_alphabets): Remove.
	(fgrep_to_grep_pattern): Fix problems when mbrtowc returns -1 or -2.
	Simplify a bit.
	(main): Don't bother optimizing 'grep -iF PAT' when PAT contains no
	alphabetics; it's so rare it's not worth the complexity.

2014-03-23  Norihiro Tanaka  <>

	grep: optimization for fgrep with changing the macher to grep macher.
	fgrep macher is only use kwset engine.  However, it's very slow for
	case-insensitive matching in multibyte locales.

	And so, if the matcher is fgrep and case-insensitive and keys including
	any alphabets, change it into grep matcher by escape of keys.  OTOH, if
	keys include no alphabet, turn match_icase flag off.

	I prepare following string to measure the performance.

	    yes $(printf '%078dm' 0)| head -1000000 | tr 0 a > in
	    A=`printf '\xef\xbc\xa1'`   # FULLWIDTH LATIN CAPITAL LETTER A

	I run three tests with this patch (best-of-5 trials):

	    env LC_ALL=en_US.UTF-8 time -p src/fgrep -i "$A" in
	        real 8.54       user 7.13       sys 1.16

	Back out that commit (temporarily), recompile, and rerun the experiment:

	    env LC_ALL=en_US.UTF-8 time -p src/fgrep -i "$A" in
	        real 0.07       user 0.02       sys 0.05

	* src/fgrep.c (Gcompile) New function.
	* src/main.c (check_any_alphabets) New function.
	(fgrep_to_grep_pattern) New function.
	(main) Use them.

2014-03-23  Paul Eggert  <>

	egrep, fgrep: go back to shell scripts
	Although egrep's and fgrep's switch from shell scripts to
	executables may have made sense in 2005, it complicated
	maintenance and recently has caused subtle performance bugs.
	Go back to the old way of doing things, as it's simpler and more
	easily separated from the mainstream implementation.  This should
	be good enough nowadays, as POSIX has withdrawn egrep/fgrep and
	portable applications should be using -E/-F anyway.
	* po/ Remove src/egrep.c, src/fgrep.c, src/main.c.
	* src/ (bin_PROGRAMS): Remove egrep, fgrep.
	(bin_SCRIPTS): New macro.
	(grep_SOURCES): Move searchutils.c, dfa.c, dfasearch.c, kwset.c,
	kwsearch.c, pcresearch.c here from libgrep_a_SOURCES.
	(egrep_SOURCES, fgrep_SOURCES, noinst_LIBRARIES, libgrep_a_SOURCES):
	(LDADD): Remove libgrep.a.
	(egrep, fgrep): New rules.
	(CLEANFILES): New macro.
	* src/grep.c: Rename from src/main.c.
	(usage, setmatcher, main):
	Simplify, since there's now just one executable.
	(Gcompile, Ecompile, Acompile, GAcompile, PAcompile, matchers):
	Move here from the (removed) src/grep.c.
	(compile_fp_t, execute_fp_t, struct matcher, matchers):
	Move here from src/grep.h, as they no longer need to be public.
	(struct Avoid one level of indirection/relocation.
	(do_execute, main): Fix a performance bug when it was compiled
	as 'fgrep', due to confusion about which matcher was which.
	(main): Fix a performance bug with -P, likewise.
	* src/grep.h (before_options, after_options): Remove.
	* src/egrep.c, src/fgrep.c, src/grep.c: Remove.

	dfa: port to freestanding DJGPP (Bug#17056)
	* src/dfa.c (setlocale) [!LC_ALL]: Define a dummy.

2014-03-16  Jim Meyering  <>

	tests: avoid false-positive failure on some AMD CPUs
	* tests/mb-non-UTF8-performance: Avoid false-positive failure
	when run on certain AMD processors.

2014-03-10  Jim Meyering  <>

	tests: make a performance-measuring test less system-sensitive
	Andreas Schwab reported in
	that this test would timeout and fail on m68k-suse-linux.
	Rather than testing absolute duration with a limit tuned
	to today's hardware, compare performance of grep with LC_ALL=C
	against that same command using LC_ALL=ja_JP.eucJP.
	* tests/init.cfg (require_hi_res_time_): New function.
	* tests/mb-non-UTF8-performance: Rewrite to use it:
	record absolute duration D of the first (normally much faster)
	command, and set a timeout of 8*D for the command running in
	an affected locale.

2014-03-09  Paul Eggert  <>

	maint: pacify 'make dist'
	* src/dfa.c (parse_bracket_exp): Reindent with spaces.
	* src/dfa.h (case_folded_counterparts): Prefix decl with 'extern'.
	* src/main.c: Don't include assert.h.

2014-03-07  Paul Eggert  <>

	fgrep: fix case-fold incompatibility with plain 'grep'
	fgrep converted to lowercase, whereas the regex code converted
	to uppercase.  The resulting behaviors don't agree in offbeat
	cases like Greek sigmas and Turkish Is.  Fix this by changing
	fgrep to agree with the regex code.
	* src/kwsearch.c (Fcompile, Fexecute):
	* src/searchutils.c (kwsinit, mbtoupper):
	Convert to uppercase, not to lowercase, for compatibility with
	plain 'grep'.
	* src/search.h, src/searchutils.c (mbtoupper):
	Rename from mbtolower, since it now converts to uppercase.
	All uses changed.
	* tests/case-fold-titlecase: Add tests for this.

	grep: fix case-fold mismatches between DFA and regex
	The DFA code and the regex code didn't use the same semantics for
	case-folding.  The regex code says that the data char d matches
	the pattern char p if uc (d) == uc (p).  POSIX is unclear in this
	area; the simplest fix for now is to change the DFA code to agree
	with the regex code.  See <>.
	* src/dfa.c (static_assert): New macro, if not already defined.
	(setbit_case_fold_c): Assume MB_CUR_MAX is 1 and that case_fold
	is nonzero; all callers changed.
	(setbit_case_fold_c, parse_bracket_exp, lex, atom):
	Case-fold like the regex code does.
	(lonesome_lower): New constant.
	(case_folded_counterparts): New function.
	(parse_bracket_exp): Prefer plain setbit when case-folding is
	not needed.
	* src/dfa.h (CASE_FOLDED_BUFSIZE): New constant.
	(case_folded_counterparts): New function decl.
	* src/main.c (trivial_case_ignore): Case-fold like the regex code does.
	(main): Try to improve comment re trivial_case_ignore.
	* tests/case-fold-titlecase: Add lots more test cases.

2014-03-06  Paul Eggert  <>

	build: update gnulib submodule to latest

	doc: do not overpromise --ignore-case's behavior
	* NEWS: Omit vague statement about titlecase that could be
	misinterpreted, and is more trouble than it's worth.
	* doc/grep.texi: Add @documentencoding.  Fix copyright range to
	use endash not hyphen.
	(Matching Control): Do not overpromise what --ignore-case will do.
	Give examples of corner cases where the documentation does not
	specify behavior.

2014-03-05  Paul Eggert  <>

	maint: remove differences from gnulib regex code
	These don't seem to be needed with GCC 4.8.2, and are making
	maintenance harder.  If we need to disable warnings with older
	compilers, we can add pragmas to the gnulib versions.  See
	* gl/lib/regcomp.c.diff, gl/lib/regex_internal.c.diff:
	* gl/lib/regex_internal.h.diff, gl/lib/regexec.c.diff:
	* (exclude_file_name_regexp--sc_prohibit_tab_based_indentation):
	Don't mention gl/* files.

2014-03-03  Paul Eggert  <>

	grep: fix comment
	* src/main.c (trivial_case_ignore): Fix comment typo.

2014-03-03  Norihiro Tanaka  <>

	grep: avoid to add same character to a bracket expression
	* src/main.c (trivial_ignore_case): Only when uppercase and/or
	lowercase is different from original character, add it to new pattern.

2014-03-02  Paul Eggert  <>

	grep: fix some unlikely bugs in trivial_case_ignore
	* src/main.c (MBRTOWC, WCRTOMB): Reformat as per usual GNU style.
	(trivial_case_ignore): Don't overrun buffer in the unusual case
	when a character has both lowercase and uppercase counterparts.
	Don't rely on undefined behavior when assigning out-of-range value
	to an 'int'.  Simplify by avoiding unnecessary buffer copies.
	Work even with shift encodings, by using mbsinit to
	disable the optimization if we are not in the initial state
	when we replace B by [BCD].

2014-03-02  Norihiro Tanaka  <>

	grep: revert removal of trivial_case_ignore
	Revive trivial_case_ignore function in order to be able to use kwset.

	* src/main.c (MBRTOWC, WCRTOMB): New macros.
	(trivial_case_ignore): New function.
	(main): Use it.

2014-03-02  Norihiro Tanaka  <>

	grep: optimization of bracket expression for non-UTF8 locales
	* src/dfa.c (addtok): Replace an MBCSET with a CSET even in
	non-UTF8 locales, and even when it has individual characters.

2014-03-01  Paul Eggert  <>

	doc: describe titlecase fix better
	* NEWS: Document behavior on lowercase text too.
	Suggested by Eric Blake in <>.
	* doc/grep.texi (Matching Control): Specify behavior of -i
	more precisely.

2014-02-28  Paul Eggert  <>

	grep: minor tuning for mb_case_map_apply
	* src/kwsearch.c (mb_case_map_apply): Avoid unnecessary widening of
	size_t to intmax_t.  Avoid unnecessary reinitialization of k.

	grep: avoid 'inline' when it doesn't matter
	These days, compilers generally do just fine without advice from
	users about 'inline', and there's little need for 'static inline',
	just as there's little need for 'register'.
	* src/dfa.c (to_uchar):
	* src/dosbuf.c (guess_type, undossify_input, dossified_pos):
	* src/main.c (undossify_input):
	No longer inline.
	* src/search.h (mb_case_map_apply): Move from here ...
	* src/kwsearch.c (mb_case_map_apply): ... to here, and
	make it no longer 'inline'.

	grep: fix bugs with -i and titlecase
	* NEWS: Document this.
	* src/dfa.c (setbit_wc): Simplify.
	(setbit_c): Remove; no longer used.
	(setbit_case_fold_c, parse_bracket_exp, atom):
	Don't mishandle titlecase.  For 'atom', this removes the need for
	the refactoring of Bug#16729.
	(lex): Use the slower approach only for letters that have a
	differing case.
	* tests/case-fold-titlecase: New file.
	* tests/ (TESTS): Add it.

	grep: remove lint
	* src/main.c (MBRTOWC, WCRTOMB): Remove no-longer-used macros.

2014-02-28  Norihiro Tanaka  <>

	grep: remove trivial_case_ignore
	* src/main.c (trivial_case_ignore): Remove.
	(main): Remove its use; this optimization is no longer needed.

	grep: don't match line-by-line for case-insensitive with grep and awk
	* src/main.c (matcher): Move decl up.
	(do_execute): With the grep or awk matchers,
	no need to match line by line.

2014-02-27  Jim Meyering  <>

	maint: dfa: pass NULL, not 0, as 2nd arg to setlocale
	* src/dfa.c (using_simple_locale): Use NULL, not 0.

2014-02-27  Paul Eggert  <>

	* src/dfa.c (prednames): POSIX allows [[:xdigit:]] to match multibyte chars.

	* src/dfa.c (parse_bracket_exp): Parenthesize.

	grep: fix multiple bugs with bracket expressions
	* NEWS: Document this.
	* src/dfa.c (using_simple_locale): New function.
	(parse_bracket_exp): Handle bracket expressions like [a-[.z.]]
	correctly.  Don't assume that dfaexec handles expressions like
	[^a-z] correctly, as they can match multiple characters in some
	* tests/posix-bracket: New file.
	* tests/ (TESTS): Add it.

2014-02-25  Stephane Chazelas  <>

	align grep -Pw with grep -w
	For the -w option, with -P, we used to look for the pattern surrounded by
	word boundaries. That's different from what grep -w does and what the
	documentation describes. Now align with grep -w and the documentation by
	using PCRE look-behind and look-ahead operators to match the pattern if
	it is not surrounded by word constituents.
	* src/pcresearch.c (Pcompile): Use (?<!\w)(?:...)(?!\w) rather than
	* NEWS (Bug fixes): Mention it.
	* tests/pcre-w: New file.
	* tests/ (TESTS): Add it.
	This complements the fix for

2014-02-24  Stephane Chazelas  <>

	grep -P: fix it so backreferences now work with -w and -x
	To implement -w and -x, we bracket the search term with parentheses.
	However, that set of parentheses had the default semantics of
	"capturing", i.e., creating a backreferenceable matched quantity.
	Instead, use (?:...), to create a non-capturing group.
	* src/pcresearch.c (Pcompile): Use (?:...) rather than (...).
	* NEWS (Bug fixes): Mention it.
	* tests/pcre-wx-backref: New file.
	* tests/ (TESTS): Add it.
	This addresses

2014-02-20  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.18
	* NEWS: Record release date.

	tests: test for the non-UTF8 multi-byte performance regression
	Test for the just-fixed performance regression.
	With a 100-200x differential, it is reasonable to expect that
	a very slow system will be able to complete the designated
	task in a few seconds, while with the bug, even a very fast
	system would exceed the timeout.
	* tests/mb-non-UTF8-performance: New file.
	* tests/ (TESTS): Add it.
	* tests/init.cfg (require_JP_EUC_locale_): New function.

	grep -i: avoid a performance regression in multibyte non-UTF8 locales
	* src/main.c: Include dfa.h.
	(trivial_case_ignore): Perform this optimization only for UTF8 locales.
	This rectifies a 100-200x performance regression in non-UTF8 multi-byte
	locales like ja_JP.eucJP.  The regression was introduced by the 10x
	UTF8/grep-i speedup, commit v2.16-4-g97318f5.
	* NEWS (Bug fixes): Mention it.
	Reported by Norihiro Tanaka in

	maint: give dfa.c's using_utf8 function external scope
	* src/dfa.c (using_utf8): Remove "static inline".
	* src/dfa.h (using_utf8): Declare it.
	* src/searchutils.c (is_mb_middle): Use using_utf8 rather than
	rolling our own.

2014-02-20  Paul Eggert  <>

	tests: test [^^-^] in unibyte locales
	This is a bug in the current dfa.c, which was reintroduced by the
	recent reversion from RRI.
	* tests/unibyte-negated-circumflex: New file.
	* tests/ (TESTS): Add it.
	* tests/init.cfg (require_unibyte_locale): New function.

	grep: fix bug with patterns like [^^-~] in unibyte locales
	* NEWS: Document this.
	* src/dfa.c (parse_bracket_exp): Escape patterns like [^^-~], or
	Awk patterns like [\^-\]], so that they are not misinterpreted by
	the system regex library.  Check for system regex failure due to
	memory exhaustion.

2014-02-17  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.17
	* NEWS: Record release date.

2014-02-17  Paolo Bonzini  <>

	revert "grep: DFA now uses rational ranges in unibyte locales"
	The correct course of action for grep is to defer range interpretation
	to regex, because otherwise you can get mismatches between regexes with
	backreferences and those without.

	For example, [A-Z]. will use RRI but ([A-Z])\1 won't, with the confusing
	result that the first regex won't match a superset of the language
	described by the second regex.

	The source of the confusion is that, even though grep's dfa.c was changed
	to use range checking instead of strcoll, that code is only invoked if
	dfaexec is called with backref = NULL, and that never happens for grep!

	In the end, all that's needed for RRI is compiling --with-included-regex,
	and in that case the patch is almost a no-op.  Almost, because there
	are corner cases that aren't handled correctly (e.g. [a-[.e.]], or
	regular expressions that include a NUL character), but this can be
	handled separately.

	* NEWS: Revert paragraph introduced by commit v2.16-7-g1078b64.
	* src/dfa.c (parse_bracket_exp): Revert back to regcomp/regexec.

2014-02-16  Mike Frysinger  <>

	maint: ignore configure.lineno
	* .gitignore: Add configure.lineno.

2014-02-11  Benno Schulenberg  <>

	help: remove surplus newline
	* src/main.c (usage): Remove inconsistent \n introduced by previous

2014-02-10  Benno Schulenberg  <>

	help: fix a line ending, and use the same word for similar things
	* src/main.c (usage): Change a stray 'n' to a newline, and use
	the word "display" for showing version info as for help text.

2014-02-09  Norihiro Tanaka  <>

	speed up mb-boundary-detection after each preliminary match
	After each kwsexec or dfaexec match, we must determine whether
	the tentative match falls in the middle of a multi-byte character.
	That is what our is_mb_middle function does, but it was expensive,
	even when most input consisted of single-byte characters.  The main
	cost was for each call to mbrlen.  This change constructs and uses
	a cache of the lengths returned by mbrlen for unibyte values.
	The largest speed-up (3x to 7x, CPU-dependent) is when most
	lines contain a match, yet few are printed, e.g., when using
	grep -v common-pattern ... to filter out all but a few lines.

	* src/search.h (build_mbclen_cache): Declare it.
	* src/main.c: Include "search.h".
	[MBS_SUPPORT] (main): Call build_mbclen_cache in a multibyte locale.
	* src/searchutils.c [HAVE_LANGINFO_CODESET]: Include <langinfo.h>.
	(mbclen_cache): New global.
	(build_mbclen_cache): New function.
	(is_mb_middle) [HAVE_LANGINFO_CODESET]: Use it.
	* NEWS (Improvements): Mention it.

2014-02-01  Jim Meyering  <>

	maint: use to_uchar function rather than explicit casts
	* src/system.h (to_uchar): Define function.
	* src/kwsearch.c (Fexecute): Use to_uchar twice in place of casts.
	* src/dfasearch.c (EGexecute): Likewise.
	* src/main.c (prepend_args): Likewise.
	* src/kwset.c (U): Define in terms of to_uchar.
	* src/dfa.c (match_mb_charset): Use to_uchar, not an explicit cast.

2014-01-27  Jim Meyering  <>

	maint: remove vestiges of support for long-disabled --mmap option
	This option was disabled in March of 2010, and began to elicit a
	warning in January of 2012.  Its time has come.
	* doc/ Remove mention.
	* doc/grep.texi: Likewise.
	(long_options, main): Remove all traces.
	* tests/ (check_PROGRAMS): Remove mention of ignore-mmap.
	* tests/ignore-mmap: Remove file.
	* NEWS (Maintenance): Mention it.

2014-01-26  Jim Meyering  <>

	maint: move two local variable declarations
	* src/dfasearch.c (kwsmusts): Move one declaration down to the point
	of definition.  Move another into the sole scope where it is used.

2014-01-26  Norihiro Tanaka  <>

	dfasearch: skip kwset optimization when multi-byte+case-insensitive
	Now that DFA searching works with multi-byte locales, the only remaining
	reason to case-convert the searched input is the kwset optimization.
	But multi-byte case-conversion is so expensive that it's not
	worthwhile even to attempt that optimization.

	* src/dfasearch.c (kwsmusts): Skip this function in ignore-case mode
	when the locale is multi-byte.
	(EGexecute): Now that this code need not handle multi-byte case-ignoring
	matches, remove the expensive copy/case-conversion code.
	With no case-converted buffer, there is no longer any need to call
	mb_case_map_apply, so remove it and associated code.
	(kwsincr_case): Remove function.  Now, every use of this function
	is equivalent to a use of kwsincr.  Replace all uses.
	* tests/turkish-eyes: Test all of -E, -F and -G.

2014-01-25  Norihiro Tanaka  <>

	dfa: remove GREP-ifdef'd code in favor of code used by gawk
	For many years, gawk and grep have used different #ifdef'd bits of
	code relating to how the DFA matcher matches multibyte characters.
	Remove the GREP-specific code in favor of the code gawk uses.  This
	permits us to avoid still more cases in which grep must resort to
	the expensive process of copying/case-converting each input line
	before matching against a case-converted regexp.
	* src/dfa.c (parse_bracket_exp, atom): As above.

2014-01-25  Jim Meyering  <>

	gnulib: update to latest

2014-01-17  Paul Eggert  <>

	grep: DFA now uses rational ranges in unibyte locales
	Problem reported by Aharon Robbins in <>.
	* NEWS:
	* doc/grep.texi (Environment Variables)
	(Character Classes and Bracket Expressions):
	Document this.
	* src/dfa.c (parse_bracket_exp): Treat unibyte locales like multibyte.

2014-01-17  Aharon Robbins  <>

	grep: add undocumented '-X gawk' and '-X posixawk' options
	See <>.
	* src/grep.c (GAcompile, PAcompile): New functions.
	(const): Use them.

2014-01-10  Pádraig Brady  <>

	tests: remove superfluous uses of printf
	* tests/turkish-eyes: Remove unnecessary uses of printf.

2014-01-09  Jim Meyering  <>

	grep: make --ignore-case (-i) faster (sometimes 10x) in multibyte locales
	These days, nearly everyone uses a multibyte locale, and grep is often
	used with the --ignore-case (-i) option, but that option imposes a very
	high cost in order to handle some unusual cases in just a few multibyte
	locales.  This change gets most of the performance of using LC_ALL=C
	without eliminating the ability to search for multibyte strings.

	With the following example, I see an 11x speed-up with a 2.3GHz i7:
	Generate a 10M-line file, with each line consisting of 40 'j's:

	    yes jjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjjj | head -10000000 > k

	Time searching it for the simple/noexistent string "foobar",
	first with this patch (best-of-5 trials):

	    LC_ALL=en_US.UTF-8 env time src/grep -i foobar k
	        1.10 real         1.03 user         0.07 sys

	Back out that commit (temporarily), recompile, and rerun the experiment:

	    git log -1 -p|patch -R -p1; make
	    LC_ALL=en_US.UTF-8 env time src/grep -i foobar k
	        12.50 real        12.41 user         0.08 sys

	The trick is to realize that for some search strings, it is easy
	to convert to an equivalent one that is handled much more efficiently.
	E.g., convert this command:

	  grep -i foobar k

	to this:

	  grep '[fF][oO][oO][bB][aA][rR]' k

	That allows the matcher to search in buffer mode, rather than having to
	extract/case-convert/search each line separately.  Currently, we perform
	this conversion only when search strings contain neither '\' nor '['.
	See the comments for more detail.

	* src/main.c (trivial_case_ignore): New function.
	(main): When possible, transform the regexp so we can drop the -i.
	* tests/turkish-eyes: New file.
	* tests/ (TESTS): Use it.
	* NEWS (Improvements): Mention it.

2014-01-07  Paul Eggert  <>

	tests: port Solaris 10 /bin/sh patch back to GNU/Linux
	Problem reported by Jim Meyering.
	* tests/bre, tests/ere, tests/spencer1-locale:
	Prefer re_shell, not re_shell_.
	* tests/ (re_shell): New var, which is exported instead of

	Port to Solaris 10 /bin/sh.
	Problem reported by Dagobert Michelsen in <>.
	* tests/bre, tests/ere, tests/spencer1-locale:
	Prefer re_shell_ to SHELL, if re_shell_ is set.
	* tests/ (re_shell_): Export if it's used.

2014-01-01  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.16
	* NEWS: Record release date.

	gnulib: update to latest, for fix

	maint: update copyright dates for 2014
	Do that by running "make update-copyright".

	gnulib: update to latest

2013-12-31  Jim Meyering  <>

	pcre: use PCRE_NO_UTF8_CHECK properly
	In order to obtain the behavior we want, i.e., to disable
	error-on-invalid-UTF-in-input, apply this PCRE option in
	pcre_exec, not when compiling.
	* src/pcresearch.c (Pexecute): Use PCRE_NO_UTF8_CHECK here, ...
	(Pcompile): ...rather than here.
	* tests/pcre-invalid-utf8-input: Adjust test case to test for this.

2013-12-26  Jim Meyering  <>

	maint: fix inconsistent spacing in expression
	* src/main.c (prline): Fix inconsistent spacing in expression:
	s/  / /.

2013-12-26  behoffski  <>

	maint: fix a garbled comment
	* src/dfa.c (XNMALLOC, etc.): Fix garbled comment wording.

2013-12-23  Jim Meyering  <>

	maint: fix/improve a comment
	* src/main.c (prline): Replace untrue FIXME comment with one
	telling how the hard-to-reach code can be exercised.

2013-12-21  Santiago Ruano Rincón  <>

	pcre: tell grep -P to relax its stance on invalid multibyte chars
	Do not exit-2 for invalid UTF-8 characters.  Just prior to this
	change, this command would match no lines and fail like this:
	  $ printf 'j\x82\nj\n'|LC_ALL=en_US.UTF-8 grep -P j|cat -A; echo $?
	  grep: invalid UTF-8 byte sequence in input
	After this change, the same command matches both lines, and succeeds:
	* src/pcresearch.c (Pcompile): Use PCRE_NO_UTF8_CHECK, too, and
	add a comment.
	* tests/pcre-utf8: Add a test and a comment.
	This change did not work with Debian unstable pcre-8.31-2
	or with some 8.33 and 8.34-based versions, but does work with
	Fedora 20's 8.33 and with a built-from-latest source library.
	Based on a patch by Santiago Ruano Rincón.

2013-12-21  Jim Meyering  <>

	tests: avoid FP failure due to exhausted memory
	* tests/long-line-vs-2GiB-read: Don't declare the test "failed"
	when running out of memory.  In that case, skip it.

2013-12-18  Jim Meyering  <>

	maint: add comments and split some long lines
	* src/main.c (do_execute): Add a comment.
	Split some lines longer than 80 bytes.

	pcre: avoid a nominal leak
	* src/pcresearch.c (Pcompile)[HAVE_LIBPCRE && !PCRE_STUDY_JIT_COMPILE]:
	We would leak "re" if built with HAVE_LIBPCRE but without
	PCRE_STUDY_JIT_COMPILE.  Move the free out one level.

	maint: indent cpp directives to reflect nesting
	* src/pcresearch.c: Insert spaces after a few "#", to indent
	cpp directives to reflect their nesting.

	grep: handle lines longer than INT_MAX on more systems
	When trying to exercize some long-line-handling code, I ran these
	  $ dd bs=1 seek=2G of=big < /dev/null; grep -l x big; echo $?
	  grep: big: Invalid argument
	grep should not have issued that diagnostic, and it should
	have exited with status 1, not 2.  What happened?
	grep read the 2GiB of NULs, doubled its buffer size,
	copied the 2GiB into the new 4GiB buffer, and proceeded
	to call "read" with a byte-count argument of 2^32.
	On at least Darwin 12.5.0, that makes read fail with EINVAL.
	The solution is to use gnulib's safe_read wrapper.
	* src/main.c: Include "safe-read.h"
	(fillbuf): Use safe_read, rather than bare read.  The latter
	cannot handle a read size of 2^32 on some systems.
	* bootstrap.conf (gnulib_modules): Add safe-read.
	* tests/long-line-vs-2GiB-read: New file.
	* tests/ (TESTS): Add it.
	* NEWS (Bug fixes): Mention it.

2013-11-25  Jim Meyering  <>

	tests: port to non-GNU sed
	* tests/multibyte-white-space (utf8_space_characters): The generation
	of test inputs relied on GNU sed's interpretation of \<, but that is
	not portable, and caused spurious test failures.  Adjust the sed regexp
	to work on all versions.
	Reported by Karl Dubost in

2013-11-22  Jim Meyering  <>

	maint: minor cleanup: xmalloc+strcpy -> xmemdup
	* src/main.c (main): Replace an xmalloc+strcpy combination
	with an equivalent use of xmemdup.

2013-11-21  Jim Meyering  <>
	    Paul Eggert  <>

	dfa: avoid undefined behavior of "1 << 31"
	* src/dfa.c (charclass): Change type from "int" to "unsigned int".
	(tstbit): Rather than shifting "1" left to form a mask, shift the
	LHS bits the right and use "1" as the mask.  Also, return bool, rather
	than "int".
	(setbit, clrbit, dfastate): Don't shift "1" (aka (int)1) left by 31 bits.
	Instead, use "1U" as the operand, to avoid undefined behavior.
	Spotted by gcc's new -fsanitize=undefined.

2013-11-02  Jim Meyering  <>

	grep: fix regression with -P vs. invalid UTF-8 input
	* src/pcresearch.c (Pexecute): Don't abort upon unexpected
	PCRE-specific error code.  Explicitly handle PCRE_ERROR_BADUTF8,
	and change the default to print a diagnostic including the unhandled
	integer PCRE error code and exit with status 2.
	* tests/pcre-invalid-utf8-input: New file.
	* tests/ (TESTS): Add it.
	* NEWS (Bug fixes): Mention it.
	* THANKS: Update.
	Reported by Dave Reisner in

	grep: fix regression involving \s and \S
	Commit v2.14-40-g01ec90b made \s and \S work with multi-byte
	characters, but it made it so any use like \s*, \s+, \s?, \s{3}
	would malfunction in a multi-byte locale.
	* src/dfa.c (lex): Also reset laststart.
	* tests/backslash-s-and-repetition-operators: New file.
	* tests/ (TESTS): Add it.
	* NEWS (Bug fixes): Mention it.
	* THANKS: Update.
	Reported by Mirraz Mirraz in

2013-11-01  Jim Meyering  <>

	maint: NEWS: document a release-related bug fix
	* NEWS (Bug fixes): Add an entry for a fix pulled from gnulib.

2013-10-26  Jim Meyering  <>

	build: update gnulib submodule to latest
	This pulls in a gnulib fix for that ensures the procedure
	described in README-release actually does what we want.  Before this
	change, that procedure resulted in a grep-2.15 tarball that would
	lead to a grep binary whose --version- reported version number was
	2.14.51... rather than the expected 2.15.

	maint: avoid automake deprecation warning re ACLOCAL_AMFLAGS
	* (ACLOCAL_AMFLAGS): Don't use this deprecated variable.
	* (AC_CONFIG_MACRO_DIRS): Use this instead.
	(AUTOMAKE_OPTIONS): Require automake-1.12.

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.15
	* NEWS: Record release date.

2013-10-25  Paul Eggert  <>

	build: port to AIX
	Problem reported by Pavel Kharitonov in <>.
	* src/ (LDADD): Add $(LIBTHREAD).

	build: avoid duplicate -funit-at-a-time etc. options
	* (WERROR_CFLAGS): Don't add -fdiagnostics-show-option
	and -funit-at-a-time, as Gnulib does that for us now, and we're
	merely piling on duplicats.

2013-10-24  Jim Meyering  <>

	tests: port more tests to bourne shells with hex-challenged printf
	* tests/pcre-utf8: Convert the hex \xHH literals for the euro symbol
	to octal \OOO.
	* tests/turkish-I: Likewise for "I with dot".
	* tests/turkish-I-without-dot: Likewise for another Turkish I: U+0131.

	maint: clean up an ugly 'while' condition
	* src/main.c (get_nondigit_option): Separate a slightly baroque
	"while" expression into two separate statements, both inside the loop.

2013-10-23  Jim Meyering  <>

	tests: port to bourne shells whose printf doesn't grok hex
	Use octal escapes, not hex, in printf(1) format strings,
	and in one case, use $AWK's printf so we can continue
	to use the table of hex values.
	* tests/char-class-multibyte: Use printf octal escapes, not hex,
	for portability to shells like dash and Solaris 10's /bin/sh.
	* tests/backslash-s-vs-invalid-multitype: Likewise.
	* tests/surrogate-pair: Likewise.
	* tests/unibyte-bracket-expr: Count in decimal and convert to octal.
	* tests/multibyte-white-space (hex_printf): New function.
	Use it in place of printf so we can retain the table of hex digits
	without hitting the limitation of some bourne shells.
	Reported by Paul Eggert in

2013-10-21  Jim Meyering  <>

	gnulib: update to latest

	maint: remove now-unused wcscoll module
	* bootstrap.conf (gnulib_modules): Remove wcscoll; no longer used.

2013-10-20  Paul Eggert  <>

	build: avoid chatter from Automake 1.14
	* (AM_INIT_AUTOMAKE): Add subdir-objects.

	build: port shell pattern to Solaris 10
	* Don't use unquoted '^' in a pattern, as this
	breaks 'configure' on Solaris 10, whose /bin/sh complains about it,
	which causes 'configure' to exit even before it finds a decent shell.
	Unix 7th edition shell accepted '^' as an alias for '|'.

	build: port to platforms that predefine _FORTIFY_SOURCE
	Problem reported by Brenton Hoff (Bug#15663).
	* (_FORTIFY_SOURCE): Don't define if already defined.
	This is what Emacs does.

2013-10-20  Jim Meyering  <>

	build: update gnulib submodule to latest

2013-10-19  Jim Meyering  <>

	tests: extend the multibyte-white-space test
	* tests/multibyte-white-space (utf8_space_characters): Add more
	single-byte whitespace characters.  Align RHS hex values and
	make the sed substitution less rigid, to accommodate.
	Also, ensure that grep '\S' exits with status 1.

	maint: update bootstrap to latest from gnulib
	* bootstrap: Update from gnulib.

	maint: fix typo in NEWS
	* NEWS: Fix/improve example commands in most recent entry.
	The LC_ALL envvar setting goes before grep, not before printf.
	Don't reference src/ in the second example command, and do specify
	the locale.

2013-10-09  Jim Meyering  <>

	tests: add a test for better coverage of some tricky code
	* tests/spencer1.tests: Add a non-range bracket expression representing the
	same regexp, to cover the alternate code path, the one that does not require
	a regcomp/exec call to interpret the regexp.

2013-10-01  Jim Meyering  <>

	tests: ensure neither \s nor \S matches an invalid multibyte character
	* tests/backslash-S-vs-invalid-multitype: New file.
	Prompted by the bug report from Roman at
	* tests/ (TESTS): Add it.

	dfa: fix \s and \S to work for multibyte
	* src/dfa.c (lex): In multibyte mode, we can't treat \s and \S as we do
	in single-byte mode.  Map them to [[:space:]] and [^[:space:]] respectively,
	to make the DFA matcher use the regex-matcher for this term.
	* tests/multibyte-white-space: New file.  Test for the bug.
	* tests/ (TESTS): Add it.
	This bug was introduced with the addition of DFA support
	for \s and \S in commit v2.5.4-112-gf979ca0.

2013-09-30  Jim Meyering  <>

	maint: change all references: s/POSIX\.2/POSIX/
	There is no longer any point in referring to POSIX.N.
	POSIX is sufficient.
	* doc/ As above.
	* src/main.c (main): Likewise.
	* tests/file: Likewise.
	* tests/options: Likewise.
	* ChangeLog: Likewise.
	* NEWS: Likewise.
	* Update, to match changed NEWS.
	Inspired by Glenn Golden's suggestion in

2013-09-22  Jim Meyering  <>

	dfa: remove dead disjunct
	* src/dfa.c (parse_bracket_exp): Remove dead disjunct.
	At that point, we know MB_CUR_MAX <= 1, so the test,
	MB_CUR_MAX > 1 && ... is always false.  Remove the disjunct.

	maint: dfa: improve comments and formatting
	* src/dfa.c (add_utf8_anychar): Correct wording/alignment of a comment.
	(dfaexec): Add curly braces around multi-line while statement within
	a "then" block.
	(ANYCHAR): Clarify comment: "." does not match an invalid UTF8 character.
	(parse_bracket_exp) Improve comment.

2013-09-08  Jim Meyering  <>

	dfa: appease a static analyzer, and save 95 stack bytes
	and decrease from 128 to 32.
	(parse_bracket_exp): Add one byte more than MAX_BRACKET_STRING_LEN
	to the length of "str" buffer, to avoid appearance that we may store
	the trailing NUL beyond the end of buffer.  A string of length 32
	or greater is rejected by earlier processing, so would never reach
	this code.  Addresses

2013-09-01  Corinna Vinschen  <>

	fix Cygwin UTF-16 surrogate-pair handling with -i
	grep -i would segfault on systems using UTF-16-based wchar_t (Cygwin)
	when converting an input string containing certain 4-byte UTF-8
	sequences to lower case.  The conversions to wchar_t and back to
	a UTF-8 multibyte string did not take surrogate pairs into account.
	* src/searchutils.c (mbtolower) [__CYGWIN__]: Detect and handle
	surrogate pairs when converting.
	* NEWS (Bug fixes): Mention it.
	* tests/surrogate-pair: New test.
	* tests/ (TESTS): Add it.
	Reported by: Jim Burwell

2013-08-19  Paul Eggert  <>

	doc: mention how to use the latest gnulib
	* README-hacking: Steal some text from coreutils/README-hacking.

2013-08-10  Jim Meyering  <>

	build: update gnulib-related code
	* gnulib: Update submodule to latest.
	* bootstrap: Update from gnulib.
	* gl/lib/regex_internal.h.diff: Update to reflect gnulib changes.
	* bootstrap.conf: Partial sync from coreutils.

2013-08-09  Jim Meyering  <>

	tests: simplify and factor newest test
	* tests/char-class-multibyte2: Simplify file names.
	Factor out $e_acute, so that the grep argument representation
	is ascii (though the value is still UTF8).

	doc: NEWS: mention the DFA segfault fix
	* NEWS (Bug fixes): List the DFA segfault fix.

2013-07-05  Paul Eggert  <>

	Redo comments and white space to better approach GNU style.

2013-07-05  Paolo Bonzini  <>

	tests: add testcase for previous change
	* tests/ (TESTS): add char-class-multibyte2.
	* tests/char-class-multibyte2: New file.

2013-07-05  Mike Haertel  <>

	dfa: fix multibyte character in brackets with repetition
	Let FOO stand for any multibyte (e.g. CJK character) in the regexp.
	It turns out the following much simpler regexp:
	is sufficient to cause the crash.

	In the first step of its parsing, DFA transforms regexp from human
	readable syntax into reverse-polish form.  For regexps of the form a{m,n}
	repeat counts, it simply builds repeated copies of the representation
	of a, with appropriate inserted CAT and QMARK operators.  For the above
	example with a regexp of the form a{1,2} it would build:

	        <RPN representation for a>
	        <RPN representation for a>

	When building repeated copies of RPN representations, additional
	copies of the RPN representations are made by calling a function
	copytoks() with arguments consisting of the start position and
	length of the original copy.

	The problem is that the current code for copytoks() is simply
	incorrect.  It operates by calling addtok() for each individual
	token in the source range being copied.  But, in the particular
	case that the token being added is MBCSET, addtok():

	(1) incorrectly assumes that the character set being added to be added
	    is the one most (addtok has no argument to indicate which cset is
	    being added, so it just uses the latest one)

	(2) attempts to do some token sequence expansion into more primitive
	    operators so things like [FOO] are matched efficiently.

	Both of these assumptions are incorrect in the case that addtok()
	is being called from copytoks(): (1) is simply not true, and
	(2) is redundant--the expansion has already been done token sequence
	being copied, so there is no need to do the expansion again.

	The correct function to add exactly one token, without further expansion,
	is addtok_mb().  So here is my proposed fix, which is that copytoks()
	should never call addtok(), but instead directly call addtok_mb()
	(which is what addtok() eventually calls).

	* src/dfa.c (copytoks): Rewrite using addtok_mb directly.

2013-05-28  Jim Meyering  <>

	maint: align backslashes consistently
	* tests/ Most backslashes were aligned with TABs,
	so adjust the few that used spaces to conform.

	grep -F: avoid an infinite loop with invalid multi-byte search string
	* src/kwsearch.c (Fexecute): Avoid an infinite loop when processing
	a fixed (-F) multibyte search string that is an invalid byte sequence
	in the current locale and that matches the bytes of the input twice
	on a line.  Reported by Daisuke GOTO in
	* tests/invalid-multibyte-infloop: New test.
	* tests/ (TESTS): Add it.
	* NEWS (Bug fixes): Mention it.

2013-04-18  Paul Eggert  <>

	* (old_NEWS_hash): Update.

	doc: document EREs like a{,10}
	Problem reported by Eric Blake in
	* NEWS: Document the bug fix.
	* doc/ Restore documentation for this feature, but mention
	that it is a GNU extension.
	* doc/grep.texi (Fundamental Structure): Mention that this feature
	is a GNU extension.

2013-04-02  Paul Eggert  <>

	build: make dfa.c closer to Gawk's
	* src/dfa.c: Include <stddef.h>, not <sys/types.h>.
	stddef.h is smaller and is all we need and is portable nowadays.
	Include <wchar.h> and <wctype.h> only if MBS_SUPPORT.

2013-01-15  Paul Eggert  <>

	grep: make dfa.h standalone
	Problem reported by Aharon Robbins in
	* src/dfa.c: Include dfa.h first, so that it's tested standalone.
	No need to include <regex.h>, since we are in charge of dfa.h and
	know that it includes <regex.h>.
	* src/dfa.h: Include <regex.h> and <stddef.h>, so that it's standalone.

2013-01-11  Stefano Lattarini  <>

	build: update gettext version to 0.18.2
	* (AM_GNU_GETTEXT_VERSION): Update to 0.18.2.
	This is necessary to have the gettext-provided m4 files to use
	AC_PROG_MKDIR_P rather than AM_PROG_MKDIR_P.  This latter macro,
	planned to disappear in Automake 1.14, has already been removed
	in the development version of Automake, so that, without this
	change, grep fails to bootstrap with bleeding-edge Automake.

2013-01-11  Paul Eggert  <>

	build: update gnulib submodule to latest

2013-01-11  Stefano Lattarini  <>

	build: remove redundant use of $(INCLUDES)
	* lib/ (INCLUDES): Remove.  Automake automatically adds
	$(srcdir) and $(top_builddir) to the C preprocessor search path.
	INCLUDES is deprecated in Automake 1.13 (causing a runtime
	warning), and will be removed in Automake 1.14.

2013-01-04  Jim Meyering  <>

	build: update gnulib submodule to latest

	maint: update all copyright year number ranges
	Run "make update-copyright".

2012-11-20  Paul Eggert  <>

	grep: normalize diagnostics
	* src/pcresearch.c (Pcompile): Use similar format diagnostics
	as elsewhere, and translate them.

2012-11-19  Paul Eggert  <>

	grep: diagnose read errors from -f dir, porting to Solaris
	Problem reported by Dennis Clarke for Solaris 10 in
	* src/main.c (main): For -f F, diagnose any read errors
	encountered when reading F.
	* tests/ (XFAIL_TESTS): Remove grep-dir.
	* tests/grep-dir: Don't assume that directories cannot be read
	via fread, as POSIX allows this and it can happen on Solaris.

2012-11-09  Paolo Bonzini  <>

	pcre: add PCRE-JIT support for grep
	* NEWS: Document new feature.
	* src/pcresearch.c [PCRE_STUDY_JIT_COMPILE] (jit_stack): New.
	[PCRE_STUDY_JIT_COMPILE] (Pcompile): JIT-compile the regular expression
	and allocate a stack for it.  Based on a patch from Zoltan Herczeg.
	* THANKS: Add Zoltan to the list.

2012-10-24  Paul Eggert  <>

	build: go back to AC_PROG_CC
	* Go back to using AC_PROG_CC rather than AC_PROG_CC_STDC,
	as the latter is obsolescent and the Autoconf bug involving the former
	has been fixed.

2012-10-24  Jim Meyering  <>

	build: use AC_PROG_CC_STDC rather than AC_PROG_CC
	* Use AC_PROG_CC_STDC rather than AC_PROG_CC,
	to accommodate autoconf-2.69-37+.

	build: update gnulib submodule to latest

2012-10-23  Eric Blake  <>

	build: default to --enable-gcc-warnings in a git tree
	Anyone building from cloned sources can be assumed to have a new
	enough environment, such that enabling gcc warnings by default will
	be useful.  Tarballs still default to no warnings, and the default
	can still be overridden with --disable-gcc-warnings.
	* (gl_gcc_warnings): Set default based on environment.

2012-10-03  Jim Meyering  <>

	maint: factor out STREQ definition
	* src/main.c (STREQ): Remove definition.
	* src/pcresearch.c: (STREQ): Likewise.
	* src/system.h (STREQ): Define it here instead.

	maint: correct syntax-check failures; adjust NEWS
	* tests/pcre-utf8: Reverse order of compare arguments.
	Remove all copyright year numbers except 2012.
	Use skip_ "diagnostic...", rather than a bare "exit 77".
	* NEWS: Start with a concise description of the bug.
	* src/pcresearch.c (STREQ): Define, so that we can...
	(Pcompile): use STREQ, not strcmp.

2012-10-03  Paolo Bonzini  <>

	tests: include UTF-8 testcases for grep -P
	* tests/ (TESTS): Add pcre-utf8.
	* tests/pcre-utf8: New file.

2012-10-03  Petr Pisar  <>

	pcresearch: set UTF-8 flag correctly for UTF-8 locales
	Otherwise, Unicode properties (\p{XXX}) do not work with characters
	outside the 7-bit ASCII character set.

	* src/pcresearch.c (Pcompile): Look for UTF-8 locales and set PCRE_UTF8
	if one is found.

2012-10-03  Jaroslav Škarvada  <>

	doc: fix a formatting bug in grep.1 template
	* doc/ Insert .TP before the paragraph describing
	--dereference-recursive (-R).

2012-10-03  Jim Meyering  <>

	maint: placate gcc's -Wjump-misses-init warning
	* src/kwsearch.c (Fexecute): Replace a "goto" and "return" with
	a simple return statement, eliminating the label, since that was
	the sole use.
	* src/dfasearch.c (EGexecute): Likewise.

2012-09-01  Jim Meyering  <>

	build: update gnulib submodule to latest

2012-09-01  Eric Blake  <>

	build: work with new glibc when not optimizing
	Starting with glibc 2.15, the system headers refuse to compile
	unconditional use of FORTIFY_SOURCE if optimization is disabled
	but -Werror is in effect.

	* (FORTIFY_SOURCE): Make conditional.

2012-08-19  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.14
	* NEWS: Record release date.

2012-08-07  Jim Meyering  <>

	build: update gnulib and bootstrap

	tests: test for bug with -i and ^$ in a multi-byte locale
	* tests/empty-line-mb: New file.
	* tests/ (TESTS): Add it.

	grep -i '^$' in a multi-byte locale could report a false match
	* src/dfasearch.c (EGexecute): Do not match the sentinel "newline"
	that is appended to each buffer.
	This bug may sound like a big deal (it certainly surprised me), but
	realize that only the empty-line-matching regular expression '^$'
	can trigger it, and then only when you add the unnecessary (and
	arguably superfluous) -i, *and* run the command in a multi-byte
	locale.  Using a multi-byte locale for such a regular expression
	is also pointless, and hurts performance.
	* NEWS (Bug fixes): Mention it.
	Reported by Alexander Katassonov <>

2012-08-06  Jim Meyering  <>

	tests: fix a skip diagnostic that mentioned the wrong locale
	* tests/init.cfg (require_tr_utf8_locale_): s/en_US/tr_TR/

2012-08-02  Jim Meyering  <>

	tests: skip failing test on FS/system that lack SEEK_HOLE support
	* tests/big-hole: Test for SEEK_HOLE support.  If not available,
	skip this test.  Hence, this test is now skipped on linux-3.5.0 with
	ext4 or tmpfs.  The test runs (and passes) with at least btrfs, xfs,
	or ocfs2.
	* bootstrap.conf (gnulib_modules): Use the perl module.

2012-07-30  Jim Meyering  <>

	maint: optimize long-line processing
	* src/main.c (grep): Use memrchr rather than an open-coded loop,
	reducing the cost of the replaced code by 50% when processing very
	long lines.  If there were a rawmemrchr function (analogous to glibc's
	rawmemchr), then the performance improvement would be even greater.

2012-07-27  Paul Eggert  <>

	maint: remove stat-size
	* bootstrap.conf (gnulib_modules): Remove stat-size.
	* src/main.c: Don't include stat-size.h; no longer needed.

	grep: don't falsely report compressed text files as binary
	* NEWS: Document this.
	* src/main.c (file_is_binary): Remove the heuristic based on
	st_blocks, as it does not work for compressed file systems.
	On Solaris, it'd be cheap to test whether the file system is known
	to be uncompressed, which allow the heuristic, but Solaris has
	SEEK_HOLE so there's little point.

	grep: don't falsely report tiny text files as binary
	* NEWS: Document this.
	* src/main.c (file_is_binary): When we are already at apparent
	EOF, skip the file-size check, as some servers use zero blocks
	to store binary files.  Reported by Martin Carroll in

2012-07-26  Paul Eggert  <>

	doc: document -r/-R in man page
	* doc/ Document -r vs. -R.

2012-07-21  Jim Meyering  <>

	tests: avoid false positive upon kernel OOM-kill
	* tests/big-match (skip_diagnostic): Handle case of 139 (SIGKILL)
	with no diagnostic.

	build: update gnulib and bootstrap

	maint: fix misspellings in old ChangeLog
	* ChangeLog-2009: Fix typos.

2012-07-19  Paul Eggert  <>

	grep: fix ptrdiff/size_t clash
	Reported by Jaroslav Škarvada in <>.
	* src/dfasearch.c (EGexecute): Use size_t, not ptrdiff_t, for lengths.
	Use regoff_t to store re_match's output, and test it before converting
	it to size_t.

2012-07-06  Jim Meyering  <>

	maint: correct log typo, to reflect in generated ChangeLog
	* (gen-ChangeLog): Use --amend, now that we must
	make our first log correction.
	* build-aux/git-log-fix: New file.

2012-07-04  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.13
	* NEWS: Record release date.

	build: update gnulib submodule, bootstrap,

2012-06-17  Jim Meyering  <>

	tests: add another turkish-I-related test case
	* tests/turkish-I-without-dot: Also exercise the case in which
	the original string and the lower-case buffer have precisely
	the same length (22 bytes here), yet internal offsets do differ.

2012-06-16  Jim Meyering  <>

	grep -i: work also when converting to lower-case inflates byte count
	Commit v2.12-16-g7aa698d addressed the case in which the lower-case
	representation of an input byte occupies fewer bytes than the original.
	However, even with commit v2.12-20-g074842d, grep -i would still
	misbehave when converting a character to lower-case increased its
	byte count.  The map-manipulation code assumed that the case conversion
	could only shrink the byte count.  With the consideration that it may
	also inflate it, the deltas recorded in the map array must be signed,
	and we must account for the one-to-two-or-more mapping when the
	original-to-lower-case conversion causes the byte count to increase.
	* src/searchutils.c (mbtolower): When a lower-case character occupies
	more than one byte, set its remaining map slots to zero.  Change the
	type of the map to be signed, and compute the change in character
	byte count as new_length - old_length.
	* src/search.h: Include <stdint.h>, for decl of intmax_t.
	(mb_case_map_apply): Adjust for signed increments:
	each map entry is now signed.
	(mb_len_map_t): Define type.  Thanks to Paul Eggert for noticing
	in review that using a bare "char" as the base type would be wrong on
	systems for which it is a signed type (as with gcc's -funsigned-char).
	* src/kwsearch.c (Fcompile, Fexecute): Likewise.
	* src/dfasearch.c (kwsincr_case, EGexecute): Likewise.
	* tests/turkish-I-without-dot: New test.  Thanks to Paolo Bonzini
	for the tip that in the tr_TR.utf8 locale, mapping "I" to lower case
	increases the character's byte count.
	* tests/ (TESTS): Add it.
	* tests/init.cfg (require_tr_utf8_locale_): New function.
	* NEWS (Bug fixes): Expand the existing entry.

2012-06-12  Paul Eggert  <>

	grep: handle -i when chars differ in length but line does not
	* src/searchutils.c (mbtolower): Return the map back to the caller
	if any input character's length differs from the corresponding output
	character's, not merely if the total string length differs.
	Problem reported by Johannes Meixner in

2012-06-07  Jim Meyering  <>

	tests: extend coverage of dfa.c's match_mb_charset
	Add a test case to increase test coverage of part of dfa.c (the DFA
	matcher used by grep and gawk).  While thinking about removing the few
	remaining uses of strncpy in dfa.c, I found that none of the existing
	tests covered the 40+ lines of code at the end of match_mb_charset,
	so constructed this test case to demonstrate that it's not dead code
	* tests/dfa-coverage: New test, for improved coverage.
	* tests/ (TESTS): Add it.

2012-06-05  Jim Meyering  <>

	build: fix a subtly twisted "make distcheck" failure
	"make distcheck" would fail when, during a test build,
	an attempt to overwrite the deliberately-write-protected
	$(srcdir)/grep.pot file would fail.
	* bootstrap.conf (bootstrap_epilogue): Don't let the existence of
	a large sparse file in the build directory induce "make distcheck"
	failure.  The existence of a large sparse test file named 8T-or-so
	would make po/'s use of grep (to search for "GNU grep"
	as an indication that this is a GNU package) exit 2 without generating
	any output, which made the first xgettext use --package-name=grep,
	while that same search for "GNU grep" would succeed when run
	from a pristine from-tarball build, thus making the second
	xgettext invocation use --package-name='GNU grep'.
	That mismatch:
	-"Project-Id-Version: grep 2.12.18-1080\n"
	+"Project-Id-Version: GNU grep 2.12.18-1080\n"
	led to the attempt by's grep.pot-update rule to
	overwrite ../../grep.pot in the read-only po/ source directory.

2012-06-03  Jim Meyering  <>

	build: update gnulib submodule, bootstrap and Exempt dfa.c from the new no-strncpy test, for now.

2012-06-02  Jim Meyering  <>

	grep: fix how -i works with a match containing the Turkish I-with-dot
	Fix a long-standing problem in the way grep's -i interacts with
	data whose byte count changes when we convert it to lower case.
	For example, the UTF-8 Turkish I-with-dot (İ) occupies two bytes,
	but its lower case analog, i, occupies just one byte.  The code
	converts both search string and the haystack data to lower case,
	and then searches for the modified string in the modified buffer.
	The trouble arose when using a lowercase buffer <offset,length>
	pair to manipulate the original (longer) buffer.

	The solution is to change mbtolower to return additional information:
	a malloc'd mapping vector.  With that, the caller maps the lowercase-
	relative <offset,length> to numbers that refer to the original buffer.
	This mapping is used only when lengths actually differ, so the cost
	in general should be small.

	* src/searchutils.c (mbtolower): Add the new map parameter.
	* src/search.h (mb_case_map_apply): New function.
	* src/kwsearch.c (Fexecute): Update mbtolower caller, and upon
	success, apply the new map.
	* src/dfasearch.c (EGexecute): Likewise.
	* tests/ (XFAIL_TESTS): Remove turkish-I from this list;
	that test is no longer expected to fail.
	* NEWS (Bug fixes): Mention it.
	Reported by Ilya Basin in and later
	by Strahinja Kustudic in

2012-06-01  Paul Eggert  <>

	grep: remove unnecessary "what-if-signal?" code
	* src/main.c (fillbuf): Don't worry about EINTR when closing --
	not possible, since we're not catching signals.

2012-05-16  Paul Eggert  <>

	grep: avoid nominal integer overflow
	* src/dfa.c (add_utf8_anychar): Avoid signed integer overflow.
	Although this works on all platforms we know about, strictly
	speaking the behavior is undefined, and Sun C 5.8 warns about it.

2012-05-15  Jim Meyering  <>

	maint: avoid nit-picky syntax-check test failure; tweak big-hole test
	* NEWS: Restore deleted newline in "old" NEWS, to fix a syntax-check
	test failure.
	* tests/big-hole: Use awk, rather than a shell loop: saves 3000 lines
	of verbose shell output in the .log file.

2012-05-15  Paul Eggert  <>

	grep: sparse files are now considered binary
	* NEWS: Document this.
	* doc/grep.texi (File and Directory Selection): Likewise.
	* bootstrap.conf (gnulib_modules): Add stat-size.
	* src/main.c: Include stat-size.h.
	(usable_st_size): New function, mostly stolen from coreutils.
	(fillbuf): Use it.
	(file_is_binary): New function, which looks for holes too.
	(grep): Use it.
	* tests/ (TESTS): Add big-hole.
	* tests/big-hole: New file.

2012-05-06  Paul Eggert  <>

	maint: quote 'like this' or "like this", not `like this'
	See <>.
	* ChangeLog-2009, HACKING, NEWS, README-hacking,,
	* lib/colorize-w32.c, m4/pcre.m4:
	* src/, src/dfa.c, src/dosbuf.c, src/main.c:
	* tests/backref, tests/help-version, tests/tests:
	In commentary, quote 'like this' or "like this" rather than
	`like this' or ``like this''.
	* (old_NEWS_hash): Update due to changed old NEWS.
	* doc/grep.texi (General Output Control): Quote sample text
	with @samp, not with `...'.
	* src/main.c (usage):
	* tests/help-version: Quote 'like this' rather than `like this'
	in diagnostics.

	exclude: process exclude and include directives in order
	Also, change exclude and include directives so that they apply to
	command-line arguments too.  This restores the pre-2.6 behavior,
	and fixes a bug reported by Quentin Arce in
	* NEWS: Document this.
	* src/main.c (included_patterns): Remove.  All uses removed.
	(skipped_file): New function.
	(grepdirent): New arg command_line; all callers changed.  This is
	needed because non-command-line files can invoke fts_open, and
	their directory entries need to be distinguished from top-level
	directory entries.  Move code into the new skipped_file function.
	(grepdesc): Check whether a command-line argument should be skipped.
	(main): --include and --exclude options now share excluded_patterns
	rather than having separate variables included_patterns and
	* tests/include-exclude: Add a test to detect the fixed bug.

	build: update gnulib submodule to latest

2012-04-30  Jim Meyering  <>

	cosmetic: binary operator goes *after* the newline, when split
	* src/dfa.c (match_mb_charset): Join split lines.
	(parse_bracket_exp): Move "||" from end of first split line
	to the beginning of the continued line.
	* src/dosbuf.c (dossified_pos): Likewise, but for "&&".

	grep: -K is not an option: remove it from list
	The presence of "K" in the short-option string meant that
	an erroneous "grep -K ..." would fail with a bare Usage/Try...
	message, without the usual "invalid option -- 'K'".  With this
	removal, now grep prints the expected invalid option diagnostic.
	* src/main.c (short_options): Remove "K".
	Reported by Петр Досычев in

2012-04-29  Paolo Bonzini  <>

	dfa: small fixes to single-byte range computation
	* src/dfa.c (parse_bracket_exp): Do not call regexec with an invalid
	subject.  Move declarations before all statements.

2012-04-27  Paolo Bonzini  <>

	dfa: do not use hard-locale
	* bootstrap.conf (gnulib_modules): Remove hard-locale.
	* src/dfa.c (hard_LC_COLLATE): Remove.
	(dfaparse): Do not initialize it.
	(parse_bracket_exp): Always go through system regex matcher to find
	single byte characters matching a range.

	drop support for Makefile.boot
	* Do not distribute README-boot and Makefile.boot.
	* NEWS: Mention this change.
	* README-alpha: Do not mention README-boot and Makefile.boot.
	* Makefile.boot: Remove.
	* README-boot: Remove.

2012-04-27  Aharon Robbins  <>

	dfa: do not use strcoll to match multibyte characters in ranges
	This does not affect the behavior of grep, which always defers
	to glibc or gnulib when matching ranges.
	* src/dfa.c (match_mb_charset): Compare wc directly to the range

	dfa: include stdbool.h explicitly
	* src/dfa.c: Include stdbool.h explicitly

2012-04-23  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.12
	* NEWS: Record release date.

	build: update gnulib submodule to latest

	tests: skip annoyingly long gnulib lock tests
	* bootstrap.conf (avoided_gnulib_modules): Define.
	(gnulib_tool_option_extras): Use it.

2012-04-22  Jim Meyering  <>

	tests: avoid spurious quote-mismatch failure on OS/X
	* tests/in-eq-out-infloop: Simplify expected error output, eliminating
	expected quotes altogether, thus avoiding spurious OS/X-specific
	failure due to mismatch of multi-byte vs. single-byte quotes.

2012-04-17  Jim Meyering  <>

	build: update gnulib submodule to latest
	* bootstrap: Also update this file.

2012-04-17  Jim Meyering  <>

	grep: fix --devices=ACTION (-D) so stdin is once again exempt
	An oversight in the 2.11 changes made it so "echo x|grep x" would
	fail for those who set GREP_OPTIONS=--devices=skip.

	* src/main.c (grepdesc): Ignore skip-related options when reading
	from standard input.
	* tests/skip-device: New file.  Test for the above.
	* tests/ (TESTS): Add it.
	* doc/grep.texi (File and Directory Selection): Clarify this point,
	documenting the stdin exemption.
	* NEWS (Bug fixes): Mention it, and add a few "[fixed in ...] notes.
	Reported by Tino Keitel in,
	and forwarded to bug-grep by Aníbal Monsalve Salazar.

2012-04-13  Jim Meyering  <>

	maint: dfa: correct bogus formatting
	* src/dfa.c (transit_state, dfaexec): s/++ * VAR/++*VAR/

	maint: dfa: add/improve comments
	* src/dfa.c (transit_state_consume_1char): Note always-ignored
	return value.
	Fix typos: s/equivalent class/equivalence class/.

	maint: dfa: avoid unnecessary uses of strcpy/strncpy
	* src/dfa.c (icatalloc): Use memcpy, not strcpy, given the length.
	(dfamust): Combine MALLOC+strcpy into cleaner xmemdup.
	(parse_bracket_exp): Likewise, but replace a use of strncpy.

	grep: handle symlinked directory loops as usual
	* src/main.c (grepfile): Treat EMLINK just like ELOOP, for
	systems like FreeBSD 9.0 on which we would otherwise report
	"Too many links" rather than ignoring that type of failure.
	E.g., "mkdir d; cd d; ln -s . a; grep -r ^" would print
	grep: a: Too many links and would exit with status 2.
	Now, it prints nothing and exits with status 1, as before.
	Reported by Nelson H. F. Beebe.

	tests: avoid spurious failure of the symlink test
	* tests/symlink: Ignore spurious "Binary file d matches" on
	systems for which reading from a directory actually succeeds.
	Reported by Bruno Haible and Nelson Beebe.

2012-04-09  Jim Meyering  <>

	tests: avoid syntax-check failure: reverse compare arguments
	* tests/repetition-overflow: Fix reversed compare arguments.

	build: update gnulib submodule to latest

2012-03-18  Paul Eggert  <>

	grep: report overflow for ERE a{1000000000}
	* NEWS: Document this.
	* src/dfa.c (MIN): New macro.
	(lex): Lexically analyze the repeat-count operator once, not
	twice; the double-scan complicated the code and made it harder to
	understand and fix.  Adjust the repeat-count parsing so that it
	better matches the behavior of the regex code, in three ways:
	1. Diagnose too-large repeat counts rather than treating them as
	literal characters.  2. Use RE_INVALID_INTERVAL_ORD, not
	RE_NO_BK_BRACES, to decide whether to treat invalid-syntax {...}s
	as literals.  3. Use the same wording for {...}-related
	diagnostics that the regex code uses.
	* tests/bre.tests, tests/ere.tests, tests/repetition-overflow:
	Adjust to match new behavior, and add a few tests.
	* (exclude_file_name_regexp--sc_error_message_uppercase):
	New macro, since the diagnostics start with uppercase letters.

2012-03-14  Paul Eggert  <>

	grep: -r no longer follows symlinks; use fts
	Change -r to follow only command-line symlinks, and by default to
	read only devices named on the command line.  This is a simple
	way to get a more-useful behavior when searching random
	directories; the idea is to use 'find' if you want something fancy.
	-R acts as before and gets a new alias --dereference-recursive.
	The code now uses fts internally, so it is more robust and
	faster with large hierarchies.
	* .gitignore: Remove lib/savedir.c, lib/savedir.h.
	* tests/symlink: New file
	* Makefile.boot (LIB_OBJS_core): Remove isdir.o, savedir.o.
	Perhaps other changes are needed too, but I'm not sure what
	this makefile is for.
	* NEWS: Document changes.
	* doc/grep.texi (File and Directory Selection): Likewise.
	* bootstrap.conf (gnulib_modules): Remove dirent, dirname, isdir, open.
	Add fstatat, fts, openat-safer.
	* lib/ (libgreputils_a_SOURCES): Remove savedir.c, savedir.h.
	* lib/savedir.c, lib/savedir.h: Remove.
	* po/ Add lib/openat-die.c.
	* src/main.c: Include fcntl-safer.h, fts_.h.  Don't include
	isdir.h, savedir.h.
	(struct stats, stats_base): Remove.
	(long_options, usage, main): Add --dereference-recursive and
	implement -r vs -R.
	(filename_prefix_len, fts_options): New static vars.
	(basic_fts_options, READ_COMMAND_LINE_DEVICES): New constants.
	(devices): Now defaults to READ_COMMAND_LINE_DEVICES.
	(reset, grep): Now takes just struct stat rather than file name and
	struct stats.  All callers changed.
	(fillbuf): Now takes struct stat reather than struct stats.
	All callers changed.
	(grep): Don't worry about recursing too deeply; fts and grepdesc
	handle this now.
	(is_device_mode, grepdirent, grepdesc, grep_command_line_args):
	New functions.
	(grepfile): New args DIRDESC, FOLLOW, COMMAND_LINE.  Remove struct stats
	arg.  All callers changed.  Use openat_safer rather than open.
	Use desc == STDIN_FILENO to tell whether we're reading "-".
	Don't worry about EINTR when closing -- not possible, since we're
	not catching signals.
	* tests/ (TESTS): Add symlink.
	* tests/symlink: New file.

2012-03-12  Paul Eggert  <>

	tests: port big-match to non-GNU dd
	* tests/big-match: Don't assume GNU dd extension "bs=1M".

	tests: test for bug with -r --exclude-dir and no file operand
	* tests/include-exclude: Test for the bug and fix.

2012-03-12  Allan McRae  <>

	grep: fix segfault with -r --exclude-dir and no file operand
	* src/main.c (grepdir): Don't invoke excluded_file_name on NULL.
	* NEWS (Bug fixes): Mention it.

2012-03-09  Jim Meyering  <>

	tests: exercise two recently-fixed bugs
	* tests/repetition-overflow: New test for bugs fixed by commit
	* tests/ (TESTS): Add it.

2012-03-03  Jim Meyering  <>

	maint: use an optimal-for-grep xz compression setting
	* (XZ_OPT): Use -6e (determined empirically, see comments).
	This sacrifices a meager 60 bytes of compressed tarball size for a
	55-MiB decrease in the memory required during decompression.  I.e.,
	using -9e would shave off only 60 bytes from the tar.xz file, yet
	would force every decompression process to use 55 MiB more memory.

	build: update gnulib submodule to latest

2012-03-02  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.11
	* NEWS: Record release date.

	tests: avoid failure when using Solaris 10's sed
	* tests/reversed-range-endpoints: Use a simpler sed expression to
	sanitize actual output, so it also works with Solaris 10's /bin/sed.

2012-03-01  Jim Meyering  <>

	maint: manually correct formatting in dfa.c's cpp definitions
	* src/dfa.c: Adjust formatting in cpp definitions.

	maint: indent dfa.c
	* src/dfa.c: Filter through indent like this:
	HOME=. indent -Tsize_t -l79 --leave-preprocessor-space \
	  --dont-format-comments --no-tabs < dfa.c > k && mv k dfa.c

	doc: correct grep.1's descriptions of \w and \W (they omitted "_")
	* doc/ Fix descriptions of \w and \W.
	They did not mention "_".
	* doc/grep.texi (The Backslash Character and Special Expressions):
	[\w, \W]: List the "_" before the char class, not after: [_[:alnum:]],
	for readability and to be consistent with the man page.

2012-03-01  Paul Eggert  <>

	maint: spelling fixes

	grep: fix integer-overflow issues in main program
	* NEWS: Document this.
	* bootstrap.conf (gnulib_modules): Add inttypes, xstrtoimax.
	Remove xstrtoumax.
	* src/main.c: Include <inttypes.h>, for INTMAX_MAX, PRIdMAX.
	(context_length_arg, prtext, grepbuf, grep, grepfile)
	(get_nondigit_option, main):
	Use intmax_t, not int, for line counts.
	(context_length_arg, main): Silently ceiling line counts
	to maximum value, since there's no practical difference between
	doing that and using infinite-precision arithmetic.
	(out_before, out_after, pending): Now intmax_t, not int.
	(max_count, outleft): Now intmax_t, not off_t.
	(prepend_args, prepend_default_options, main):
	Use size_t, not int, for sizes.
	(prepend_default_options): Check for int and size_t overflow.

	grep: avoid mishandling of long lines
	* src/pcresearch.c (Pexecute): Do not pass a line longer than
	INT_MAX to pcre_exec, since its API does not permit that.

	grep: remove no-longer-used setrlimit code
	This code has been unused and obsolescent ever since the regex
	code stopped using the stack for large regular expressions.
	* src/main.c [HAVE_SETRLIMIT]: Do not include <sys/time.h> or
	or <sys/resource.h>; no longer needed.
	(set_rlimits): Remove.  All callers changed.

	grep: fix some core dumps with long lines etc.
	These problems mostly occur because the code attempts to stuff
	sizes into int or into unsigned int; this doesn't work on most
	64-bit hosts and the errors can lead to core dumps.
	* NEWS: Document this.
	* src/dfa.c (token): Typedef to ptrdiff_t, since the enum's
	range could be as small as -128 .. 127 on practical hosts.
	(position.index): Now size_t, not unsigned int.
	(leaf_set.elems): Now size_t *, not unsigned int *.
	(dfa_state.hash, struct mb_char_classes.nchars, .nch_classes)
	(.nranges, .nequivs, .ncoll_elems, struct dfa.cindex, .calloc, .tindex)
	(.talloc, .depth, .nleaves, .nregexps, .nmultibyte_prop, .nmbcsets):
	(.mbcsets_alloc): Now size_t, not int.
	(dfa_state.first_end): Now token, not int.
	(state_num): New type.
	(struct mb_char_classes.cset): Now ptrdiff_t, not int.
	(struct dfa.utf8_anychar_classes): Now token[5], not int[5].
	(struct dfa.sindex, .salloc, .tralloc): Now state_num, not int.
	(struct dfa.trans, .realtrans, .fails): Now state_num **, not int **.
	(struct dfa.newlines): Now state_num *, not int *.
	(prtok): Don't assume 'token' is no wider than int.
	(lexleft, parens, depth): Now size_t, not int.
	(charclass_index, nsubtoks)
	(parse_bracket_exp, addtok, copytoks, closure, insert, merge, delete)
	(state_index, epsclosure, state_separate_contexts)
	(dfaanalyze, dfastate, build_state, realloc_trans_if_necessary)
	(transit_state_singlebyte, match_anychar, match_mb_charset)
	(check_matching_with_multibyte_ops, transit_state_consume_1char)
	(transit_state, dfaexec, free_mbdata, dfaoptimize, dfafree)
	(freelist, enlist, addlists, inboth, dfamust):
	Don't assume indexes fit in 'int'.
	(lex): Avoid overflow in string-to-{hi,lo} conversions.
	(dfaanalyze): Redo indexing so that it works with size_t values,
	which cannot go negative.
	* src/dfa.h (dfaexec): Count argument is now size_t *, not int *.
	(dfastate): State numbers are now ptrdiff_t, not int.
	* src/dfasearch.c: Include "intprops.h", for TYPE_MAXIMUM.
	(kwset_exact_matches): Now size_t, not int.
	(EGexecute): Don't assume indexes fit in 'int'.
	Check for overflow before converting a ptrdiff_t to a regoff_t,
	as regoff_t is narrower than ptrdiff_t in 64-bit glibc (contra POSIX).
	Check for memory exhaustion in re_search rather than treating
	it merely as failure to match; use xalloc_die () to report any error.
	* src/kwset.c (struct trie.accepting): Now size_t, not unsigned int.
	(struct kwset.words): Now ptrdiff_t, not int.
	* src/kwset.h (struct kwsmatch.index): Now size_t, not int.

	tests: test for problems with long matches
	The new test is expensive, so add a category of expensive tests,
	which are normally not run, and put the new test in this new
	category.  The idea of having expensive tests is taken from coreutils.
	* HACKING: Mention RUN_EXPENSIVE_TESTS and similar env vars.
	* (check-expensive): New rule.
	* tests/ (TESTS): Add big-match.
	* tests/init.cfg (expensive_): New function, from coreutils.
	* tests/big-match: New file.

2012-02-29  Paul Eggert  <>

	maint: use gnulib _Noreturn rather than __attribute__ ((noreturn))
	* src/grep.h (__attribute__): Remove.
	* src/dfa.h (__attribute__): Likewise.
	(dfaerror): Use noreturn rather than __attribute__ ((noreturn)).
	* src/main.c (usage): Likewise.

2012-02-26  Jim Meyering  <>

	build: update submodule, bootstrap, tests/ from gnulib
	* gl/lib/regcomp.c.diff: Adjust.
	* bootstrap: Update from gnulib.
	* tests/ Update from gnulib.

2012-02-26  Paolo Bonzini  <>

	dfa: merge calls to SUCCEEDS_IN_CONTEXT
	* src/dfa.c (state_index): use a single call to SUCCEEDS_IN_CONTEXT.

	dfa: fix a subtle constraint encoding bug
	PREV_LETTER_DEPENDENT): Rewrite to handle all 3*3=9 possible
	combinations of previous and next character contexts.
	NOTLIMWORD_CONSTRAINT): Switch to new encoding.
	* NEWS: Document resulting bugfix.
	* tests/spencer1.tests: Add regression test.

	dfa: do not use MATCHES_*_CONTEXT directly
	* src/dfa.c (dfastate): Use SUCCEEDS_IN_CONTEXT.

	dfa: change meaning of a state context
	(state_separate_contexts): Remove second argument.
	(state_index): Do not mask away CTX_NONE.
	(dfaanalyze): Adjust call to state_index and state_separate_contexts.
	(dfastate): Adjust calls to state_index and state_separate_contexts.

2012-02-13  Paul Eggert  <>

	tests: fix loop in epipe test
	* tests/epipe: Don't loop forever if the bug is present.
	Problem reported by Jaroslav Skarvada.

2012-02-08  Paul Eggert  <>

	tests: work portably even if SIGPIPE is ignored
	* tests/epipe: Don't rely on "trap - PIPE"; that's not portable.
	Problem reported by Eric Blake in
	Also, use "ls -al" rather than "echo", in case "echo" is done by a
	buggy shell that ignores write errors.  And close grep's fd 3, as
	a sanity check.

2012-02-07  Paul Eggert  <>

	tests: work even if SIGPIPE is ignored
	* tests/epipe: Do not infinite-loop if SIGPIPE is already ignored.
	It could be that the invoker of 'make check' ignores SIGPIPE,
	for example.

2012-02-05  Jim Meyering  <>

	build: accommodate -Wshadow and -Werror=suggest-attribute=pure
	* src/dfa.c (state_separate_contexts): Add _GL_ATTRIBUTE_PURE.
	(dfaexec): Rename parameter, s/newline/allow_nl/, to avoid
	shadowing the global.

2012-02-05  Paolo Bonzini  <>

	dfa: refactor common context computations
	* src/dfa.c (CTX_ANY, charclass_context, state_separate_contexts): New.
	(dfaanalyze): Use state_separate_contexts.
	(dfastate): Use charclass_context and state_separate_contexts.  Rename
	prev_context to separate_contexts.

	dfa: change newline/letter to a single context value
	SUCCEEDS_IN_CONTEXT, ACCEPTS_IN_CONTEXT): Take a single context value
	for prev and curr.
	(struct dfa_state): Replace newline and letter with context.
	(wchar_context): New.
	(state_index): Replace newline and letter with context.  Compare
	context values in the state struct.  Adjust calls to pass contexts.
	(wants_newline): Replace with wanted_context.  Adjust calls to pass
	(dfastate): Replace wants_newline and wants_letter with wanted_context.
	Adjust calls to pass contexts.
	(build_state): Adjust calls to pass contexts.
	(match_anychar, match_mb_charset, transit_state): Use wchar_context.
	Adjust calls to pass contexts.

2012-02-05  Paolo Bonzini  <>

	dfa: introduce contexts for the values in d->success
	Also initialize all tables in a single place in dfasyntax.

	* src/dfa.c (CTX_NONE, CTX_LETTER, CTX_NEWLINE, char_context): New.
	(sbit, letters, newline): New.
	(dfasyntax): Fill them.
	(dfastate): Remove letters, newline, initialized.
	(build_state): Use CTX_* constants.
	(dfaexec): Remove sbit and sbit_init.

2012-02-05  Paolo Bonzini  <>

	dfa: remove useless check
	* src/dfa.c (state_index): There is nothing that is a newline *and*
	a letter.  Remove redundant call to SUCCEEDS_IN_CONTEXT.

2012-01-22  Jim Meyering  <>

	build: update bootstrap from gnulib and adapt
	* bootstrap: Update from gnulib.
	* tests/ Update from gnulib.
	* bootstrap.conf (bootstrap_epilogue): Remove now-unnecessary,
	snippet that edited gnulib-tests/
	(gnulib_tool_option_extras): Add both --symlink and  Remove use of $bt.
	* lib/ Initialize numerous automake variables so that
	generated code in may use += to append to them.

	maint: convert `this' to 'this' quoting style in diagnostics
	Now that gnulib's quote and quotearg modules use 'this' style,
	change the few explicit uses in diagnostics to conform.
	* src/egrep.c (after_options): Use 'this' style of quotes.
	* src/fgrep.c (after_options): Likewise.
	* src/grep.c (after_options): Likewise.
	* src/main.c (usage): Likewise.

	build: update gnulib to latest; adjust quoting in tests
	* gnulib: Update.
	* tests/in-eq-out-infloop: Convert expected diagnostics to match
	new quoting.

2012-01-22  Paul Eggert  <>

	doc: document recent diagnostics-related changes
	* NEWS: Document changes re diagnostics related to GREP_COLORS,
	directory loops, -s, "write error".

	grep: be quiet about GREP_COLORS syntax
	* src/main.c (struct color_cap): fct now returns void,
	since there's no longer need to use what it returns.
	(color_cap_mt_fct, color_cap_rv_fct, color_cap_ne_fct): Return void.
	(parse_grep_colors): Do not output diagnostics and then exit with
	status 0.  Instead, ignore errors in GREP_COLORS.  This is more
	consistent with programs that (e.g.) ignore errors in termcap entries,
	and it's more internally-consistent as some GREP_COLORS errors
	were ignored but not others.

	grep: exit with nonzero status if directory loop
	* src/main.c (grepdir): Exit with status 2 if a directory loop is
	found, since the output might not be "right" (i.e., infinite...).

	grep: suppress read errors if -s
	* src/main.c (reset, grep, grepfile): Do not report an input error
	if -s is given.

	grep: don't say "write error" over and over
	Problem reported by Travis Gummels in
	* src/main.c (write_error_seen): New static var.
	(clean_up_stdout): New function.
	(prline): Do not output 'write error' more than once; exit
	after the first one.  Use the same wording for the diagnostic
	that close_stdout uses.
	(main): Clean up with clean_up_stdout, not close_stdout, so that
	grep doesn't output multiple "write error" diagnostics.
	* tests/ (TESTS): Add epipe.
	* tests/epipe: New file.

2012-01-12  Paul Eggert  <>

	dfa: non-glibc word-constituent unibyte fix
	* src/dfa.c (is_valid_unibyte_character): Fix typo that caused
	this to incorrectly return 0 on unibyte non-glibc systems.
	Problem reported by Aharon Robbins in

2012-01-04  Paul Eggert  <>

	doc: document empty pattern better
	* doc/grep.texi (Top, Fundamental Structure, Usage):
	Explain how grep deals with the empty pattern.
	Problem spotted by Bernhard Voelker in

	grep: with no args, search "." only if command-line -r
	* NEWS: Document this.
	* doc/grep.texi (Environment Variables, grep Programs): Likewise.
	* src/main.c (usage): Likewise.
	(main): Implement this.
	(prepend_default_options): Return a count of prepended options.
	* tests/r-dot: Test the above.

2012-01-03  Jim Meyering  <>

	tests: adjust test to match code, now that --mmap writes to stderr
	* tests/ignore-mmap: Separate stdout and stderr; test both.

	deprecate the --mmap option
	* src/main.c (main): Deprecate the --mmap option: issue a warning
	when it is used.
	(usage): Change description.
	* doc/grep.texi (Other Options): Document the new behavior.
	* NEWS (Changes in behavior): Mention it.

2012-01-03  Paolo Bonzini  <>

	dfa: fix incorrect comment
	* src/dfa.c (dfastate): Fix comment for newline.

	dfa: fix rebase conflict
	* src/dfa.c (dfaanalyze): Fix reference to nalloc.

	dfa: automatically resize position_sets
	* src/dfa.c (insert, copy, merge): Resize arrays here.
	(dfaanalyze): Do not track number of allocated elements here.
	(dfastate): Allocate mbps with only one element.

	dfa: change position_set nelem to size_t
	* src/dfa.c (REALLOC_IF_NECESSARY): Disable assertion, to avoid
	warnings from -Wtype-limits.
	(position_set): Change nelem to a size_t.

	dfa: move nalloc to position_set structure
	* src/dfa.c (position_set): Add alloc.
	(alloc_position_set): Initialize it.
	(dfaanalyze): Use it instead of the nalloc array or nelem.

	dfa: remove dead assignment
	* src/dfa.c (transit_state): transit_state_consume_1char will clear follows,
	do not do this ourselves.

	dfa: introduce alloc_position_set
	* src/dfa.c (alloc_position_set): New function, use it throughout.

	dfa: use a more compact data type for grps
	* src/dfa.c (leaf_set): New.
	(dfastate): Use the smaller type, leaf_set, for grps.  Its prior type
	contained an unused constraint field.

	dfa: use MALLOC/REALLOC always
	src/dfa.c (dfastate, enlist, dfamust): Use MALLOC and REALLOC.

	dfa: remove unnecessary braces
	* src/dfa.c (dfastate): Remove unnecessary braces.

	dfa: x2nrealloc starting from a NULL pointer works
	* src/dfa.c (parse_bracket_exp): Do not MALLOC mbcset parts the first time
	they are encountered.  Initialize chars_al correctly.

2012-01-03  Jim Meyering  <>

	build: avoid build failure with --enable-gcc-warnings and recent gcc
	* lib/colorize-posix.c: Disable -Wsuggest-attribute=const, to avoid
	warning about this empty init_colorize function.

2012-01-03  Paolo Bonzini  <>

	remove lib/ms/
	* Create lib/colorize.c as a symbolic link.
	* lib/colorize-posix.c: New name of lib/colorize-impl.c.
	* lib/colorize-w32.c: New name of lib/ms/colorize-impl.c.
	* lib/colorize.c: Delete.
	* lib/ (EXTRA_DIST): Adjust.
	* .gitignore: Adjust.
	* Adjust syntax-check exclusions.

	unify colorize.h headers
	* lib/ (EXTRA_DIST): Adjust.
	* lib/colorize.h: Remove inline functions.
	* lib/colorize-impl.c: Move them here as functions.
	* lib/ms/colorize.h: Remove.
	* src/ (DEFAULT_HEADERS): Remove.

2012-01-02  Paolo Bonzini  <>

	colorize: use isatty module
	* bootstrap.conf: Add isatty module.
	* gnulib: Update to latest.
	* lib/colorize.h: Remove argument from should_colorize.
	* lib/ms/colorize.h: Likewise.
	* lib/colorize-impl.c: Factor isatty call out of here...
	* lib/ms/colorize-impl.c: ... and here...
	* src/main.c: ... into here.

2012-01-02  Jim Meyering  <>

	tests: avoid minor "make check" failure
	* tests/r-dot: Make executable, to avoid triggering a failed
	consistency test in "make check".

2012-01-02  Paul Eggert  <>

	grep: -r with no args now searches "."
	This is a patch I've been meaning to put in for years.
	When I added support for "grep -r", I forgot to have "grep -r PAT"
	search the working directory by default, instead of searching
	standard input (which makes no sense, even if stdin is a directory).
	This is not an upward compatible change, since "grep -r PAT <file"
	will no longer search standard input, but that's OK; nobody should
	be using "grep -r" that way anyway.
	* NEWS: Document this.
	* doc/grep.texi (File and Directory Selection, grep Programs, Usage):
	* src/main.c (usage): Likewise.
	(grepdir): If DIR is null, search the working directory, but do
	not prepend "./" to the file names.
	(main): If recursing and no operands are given, search ".".
	* tests/ (TESTS): Add r-dot.
	* tests/r-dot: New file.

	grep: prefer fgets to printf, _ to gettext
	* lib/colorize.h (print_end_colorize):
	* lib/ms/colorize-impl.c (print_end_colorize):
	Use fputs instead of printf.
	* src/main.c (usage): Likewise.  Use _ instead of gettext.

2012-01-01  Paul Eggert  <>

	grep: check stdin like other files
	* NEWS: Document this.
	* src/main.c (grepfile): Revamp tests for input files so that
	standard input is tested like other files.  For example, report
	an error if standard input equals standard output.
	Prefer open+fstat to stat+open if possible, as open+fstat is
	usually a bit faster and avoids a race condition.
	* tests/in-eq-out-infloop: Add tests for cases like
	'grep pat <file >>file'.

2012-01-01  Jim Meyering  <>

	maint: update all copyright year number ranges
	Run "make update-copyright".

2011-12-31  Paul Eggert  <>

	grep: lower-case function names
	These names used to be macros, but they're functions now.
	All callers changed.
	* src/main.c (pr_sgr_start): Rename from PR_SGR_START.
	(pr_sgr_end): Rename from PR_SGR_END.
	(pr_sgr_start_if): Rename from PR_SGR_START_IF.
	(pr_sgr_end_if): Rename from PR_SGR_END_IF.

	ms: move Microsoft-specific stuff to lib/ms
	* (exclude_file_name_regexp--sc_prohibit_strcmp)
	New rules.
	* lib/colorize.c, lib/colorize.h, lib/colorize-impl.c:
	* lib/ms/colorize.h, lib/ms/colorize-impl.c: New files.
	* (GREP_SRC_INCLUDES): New macro.
	* lib/ (libgreputils_a_SOURCES): Add colorize.[ch].
	(EXTRA_DIST): New macro.
	* src/ (DEFAULT_INCLUDES): New macro.
	* src/main.c: Include colorize.h.
	Now static functions, not macros.
	(hstdout, norm_attr, w32_console_init, w32_sgr2attr)
	(w32_clreol) [__MINGW32__]: Move to lib/ms/colorize-impl.c.
	(pr_sgr_start, pr_sgr_end): Remove; callers changed to use new
	print_start_colorize, print_end_colorize from colorize.h.
	(init_colorize): Rename from w32_console_init and move to
	colorize module; caller changed.
	(should_colorize): Move to colorize module.

	grep: do input==output check more like dir loop check
	* src/main.c (grepfile): Just use SAME_INODE; don't bother
	with SAME_REGULAR_FILE.  This works better on properly-working
	POSIX hosts, since it handles the case where the file is changing
	as we grep it.  It works worse on hosts that don't support st_ino
	properly, but in practice this isn't that much of a problem here.
	* src/system.h (same_file_attributes, SAME_REGULAR_FILE):
	Remove; no longer needed.

	build: update gnulib submodule to latest

2011-12-28  Paul Eggert  <>

	maint: remove now-unused/obsolete files
	* README.DOS: Remove file.
	* m4/djgpp.m4: Likewise.
	* .gitignore: Remove reference to m4/djgpp.m4.

2011-12-28  Jim Meyering  <>

	maint: distribute ChangeLog-2009
	* (EXTRA_DIST): Add ChangeLog-2009.
	Spotted by Eli Zaretskii.

2011-12-28  Jim Meyering  <>

	main.c: add some 'const' directives
	* src/main.c (color_dict, fg_color, bg_color, cap): Declare const.

	No semantic change.

2011-12-28  Jim Meyering  <>

	main.c: correct indentation and formatting style
	* src/main.c: Correct many formatting inconsistencies.
	No semantic change.

	avoid new syntax-check failures
	* (old_NEWS_hash): Update, to accommodate old NEWS modification.
	* src/main.c: Indent solely with spaces, never with TABs.
	(should_colorize): Remove useless parens in #if directive.

2011-12-28  Eli Zaretskii  <>

	Fix whitespace, indentation and documentation
	* src/main.c (parse_grep_colors): Fix indentation.
	(usage): Mention MS-Windows in help text for -U and -u options.

	update NEWS for MS-Windows changes
	* NEWS: Mention MS-Windows related bugfixes and enhancements.

	Fix the test suite for MS-Windows.
	* tests/include-exclude: Use --directories=skip, to avoid
	gratuitous failures on systems that cannot grep directories.
	* tests/reversed-range-endpoints: Don't reject program names with
	leading directories and drive letters.
	* tests/warn-char-classes: Likewise.

	Support color highlighting on MS-Windows
	* src/main.c (SGR_START, SGR_END, PR_SGR_FMT, PR_SGR_FMT_IF): Remove.
	(PR_SGR_START, PR_SGR_START_IF): Replace with pr_sgr_start.
	(PR_SGR_END, PR_SGR_END_IF): Replace with pr_sgr_end.
	(pr_sgr_start, pr_sgr_end, should_colorize): New functions.
	(w32_console_init, w32_sgr2attr, w32_clreol) [__MINGW32__]: New functions.
	(main): Use should_colorize.  Invoke w32_console_init.

2011-12-24  Paul Eggert  <>

	don't ignore errors when reading a directory
	grep no longer silently suppresses errors when reading a directory
	as if it were a text file.  For example, "grep x ." now reports a
	read error on most systems; formerly, it ignored the error.
	Problem reported as an aside by Bob Proulx (Bug#10355).
	* NEWS: Document this.
	* src/main.c (grep, grepfile): Implement this.  Simplify the code
	* src/system.h (is_EISDIR): Remove; no longer needed.

	--include etc. now work on command-line args more consistently
	--include and --exclude apply only to non-directories and
	--exclude-dir applies only to directories.  "-" (standard input)
	is never excluded, since it is not a file name.
	This bug was discovered while fixing a read-directory bug (Bug#10355).
	* NEWS: Document this.
	* src/main.c (main): Implement this.
	* tests/include-exclude: Test for it.

2011-12-24  Jim Meyering  <>

	build: update gnulib submodule to latest

2011-12-12  Arnold D. Robbins  <>

	doc: improve grep.texi
	* doc/grep.texi: General editing for improved aesthetics.
	Also fix a few problems.

2011-12-12  Jim Meyering  <>

	build: use gnulib's iswctype wcscoll
	* bootstrap.conf (gnulib_modules): Add iswctype and wcscoll.
	* Remove explicit checks for those functions.
	* src/mbsupport.h (MBS_SUPPORT): Define to 1 if not already defined.
	Remove the conditional, now that we're guaranteed by gnulib to have
	wcscoll and iswctype.
	Suggested by Alan Hourihane in

	disable the new input==output guard for additional options
	* src/main.c (grepfile): Do not reject input == output also
	when using a few other options.
	* tests/in-eq-out-infloop: Test these new cases.
	* NEWS (Bug fixes): Mention it

2011-12-11  Nicolas Vigier  <>

	do not reject "grep -qr . > out"
	The recent fix to avoid an infinite disk-filling loop, commit 5e20a38a,
	introduced a minor regression.  If you use grep with -q and -r, and
	redirect output to a file that will be traversed, then grep would
	reject the command, even though it will generate no output.
	In that case, there is no risk of an infinite loop.
	* src/main.c (grepfile): Do not reject input == output when
	using --quiet/--silent (-q).
	Reported by J H Wilson in
	forwarded by Nicolas Vigier to

2011-11-29  Arnold Robbins  <>

	dfa: do not call nl_langinfo in !MBS_SUPPORT mode
	* src/dfa.c (using_utf8) [!MBS_SUPPORT]: Remove erroneous "defined"
	in cpp test for MBS_SUPPORT.  Since commit a163349d, MBS_SUPPORT is 0/1.
	This error caused trouble only in the !MBS_SUPPORT case.

	dfa: avoid warning from deficient compiler in !MBS_SUPPORT mode
	* src/dfa.c (setbit_wc) [!MBS_SUPPORT]: Add explicit "return false;"
	after "abort ();", to avoid a warning from deficient compilers.

2011-11-29  Jim Meyering  <>

	tests: use "compare exp out", not "compare out exp"
	Likewise, when an empty file is expected, use "compare /dev/null out",
	not "compare out /dev/null". I.e., specify the expected/desired contents
	via the first file name.  Prompted by a suggestion from Bruno Haible

	Run these commands:

	    git grep -l -E 'compare [^ ]+ exp' \
	      |xargs perl -pi -e 's/(compare) (\S+) (exp\S*)/$1 $3 $2/'
	    git grep -l -E 'compare [^ ]+ /dev/null' \
	      |xargs perl -pi -e 's/(compare) (\S+) (\/dev\/null)/$1 $3 $2/'

2011-11-29  Jim Meyering  <>

	build: update gnulib submodule to latest

2011-11-28  Jim Meyering  <>

	build: accommodate -Werror=suggest-attribute=pure
	Now that we're using the latest manywarnings module from gnulib,
	accommodate gcc's -Werror=suggest-attribute=pure option by marking
	suggested functions with gnulib-defined _GL_ATTRIBUTE_PURE.
	* src/kwset.c (hasevery): Mark function with pure attribute.
	(bmexec): Likewise.
	* src/dfa.c (nsubtoks, istrstr, find_pred, dfamusts): Likewise.
	* Disable (for lib/) options that seem not to be worth
	the trouble: -Wunsuffixed-float-constants and -Wformat-nonliteral.

2011-11-21  Bruno Haible  <>

	build: fix "make check" error on OSF/1
	* tests/ (TESTS_ENVIRONMENT): Test the value of the variable
	BASH_VERSION, not the literal ASH_VERSION.

2011-11-21  Jim Meyering  <>

	portability: work consistently on *BSD systems
	* src/dfa.c (is_valid_unibyte_character): Define.
	(IS_WORD_CONSTITUENT): Use it here, to make grep work consistently
	even on *BSD systems, which use different tables for ctype macros
	like isalpha.
	With help from Bruno Haible.

2011-11-20  Jim Meyering  <>

	maint: consistently use NULL, not 0, when comparing pointers
	* src/dfa.c (dfaanalyze): Compare trans[s] with NULL, not 0.

	maint: remove an avoidable #ifdef/#endif pair
	* src/dfa.c (dfaanalyze): Remove avoidable #ifdef around "{".

	tests: fix typo in last change
	* tests/word-delim-multibyte: Use double quotes around $e_acute,
	not single quotes.  Spotted by Bruno Haible.
	This and the preceding change do not resolve the XPASS failure
	on OpenBSD 4.9 after all.  See the explanation at

	tests: avoid unwarranted test failure on *BSD-based systems
	* tests/word-delim-multibyte (e_acute): Use a more portable
	representation of e-acute.  Reported by Bruno Haible.

2011-11-19  Jim Meyering  <>

	maint: accommodate -Wdeclaration-after-statement, but only in dfa.c,
	and because doing so does not impact readability/maintainability.
	This is solely to accommodate gawk users who are stuck with ancient gcc.
	This is no excuse to change any other code in grep.
	* src/dfa.c (dfaoptimize, parse_bracket_exp): Move declaration
	to precede first statement in block.

2011-11-16  Jim Meyering  <>

	maint: post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.10
	* NEWS: Record release date.

	build: update gnulib submodule to latest

2011-11-13  Jim Meyering  <>

	maint: update bootstrap and from gnulib
	* tests/ Update from gnulib.
	* bootstrap: Likewise.

2011-11-12  Jim Meyering  <>

	build: update gnulib for exclude-test fixes

	tests: make our "export" replacement efficient with modern shells
	* tests/ (TESTS_ENVIRONMENT): Use a trivial and efficient
	implementation with a shell that supports "export var=val".
	Use the sed-invoking replacement only when necessary.
	Improved by Stefano Lattarini.

	tests: make the replacement export function more robust
	* tests/ (sed_quote_value): Also quote single quotes.
	Remove sed's -e options.  Not needed.

2011-11-12  Bruno Haible  <>

	tests: fix test suite execution failure on OSF/1 5.1
	* tests/ (TESTS_ENVIRONMENT): Use a shell function to
	ensure that we use only the portable form of the 'export' shell

	tests: don't assume that /bin/bash exists
	* tests/fedora: Run using /bin/sh, not /bin/bash.

	tests: avoid unwarranted failures due to SATAN's timeout
	* tests/init.cfg (require_timeout_): Also ensure that
	timeout exits with its child's exit status.

	build: fix compilation error on MSVC 9 to due Pexecute() declaration
	* src/pcresearch.c (WITHOUT_PCRE_NORETURN): Remove macro.
	(Pexecute): Replace abort() call with code that does not trigger GCC

	tests: fix high-bit-range test failure on OSF/1 5.1
	* tests/high-bit-range: Use octal escape instead of hexadecimal escape

2011-11-11  Jim Meyering  <>

	build: update gnulib for solaris test fix

2011-11-10  Jim Meyering  <>

	build: update gnulib submodule to latest

	maint: adjust the URL that will appear in the generated announcement
	* (url_dir_list): Use this$(PACKAGE)
	for the first link listed in the generated announcement.
	announce-gen now provides the faster mirror link automatically.

2011-11-06  Jim Meyering  <>

	build: stop distributing gzip'd releases;  xz is enough
	* (AM_INIT_AUTOMAKE): Add no-dist-gzip.
	* NEWS (Build-related): Mention that we're dropping .tar.gz.

	build: update gnulib submodule to latest

2011-10-14  Stefano Lattarini  <>

	distcheck: ensure dist-hook fails if syntax-check fails
	* (run-syntax-check): Fix logic, to ensure that
	the recipe of this target returns a non-zero exit status if
	"make syntax-check" fails.

2011-10-12  Jim Meyering  <>

	build: update gnulib submodule to latest
	This should fix a few portability problems, including one on HP-UX
	and a test-float failure on PPC, reported by Andreas Metzler.

2011-10-10  Stefano Lattarini  <>

	gitignore: merge top-level and tests/ .gitignore files
	* tests/.gitignore: Remove; what little remained of its
	contents has been moved ...
	* .gitignore: ... here.

	tests: tiny simplification in TESTS_ENVIRONMENT definition
	* tests/ (TESTS_ENVIRONMENT): Remove redundant use of

2011-10-10  Stefano Lattarini  <>

	tests: support development version of automake too
	This change implements a more correct and idiomatic use of the
	features of the Automake-provided 'parallel-tests' harness.
	Moreover, this change is required in order for the testsuite to
	continue to work with the new testsuite harness that is planned
	to be introduced in Automake 1.12 (which, as of the writing date,
	is still under development and in late alpha state).

	* tests/ (TESTS_ENVIRONMENT): The development version of
	automake dos not support setting the interpreter delegated to run
	the tests scripts in this variable; instead, use ...
	(LOG_COMPILER): ... this variable.
	* .gitignore: Ignore `.trs' files in directory `tests/'.
	* build-aux/.gitignore: Ignore `test-driver' script.

2011-10-03  Eli Zaretskii  <>

	dfa: don't mishandle high-bit bytes in a regexp with signed-char
	This appears to arise only on systems for which "char" is signed.
	* src/dfa.c (FETCH_WC, FETCH): Produce an unsigned value, rather
	than a sign-extended one.  Fixes a bug on MS-Windows with compiling
	patterns that include characters with the 8-th bit set.
	(to_uchar): Define.  From coreutils.
	Reported by David Millis <>.
	* NEWS (Bug fixes): Mention it.

2011-09-16  Jim Meyering  <>

	maint: dfa: simplify multi-byte-related conditionals
	* src/dfa.c (setbit_case_fold_c, parse_bracket_exp, lex):
	(addtok_mb, dfaparse): Change each "MBS_SUPPORT && MB_CUR_MAX > 1"
	test to just "MB_CUR_MAX > 1".
	* src/dfasearch.c (kwsincr_case, EGexecute): Likewise.
	* src/kwsearch.c (Fcompile, Fexecute): Likewise.
	* src/searchutils.c (kwsinit): Likewise.
	* src/dfa.c (parse_bracket_exp): Convert
	"if (!MBS_SUPPORT || MB_CUR_MAX == 1)" to
	"if (MB_CUR_MAX == 1)" and do this:
	-      assert(!MBS_SUPPORT || MB_CUR_MAX == 1);
	+      assert(MB_CUR_MAX == 1);

	maint: dfa: simplify several expressions
	* src/dfa.c (dfainit): Set d->mb_cur_max unconditionally, now
	that MB_CUR_MAX is always usable.  With that, simplify all
	"MBS_SUPPORT && d->mb_cur_max > 1" to simply "d->mb_cur_max > 1".
	(dfastate, dfaexec, dfainit, dfafree): Simplify, removing each
	now-unnecessary "MBS_SUPPORT &&".

	maint: dfa: avoid in-function "#if MBS_SUPPORT" tests
	* src/dfa.c (setbit_case_fold_c): Remove "#if MBS_SUPPORT" in favor
	of simple "if (MBS_SUPPORT ...".
	(dfaexec, addtok): Likewise.

	maint: ensure that MB_CUR_MAX is defined even when !MBS_SUPPORT
	* src/mbsupport.h [!MBS_SUPPORT] (MB_CUR_MAX): Define to 1.

	build: fix compilation failure when MBS_SUPPORT is 0
	* src/dfa.c (add_utf8_anychar): Always compile this function,
	but when MBS_SUPPORT is 0, give it an empty body.
	(prepare_wc_buf): Likewise.
	[! MBS_SUPPORT] (setbit_wc): Define to always abort.

	maint: dfa: simplify dfaoptimize
	* src/dfa.c (dfaoptimize): Simplify.
	(dfacomp): Remove now-redundant "if (MBS_SUPPORT)" guard,
	since dfaoptimize does nothing if !MBS_SUPPORT.

	maint: dfa: remove some #if MBS_SUPPORT guards
	* src/dfa.c: Replace a few "#if MBS_SUPPORT" directives with
	"if (MBS_SUPPORT)".  Remove some altogether.

	maint: dfa: convert #if-MBS_SUPPORT (dfastate)
	* src/dfa.c (dfastate): Use regular "if", not #if MBS_SUPPORT.

	maint: dfa: convert #if-MBS_SUPPORT (dfastate)
	* src/dfa.c (dfastate): Use regular "if", not #if MBS_SUPPORT.

	maint: dfa: convert #if-MBS_SUPPORT (state_index)
	* src/dfa.c (state_index): Use regular "if", not #if MBS_SUPPORT.

	maint: dfa: convert #if-MBS_SUPPORT (dfaparse)
	* src/dfa.c (dfaparse): Use regular "if", not #if MBS_SUPPORT.'

	maint: dfa: convert #if-MBS_SUPPORT (copytoks)
	* src/dfa.c (copytoks): Use regular "if", not #if MBS_SUPPORT.'

	maint: dfa: convert #if-MBS_SUPPORT (lex)
	* src/dfa.c (lex): Use regular "if", not #if MBS_SUPPORT.'

	maint: dfa: convert #if-MBS_SUPPORT (parse_bracket_exp)
	* src/dfa.c (parse_bracket_exp): Use regular "if", not #if MBS_SUPPORT.

	maint: dfa: convert #if-MBS_SUPPORT (parse_bracket_exp)
	* src/dfa.c (parse_bracket_exp): Use regular "if", not #if MBS_SUPPORT.

	maint: dfa: convert #if-MBS_SUPPORT (parse_bracket_exp)
	* src/dfa.c (parse_bracket_exp): Use regular "if", not #if MBS_SUPPORT.

	maint: dfa: convert #if-MBS_SUPPORT (dfaexec)
	* src/dfa.c (dfaexec): Use regular "if", not #if MBS_SUPPORT.

	maint: dfa: convert #if-MBS_SUPPORT (dfaexec)
	* src/dfa.c (dfaexec): Use regular "if", not #if MBS_SUPPORT.
	Also add curly braces around multi-line if/else blocks.

	maint: dfa: remove #if-MBS_SUPPORT (free_mbdata)
	* src/dfa.c (free_mbdata): Remove the #if guard altogether.

	maint: dfa: convert #if-MBS_SUPPORT (dfaoptimize, dfacomp)
	* src/dfa.c (dfaoptimize, dfacomp): Use regular "if",
	not #if MBS_SUPPORT.

	maint: dfa: convert #if-MBS_SUPPORT (dfafree)
	* src/dfa.c (dfafree): Use regular "if", not #if MBS_SUPPORT.

	maint: dfa: convert #if-MBS_SUPPORT (parse_bracket_exp, part1)
	* src/dfa.c (parse_bracket_exp): Remove in-function #if MBS_SUPPORT.

	maint: remove #if-MBS_SUPPORT declaration guards
	* src/search.h: Don't bother to #if-out declarations.

	maint: convert #if-MBS_SUPPORT (EGexecute)
	* src/dfasearch.c (EGexecute): Remove in-function #if MBS_SUPPORT.

	maint: convert #if-MBS_SUPPORT (kwsincr_case)
	* src/dfasearch.c (kwsincr_case): Remove in-function #if MBS_SUPPORT.
	Move decl's down.

	maint: convert #if-MBS_SUPPORT (Fcompile, etc.)
	* src/kwsearch.c (Fcompile, Fexecute): Remove in-function #if MBS_SUPPORT.
	(Fcompile): Rearrange some declarations.  No semantic change.

	maint: convert #if-MBS_SUPPORT (kwsinit)
	* src/searchutils.c (kwsinit): Remove in-function #if MBS_SUPPORT.

	maint: dfa: remove case-guarding #if-MBS_SUPPORT
	* src/dfa.c [DEBUG] (prtok): Remove now-useless #if-MBS_SUPPORT.

2011-09-15  Jim Meyering  <>

	maint: remove #if MBS_SUPPORT around member declaration
	* src/dfa.c (dfastate): Don't #ifdef-out "mbps" position_set member.

	maint: dfa: remove #if MBS_SUPPORT around struct definition
	* src/dfa.c (struct mb_char_classes): Don't #ifdef-out declarations.

	build: avoid compilation failure when building without PCRE support
	* src/pcresearch.c [!HAVE_LIBPCRE] (WITHOUT_PCRE_NORETURN): Define
	to _Noreturn, not obsoleted-by-gnulib _GL_ATTRIBUTE_NORETURN.
	Reported by Eric Blake.

	tests: stop using skip_test_; use skip_ instead
	* tests/init.cfg (skip_test_): Remove definition.  Use the improved
	skip_ function from, now that it has the same feature.
	* tests/euc-mb: s/skip_test_/skip_/
	* tests/sjis-mb: Likewise.
	* tests/fmbtest: Likewise.

	tests: skip tests that require MBS support
	* tests/init.cfg (require_compiled_in_MB_support): New function.
	* tests/char-class-multibyte: Use it here, since this test cannot
	succeed without MBS support.
	* tests/equiv-classes: Likewise.
	* tests/euc-mb: Likewise.
	* tests/fgrep-infloop: Likewise.
	* tests/init.cfg: Likewise.
	* tests/prefix-of-multibyte: Likewise.
	* tests/turkish-I: Likewise.
	* tests/sjis-mb: Likewise.

	tests: make fmbtest explain (to stderr, not log) why it is skipped
	* tests/fmbtest: Use skip_ and fail_ to give better diagnostics.

	maint: dfa: improve comments
	* src/dfa.c (match_mb_charset, match_anychar): Improve comments.

2011-09-14  Jim Meyering  <>

	build: update gnulib submodule to newer

	maint: correct indentation
	* src/dfa.c (dfaexec): Reposition curly braces to match indentation style.
	Remove useless comment.

	maint: move declaration "down" to inner scope where it is used
	* src/dfa.c (dfaexec): Move decl of local down into scope where used.

2011-09-07  Jim Meyering  <>

	doc: use "file name" consistently in grep's --help output
	* src/main.c (usage): Use "file name", not "filename" in descriptions
	of --with-filename (-H), --no-filename (-h) and --label=LABEL.
	Suggested by Sequoia McDowell.

	bug: requires ru_RU.KOI8-R".  [bug introduced in grep-2.9]

2011-08-31  Matthew Burgess  <>

	tests: remove debug code that would cp to /t
	* tests/unibyte-bracket-expr: Remove debug artifact introduced
	by 2011-06-02 commit de5f7000, "tests: exercise a uni-byte [...]
	bug: requires ru_RU.KOI8-R".  [bug introduced in grep-2.9]

2011-08-20  Jim Meyering  <>

	build: use largefile module and update to latest gnulib
	* Remove AC_SYS_LARGEFILE, subsumed by ...
	* bootstrap.conf (gnulib_modules): ...this.  Use largefile module.
	* gnulib: Update to latest.

	maint: clean up and plug a leak-on-OOM
	* src/dfa.c (icatalloc): Clean up; use xrealloc in place of malloc
	and realloc; remove conditionals that are unnecessary, now that
	failed allocation results in exit.
	(enlist): Use xrealloc in place of realloc; remove conditional.
	(comsubs): Avoid leak upon failed enlist call.
	(dfamust): Use xmalloc in place of malloc.
	Remove conditionals, now that icpyalloc and icatalloc never return NULL.

	maint: use x2nrealloc, not xrealloc
	* src/main.c (main): Use x2nrealloc, not xrealloc

2011-07-24  Jim Meyering  <>

	tests: add a test to trigger the bug
	* tests/ (TESTS): Add it.
	* tests/in-eq-out-infloop: Exercise the bug/fix.

	exit 2 (rather than infloop) when an input file is also on stdout
	This avoids a potential "infinite" disk-filling loop.
	Reported in
	* src/main.c: Include "quote.h".
	(out_stat): New global.
	(grepfile): Compare each regular file's dev/ino/etc.
	with those from the file on stdout (if it too is regular).
	(main): Set out_stat, if stdout is a regular file.
	* src/system.h: Include "same-inode.h".
	(same_file_attributes): Define.  From diffutils.
	* bootstrap.conf (gnulib_modules): Use quote, not quotearg.
	Use same-inode.
	* NEWS (Bug fixes): Mention it.

2011-07-15  Reuben Thomas  <>

	doc: improve documentation of character classes in the man page
	* doc/ Reword documentation of character classes.

2011-07-12  Jim Meyering  <>

	dfa: remove unnecessary inclusion of verify.h
	* src/dfa.c: Don't include "verify.h".

	dfa: simplify use of *ALLOC macros
	* src/dfa.c (XNMALLOC, XCALLOC): Redefine without outer cast-to-(t *).
	(CALLOC, MALLOC, REALLOC): Remove type "t" parameter and adjust callers.

	dfa: change semantics of REALLOC_IF_NECESSARY's 3rd parameter
	* src/dfa.c (REALLOC_IF_NECESSARY): Change meaning of 3rd param,
	from "maximum index" to 1 greater than that: the required number
	of *P-sized elements.  Note that only some of the uses of
	REALLOC_IF_NECESSARY needed to be adjusted, the others had already
	required an extra element.

	dfa: rename REALLOC_IF_NECESSARY param/local for clarity
	* src/dfa.c (REALLOC_IF_NECESSARY): Rename nalloc and new_nalloc
	to n_alloc and new_n_alloc.

	dfa: prepare for a semantic change in REALLOC_IF_NECESSARY
	* src/dfa.c (REALLOC_IF_NECESSARY): Remove "t" (type) parameter.
	Use (*p) instead.  Adjust all callers.

	dfa: add braces to REALLOC_IF_NECESSARY definition
	* src/dfa.c (REALLOC_IF_NECESSARY): Add curly braces; use TABs
	to right-indent.

2011-06-28  Paolo Bonzini  <>

	doc: improve documentation of character classes
	* doc/grep.texi (Character classes): Mention explicitly when
	examples refer to the C locale, explain better the general
	meaning of character classes.

2011-06-28  Jim Meyering  <>

	dfa: fix the root cause of the heap overrun
	dfa's "insert" function was supposed to be maintaining the position
	list sorted on *decreasing* index, but since the 2009-12-09 "Speed
	up insert" commit, 62458291, it was using code that assumed the data
	were sorted on *increasing* index.  As such, sometimes it would no
	longer merge constraints (not finding a match) and would append
	entries that normally would have matched and been merged.  Those
	erroneous append operations resulted in the heap overrun fixed by
	2011-06-17 commit 0b91d692 by doubling the array size.
	* src/dfa.c (insert): Fix the comparison.
	(dfaanalyze): Now that that's fixed, revert commit 0b91d692,
	allocating space for only d->nleaves entries, not double that.
	As far as I can tell, this change has no effect other than
	decreased memory usage, although it may improve performance
	slightly, since the resulting list of positions is half as long
	as it used to be.

2011-06-28  Paolo Bonzini  <>

	dfa: use memcpy to copy position_sets
	* src/dfa.c (copy): Use memcpy.

	dfa: use copyset to copy charclasses
	* src/dfa.c (add_utf8_anychar): Change memcpy to copyset.

	gnulib: Update
	Fixes mmap-anon.m4 conflict with fn_grep, reported by Rainer Orth.

2011-06-21  Jim Meyering  <>

	maint: update bootstrap from gnulib
	* bootstrap: Update to latest, so it no longer inserts empty lines
	in .gitignore files.
	* .gitignore: Let bootstrap move "!..." lines to end of file.

	post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.9
	* NEWS: Record release date.

	build: avoid a warning when building with --disable-perl-regexp...
	and --enable-gcc-warnings.
	* src/pcresearch.c (WITHOUT_PCRE_NORETURN): Define.
	Remove the unreachable return statement.
	Reported by Eric Blake.

	tests: ensure that each test script is executable
	This adds a rule run at "make check" time to ensure that
	test scripts are consistently executable.
	This change is not required for "make check", but makes it easier
	for people to run scripts manually, but that is discouraged because
	doing so makes it easy to omit important variable settings that
	are normally provided via TESTS_ENVIRONMENT.
	This change also makes each of the existing TESTS executable.
	* tests/ (check_executable_TESTS): New rule.
	(check): Depend on it.
	* tests/{all_scripts}: chmod 755.
	Prompted by a report from Eric Blake.

	maint: update bootstrap from gnulib
	* bootstrap: Update from gnulib.

	maint: update po/
	* po/ Remove dfasearch.c, now that it no longer
	contains a translatable diagnostic.

	tests: include-exclude: avoid false positive failure on FreeBSD
	* tests/include-exclude: Avoid false-positive failure due to
	matching "a" in a directory on FreeBSD, when searching a directory
	without "-r".  Search for '^aaa$' rather than just 'a'.
	Adjust test inputs and expected output files accordingly.

	dfa: remove some useless casts
	* src/dfa.c (icatalloc): Change type of "old" parameter
	from "char const *" to "char *".
	Don't cast-away const on realloc argument.
	Remove now-unnecessary const-discarding cast.
	Don't (void)-cast strcpy result.
	* src/dosbuf.c (undossify_input): Remove anachronistic
	cast-to-"char *" of realloc argument.

	dfa: more heap-allocation-related overflow protection
	* src/dfa.c (enlist): Use xnrealloc, not realloc.
	Also, remove unnecessary cast-to-(char *).
	(dfamust): Use xnmalloc, not malloc.  Before, this code would
	return upon malloc failure (xnmalloc exits upon failure), but
	later, via the *ALLOC macros, it could already exit, so this
	new potential exit point is nothing new.  The same applies
	to enlist, since it is called only through dfamust.

	tests: update; simplify TESTS_ENVIRONMENT
	* tests/ Update from coreutils.
	* tests/ (TESTS_ENVIRONMENT): Remove shell_or_perl_
	function.  Instead, just use $(SHELL), since grep has no test
	that starts with #!/usr/bin/perl.

2011-06-20  Jim Meyering  <>

	build: update gnulib submodule to latest

	build: avoid configure/gnulib-related errors
	* bootstrap.conf: Remove now-unnecessary code to exclude
	gettext/intl-related m4 tests.

2011-06-19  Jim Meyering  <>

	maint: tighten up superfluous code
	* src/main.c (parse_grep_colors): Use xstrdup in place of xmalloc,
	a useless test, strlen, and strcpy.

2011-06-19  Paul Eggert  <>

	dfa: avoid possibility of overflow
	Use functions from xalloc.h to avoid overflow.
	* src/dfasearch.c (GEAcompile): Use xnrealloc rather than realloc.
	* src/pcresearch.c (Pcompile): Use xnmalloc, not xmalloc.

2011-06-17  Jim Meyering  <>

	build: update gnulib submodule to latest

	dfa: correct two uses of btowc
	* src/dfa.c (setbit_c, setbit_case_fold_c): Compare the btowc
	return value against WEOF, not EOF.  Suggested by Eli Zaretskii.
	On a system like MinGW with unsigned wint_t, comparing a btowc
	return value against EOF (-1) would always be false.

	dfa: don't overrun a malloc'd buffer for certain regexps
	* src/dfa.c (dfaanalyze): Allocate space for twice as many
	positions as there are leaves.  Before this change, for some
	regular expressions, DFA analysis would have inserted far more
	"positions" than dfa->nleaves (up to double).
	Reported by Raymond Russell in
	* tests/dfa-heap-overrun: Trigger the overrun.
	* tests/ (TESTS): Add it.
	* NEWS (Bug fixes): Mention it.

2011-06-08  Jim Meyering  <>

	tests: don't ignore sjis-mb test failure
	I made changes that caused grep to segfault during "make check" --
	as seen in dmesg output -- yet no test failed(!), and there was no
	trace of the segfault in the logs.
	* tests/sjis-mb (test_grep_reject): Ensure that output is empty.
	Don't ignore test failure.

2011-06-07  Paolo Bonzini  <>

	dfa: optimize wide characters in a bracket expression
	* src/dfa.c (addtok): Compile characters to an alternation.  Handle the
	case when nothing else remains in the MBCSET.

	dfa: refactor to prepare for upcoming optimizations
	* src/dfa.c (parse_bracket_exp): Move optimization of MBCSET from here...
	(addtok): ... to here.

2011-06-07  Paolo Bonzini  <>

	dfa: correct handling of single-byte character ranges
	This provides a better fix for the unibyte-bracket-expr and high-bit-range
	testcases, and fixes the latent bug tested by bogus-wctob.

	* src/dfa.c (setbit_case_fold): Remove, replace with...
	(setbit_wc, setbit_c, setbit_case_fold_c): ... these.
	(parse_bracket_exp): Use setbit_case_fold_c when iterating over
	single-byte sequences.  Use setbit_wc for multi-byte character sets,
	and setbit_case_fold_c for single-byte character sets.
	(lex): Use setbit_case_fold_c for single-byte character sets.

2011-06-07  Paolo Bonzini  <>

	tests: exercise latent bug in character ranges
	* tests/bogus-wctob: New.
	* (TESTS): Add it.

2011-06-07  Jim Meyering  <>

	tests: exercise a uni-byte [...] bug: requires ru_RU.KOI8-R
	* tests/unibyte-bracket-expr: New file.
	* tests/ (TESTS): Add it.
	* init.cfg (require_ru_RU_koi8_r): New function.

	fix the [...] bug also for relatively unusual uni-byte encodings
	* src/dfa.c (setbit_case_fold): Also handle uni-byte locales
	like the one mentioned in the original report: see 2011-05-07
	commit d98338eb.  Re-reported by Santiago Ruano Rincón.
	Note that most uni-byte locales are not affected.
	* NEWS (Bug fixes): Mention it.

	tests: use skip_test_, not skip_
	Use skip_test_, not skip_.  The former prints its message both to
	the log file and to FD 9 (redirected to tty via tests/,
	while skip_ prints only to stderr, which goes to the log file.
	* tests/init.cfg (skip_test_): New function.
	Use skip_test_ in place of skip_ everywhere.
	* tests/fmbtest: s/skip_/skip_test_/
	* tests/sjis-mb: Likewise.
	* tests/euc-mb: Likewise.

	tests: fmbtest: factor
	* tests/fmbtest: Factor out locale-name duplication.

	tests: fix skip-inducing typo in fmbtest
	* tests/fmbtest: Fix locale name typo (s/cz_CZ/cs_CZ/)
	that would cause this test to be skipped every time.

2011-06-07  Paolo Bonzini  <>

	gnulib: adjust included modules
	* bootstrap.conf (gnulib_modules): Drop strtoul, rename wctype to

2011-05-21  Jim Meyering  <>

	grep -P: don't abort upon exceeding PCRE's backtracking limit
	* src/pcresearch.c (Pexecute): Handle PCRE_ERROR_MATCHLIMIT.
	* tests/ (XFAIL_TESTS): Remove pcre-abort.
	* tests/pcre-abort: Expect failure, no output, and increase
	the length of the input string, in case the backtracking limit
	is ever raised.  Adjust comment.
	* NEWS (Bug fixes): Mention it.

	tests: show how to make grep -P abort
	* tests/pcre-abort: New file.
	Minimal testcase by Paolo Bonzini, derived from a report
	* tests/ (TESTS): Add it.
	(XFAIL_TESTS): Add it here, too, since this test always fails, for now.

	tests: fix oddities in pcre-z
	* tests/pcre-z: Redirect stderr inside $(), not outside.
	Remove double quotes around $REGEX (which is just 'a') within
	double-quoted "$(...)".  Split a long line.

	tests: factor out a new require_pcre_ function
	* tests/init.cfg (require_pcre_): New function, factored out of...
	* tests/pcre-z:  Use the function.
	* tests/pcre: Likewise.

	tests: clean up pcre
	* tests/pcre: Skip (don't pass) the test when PCRE support is disabled.
	Don't redirect so much to /dev/null, now that all test output goes to
	pcre.log.  Remove unnecessary braces and diagnostic about failing test.

2011-05-13  Jim Meyering  <>

	post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.8
	* NEWS: Record release date.

	build: update gnulib, for fixed getcwd test

	build: update gnulib submodule to latest

	maint: remove syntax-checking sc_tight_scope rule
	* src/ (sc_tight_scope): Remove rule.
	Now it's provided via gnulib's
	* (sc_tight_scope): Likewise.

2011-05-08  Jim Meyering  <>

	maint: use consistent declaration syntax
	* src/grep.h (matchers): Declare consistently, so the sc_tight_scope
	rule detects this as an extern-marked variable.

2011-05-07  Jim Meyering  <>

	maint: use gnulib's new readme-release module
	* bootstrap.conf (gnulib_modules): Add readme-release.
	(bootstrap_epilogue): Add the recommended perl one-liner.
	* README-release: Remove file; it is now generated from gnulib.
	* .gitignore: Add it.
	* gnulib: Update submodule to latest.

	tests: exercise bug with 0x80..0xff in [...]
	* tests/high-bit-range: New test, inspired by an example in the
	report by Igor O. Ladygin:,
	via Santiago Ruano Rincón's
	* tests/ (TESTS): Add it.

	fix a bug whereby echo c|grep '[c]' would fail for any c in 0x80..0xff
	* src/dfa.c (setbit_case_fold) [MBS_SUPPORT]: Set the bit also
	when wctob returns EOF.
	* NEWS (Bug fixes): Mention it.

2011-05-02  Reuben Thomas  <>

	doc: correct comment about mmap
	* doc/grep.texi (Other Options) [--mmap]: This option is now
	ignored, so using it can have no effect on performance.

2011-05-02  Arnold D. Robbins  <>

	build: move add_utf8_anychar into MBS ifdef

2011-05-01  Arnold D. Robbins  <>

	maint: remove GAWK ifndef; no longer needed

2011-05-01  Jim Meyering  <>

	maint: remove now-unnecessary use of gnulib's strtol module
	* bootstrap.conf (gnulib_modules): Remove now-obsolete "strtol".

2011-04-29  Jim Meyering  <>

	maint: tweak README-release
	* README-release: Add note to check the NixOS/Hydra autobuilder results.

2011-04-28  Jim Meyering  <>

	build: update gnulib submodule to latest

	maint: add the tight_scope syntax-checking rule
	This ensures that the only externally scoped symbols are ones
	that are explicitly marked as "extern" or white-listed like "main".
	* src/ (sc_tight_scope): New rule, copied from coreutils.
	* (sc_tight_scope): Define, to hook to it from the top level.

	maint: mark some function declarations as extern
	* src/search.h: Add "extern" keyword to each function declaration.

2011-04-23  Jim Meyering  <>

	maint: fix doubled-word typos in comments
	* src/dfa.c (SUCCEEDS_IN_CONTEXT): Remove doubled "a".
	* src/dfa.c (BACKREF): s/it it/it is/

2011-04-09  Jim Meyering  <>

	maint: fix typos in comments: s/can not/cannot/
	* src/dfa.c (check_matching_with_multibyte_ops, dfastate): As above.

2011-03-19  Jim Meyering  <>

	maint: stop using .x-sc_* files to list syntax-check exemptions
	Instead, use the new mechanism with which you merely use a
	variable (derived from the rule name) defined in to an ERE
	matching the exempted file names.
	* gnulib: Update to latest, to get that implements this.
	* .x-sc_bindtextdomain: Remove file.
	* .x-sc_prohibit_tab_based_indentation: Likewise.
	* .x-sc_prohibit_xalloc_without_use: Likewise.
	* .x-sc_space_tab: Likewise.
	* Define variables to exempt the same files.

	build: correct my change of 2011-01-28
	Do not override original dist-hook rule.
	* (run-syntax-check): Rename from overriding dist-hook.
	(dist-hook): Depend on run-syntax-check.

2011-02-27  Jim Meyering  <>

	maint: update from gnulib
	* bootstrap: Update from gnulib.
	* tests/ Likewise.
	* gnulib: Update to latest.

2011-01-27  Jim Meyering  <>

	build: update gnulib submodule to latest

	build: run syntax-check rules as part of "make dist"
	* (dist-hook): Depend on syntax-check.
	Suggested by Reuben Thomas.

2011-01-26  Jim Meyering  <>

	maint: remove unneeded #include directives
	* lib/savedir.c: Don't include <stddef.h>.  Not needed.
	* src/dfa.c: Likewise.

2011-01-22  Jim Meyering  <>

	build: avoid new syntax-check failures
	* .x-sc_bindtextdomain: New file, used to avoid a spurious
	failure from the new syntax-check rule.
	* NEWS: Remove a trailing space.

2011-01-19  Jim Meyering  <>

	tests: add a known-to-fail test
	* tests/turkish-I: New test.
	* tests/ (TESTS): Add it.
	(XFAIL_TESTS): Add here, too.
	Reported by Ilya Basin.

	maint: sort test names in
	* tests/ (TESTS): Sort test names.

2011-01-05  Jim Meyering  <>

	doc: remove erroneous "{,m}" item from grep man page
	* doc/ Remove item describing bogus {,m} regex notation.
	Reported by Fernando Basso.

2011-01-03  Jim Meyering  <>

	maint: update copyright year ranges to include 2011
	Run "make update-copyright", so "make syntax-check" works in 2011.

	build: update gnulib submodule to latest

2010-12-20  Paolo Bonzini  <>

	main: fix exit status on xmalloc failures
	* NEWS: Update.
	* src/main.c (main): Set exit_failure.  Reported by Guy Shaw.

	add comment above fn_grep
	* (fn_grep): Add comment suggested by Bruno Haible.

2010-11-14  Paolo Bonzini  <>

	grep: add include guards
	* src/system.h: Add multiple inclusion guards.
	* src/grep.h: Likewise.

	configure: fix M4 quotation
	* Add extra brackets around [...] patterns.

	configure: remove dependency on grep that supports long lines and -e
	* (fn_grep): New.  Set GREP and EGREP to it, replace
	with newly-built grep before AC_OUTPUT.  Reported by Florin Iucha

2010-11-04  Jim Meyering  <>

	build: update gnulib to latest

	tests: don't hard-code a 5-second timeout; that's not always enough
	Instead, time the command in the C locale and use 10 times that
	duration -- rounded up to whole seconds -- as the timeout when running
	it in the UTF-8 locale.
	* tests/backref-multibyte-slow: Compute a performance-relative timeout.
	Reported by Gilles Espinasse, regarding an imac 400.  For more details,

2010-10-09  Jim Meyering  <>

	maint: describe policy on copyright year number ranges
	* README: Mention coreutils' long-standing policy on use of M-N
	ranges in copyright year lists.  Requested by Richard Stallman.

2010-10-04  Dmitry V. Levin  <>

	build: compile gnulib without -Wcast-align to avoid warnings on ARM
	* (GNULIB_WARN_CFLAGS): Remove -Wcast-align.

2010-09-30  Jim Meyering  <>

	maint: don't define a gpg_key_ID.  now it's obtained automatically
	* (gpg_key_ID): Remove definition.  No longer needed.

2010-09-23  Paolo Bonzini  <>

	tests: add testcase for previous fix
	* tests/inconsistent-ranges: New.
	* tests/ (TESTS): Add it.

2010-09-23  Paolo Bonzini  <>

	dfa: process range expressions consistently with system regex
	The actual meaning of range expressions in glibc is not exactly strcoll,
	which makes the behavior of grep hard to predict when compiled with the
	system regex.  Leave to the system regex matcher the decision of which
	single-byte characters are matched by a range expression.

	This partially reverts a change made in commit 0d38a8bb (which made
	sense at the time, but not now that src/dfa.c is not doing multibyte
	character set matching anymore).

	* src/dfa.c (in_coll_range): Remove.
	(parse_bracket_exp): Use system regex to find which single-char
	bytes match a range expression.

2010-09-23  Bruno Haible  <>

	build: fix link error on systems that have libiconv but not libintl
	* src/ (LDADD): Add $(LIBICONV).

2010-09-21  Jim Meyering  <>

	build: avoid compilation failure on the Hurd
	* src/dfasearch.c (dfawarn): Rename enum symbols to use DW_ prefix,
	so as not to collide with "GNU", which is defined by the Hurd.
	Reported by Matthias Lanzinger in

2010-09-20  Jim Meyering  <>

	maint: avoid obsolete gnulib modules
	* bootstrap.conf (gnulib_modules): Don't use obsolete atexit module.
	Use malloc-gnu and realloc-gnu -- malloc and realloc are obsolete.

	maint: update README-release
	* README-release: Reflect changes in coreutils' version of this file.

2010-09-20  Aharon Robbins  <>

	dfa: fix compilation when not using MBS
	* src/dfa.c (prepare_wc_buf) [!MBS_SUPPORT]: Do not compile this

2010-09-16  Jim Meyering  <>

	post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.7
	* NEWS: Record release date.

2010-09-13  Paolo Bonzini  <>

	tests: add equiv-classes
	* (USE_INCLUDED_REGEX): Add Automake conditional.
	* tests/equiv-classes: New test.
	* tests/ (TESTS): Add it.
	(XFAIL_TESTS) [USE_INCLUDED_REGEX]: Mark it as expected failure.

2010-09-13  Paolo Bonzini  <>

	dfa: fall back to glibc matcher if a MBCSET is found
	This patch enables full support of equivalence classes and multicharacter
	collation symbols.  It can also improve performance problems in some
	cases for multibyte grep.  Both of these changes however depend on the
	glibc version installed in the system.

	For UTF-8 it will trigger only in the presence of MBCSET, e.g. [a-z].
	For other character sets all brackets and `.` as well will trigger it.

	* NEWS: Document this.
	* src/dfa.c (dfaexec): Fall back to glibc for multibyte matches,
	if possible.

2010-09-13  Paolo Bonzini  <>

	build: update gnulib submodule to latest
	This is done to include commit "regex: Pass the system regex if its only
	problem is 32-bit regoff_t".

	* gnulib: Update to e2b0e1a.

2010-09-12  Jim Meyering  <>

	build: update gnulib submodule to latest

	tests: update from gnulib
	* tests/ Update from gnulib.

2010-09-08  Patrick Boyd  <>

	dfa: reduce stack usage
	* src/dfa.c (dfaanalyze): Allocate GRPS and LABELS arrays from heap,
	not on the stack.  With this change, grep can now run in these UEFI

2010-09-08  Jim Meyering  <>

	tests/portability: avoid spurious failure with OpenBSD's /bin/sh
	* tests/warn-char-classes: Don't use "set -x" here.  It causes
	a spurious test failure on openbsd 4.7 when using its /bin/sh,
	since the command, /bin/sh -xc 'P=1 : 2> err' emits "P=1" into err.
	To enable set -x, run the test with "VERBOSE=yes", e.g.,
	make check -C tests TESTS=warn-char-classes VERBOSE=yes

2010-09-07  Jim Meyering  <>

	build: update gnulib submodule to latest

2010-09-03  Jim Meyering  <>

	tests: remove .sh suffix from remaining test scripts.
	* tests/backref: Rename from
	* tests/bre: Rename from
	* tests/ere: Rename from
	* tests/file: Rename from
	* tests/khadafy: Rename from
	* tests/options: Rename from
	* tests/pcre: Rename from
	* tests/spencer1: Rename from
	* tests/spencer2: Rename from
	* tests/status: Rename from
	* tests/yesno: Rename from
	* tests/ Reflect renamings.

	tests: convert remaining tests to use
	* tests/ Use Use Exit, not exit. Use grep, not ${GREP}.
	* tests/ Likewise.
	* tests/ Likewise.
	* tests/ Likewise.
	* tests/ Likewise.
	* tests/ Likewise.
	* tests/spencer1.awk: Use grep, not ${GREP}.
	Don't ignore failure to generate intermediate shell script.
	* tests/ (CLEANFILES): Remove altogether, now that
	all tests use
	(TESTS_ENVIRONMENT): Don't set GREP.  It's no longer used.

	tests: remove
	* tests/ Remove file.  All it did was print a warning.
	* tests/ (TESTS): Remove

	tests: convert to use
	* tests/ Use Use Exit, not exit. Use grep, not ${GREP}.

	tests: convert to use
	* tests/ Use
	Use Exit, not exit.
	Use "$abs_top_srcdir/tests/", not "$srcdir/" to specify inputs.
	Source generated bre.script, rather than invoking $SHELL.
	* tests/ Likewise.
	* tests/bre.awk: Use grep, not ${GREP}.
	* tests/ere.awk: Likewise.
	* tests/ (CLEANFILES): Remove bre.script and ere.script.

	tests: convert to use
	* tests/ Use
	Use Exit, not exit.
	Use grep, not $GREP.
	* tests/ Likewise.
	* tests/ (CLEANFILES): Remove yesno.txt.

	build: update gnulib submodule to latest

	build: update build/test tools from gnulib
	* bootstrap: Update from gnulib.
	* tests/ Likewise.

2010-09-01  Jim Meyering  <>

	maint: add lib/version-etc.c to the list in
	* po/ Add lib/version-etc.c.

2010-09-01  Jim Meyering  <>

	grep: diagnose and exit-2 for bogus REs like [:space:], [:digit:], etc.
	When I make a mistake like this:
	  grep '[:lower:]' ...
	be it in a script or on the command line, I want to know about
	it as soon as possible.  I don't want grep to print a mere warning
	that it is interpreting this suspicious and almost guaranteed-wrong
	regular expression as a set of just 6 bytes.  And I certainly don't
	want grep to silently do the wrong thing, even if that would be
	officially standards-conforming.  It's obvious that I intended
	[[:lower:]], and I want my error to be diagnosed in a way that is
	most likely to get my attention.  Thus, with this change, grep now
	prints a diagnostic and exits with status 2 the moment it
	encounters an offending [:char_class:] construct.

	This changes the way grep works by default, rather than
	putting this new behavior on an option.  A new option
	would seldom be used in scripts (not portable), and would
	probably be used only rarely by those who need it the most.
	This new functionality provides a valuable safety measure
	and incurs truly negligible risk.

	For strict POSIX compliance, set POSIXLY_CORRECT in
	your environment.  That disables this new feature.

	Revert the changes from commit 2cd3bcea, "grep: add
	--warnings={always,never,auto}.", and then do the following:

	* src/dfasearch.c (dfawarn): Call getenv("POSIXLY_CORRECT") here;
	Remove "warning: " from the diagnostic, now that it's more than
	a warning, and exit with status 2.
	* NEWS (New features): Describe the new semantics.
	* tests/warn-char-classes: Adjust one test to accommodate this change.
	* doc/grep.texi (Character Classes and Bracket Expressions): Document.
	(Environment Variables): Cross-reference it.
	Remove reference to obsolete getopt illegal vs. invalid difference.
	Thanks to Paul Eggert for suggestions and an initial prod.

2010-08-30  Jim Meyering  <>

	maint: use gnulib's standard --version-printing code
	This includes author names and keeps the copyright year up to date.
	* bootstrap.conf (gnulib_modules): Add propername and version-etc-fsf.
	* src/main.c (AUTHORS): Define.
	(main): Use version_etc, rather than hard-coding the copyright text.
	Prompted by a patch from Paolo Bonzini.

2010-08-27  Paolo Bonzini  <>

	dfa: warn on [:space:] and similar
	* src/dfa.c (parse_bracket_exp): Warn on regular expressions such as
	* src/dfa.h (dfawarn): New prototype.
	* src/dfasearch.c (dfawarn): New.
	* NEWS: Document.

	tests: add test for warnings
	* tests/ (TESTS): Add warn-char-class.
	* tests/warn-char-class: New.

	grep: add --warnings={always,never,auto}.
	* src/grep.h (no_warnings): New declaration.
	* src/main.c (no_warnings): New.
	(WARNINGS_OPTION): Add to enum.
	(main): Add --warnings.  Handle color_option == 2 together with it.

	tests: add failing test for grep from a directory
	* tests/ (TESTS, XFAIL_TESTS): Add grep-dir.
	* tests/grep-dir: New.

	tests: add test for previous commit
	* tests/ (TESTS): Add grep-dev-null.
	* tests/grep-dev-null: New.

	search: fix "grep -Fif /dev/null"
	* bootstrap.conf: Include gnulib module minmax.
	* src/searchutils.c (mbtolower): Handle *N == 0 case.
	* src/system.h: Include minmax.h from gnulib.

2010-08-27  Adam Katz  <>

	Remove declaration after statement in dfa.c
	* dfa.c (dfaexec): Declare saved_end at the beginning of the function.

2010-08-13  Jim Meyering  <>

	make --include=FILE work once again
	The semantics of excluded_file_name changed (when operating on
	an "included" file name list).
	* src/main.c (main): Adjust for changed semantics of excluded_file_name
	simply by removing a negation.
	* NEWS (Bug fixes): Mention this fix.
	* tests/include-exclude: Add a test for this.
	Reported by Joe Perches in

2010-07-16  Paolo Bonzini  <>

	doc: document \s and \S
	* doc/grep.texi (The Backslash Character and Special Expressions):
	Document \s and \S escapes.

2010-05-29  Karl Berry  <>

	doc: discuss matches that span two or more lines
	* doc/grep.texi (Usage): Discuss matching across lines.
	(Character Classes and Bracket Expressions) <[:space:]>: refer to it.

2010-05-25  Jim Meyering  <>

	build: use latest gettext: 0.18
	* Use gettext-0.18.
	* bootstrap.conf (gnulib_modules): Use gettext-h, not gettext.
	since the latter drags in a depedency on gettext 0.18.
	Suggested by Bruno Haible.

	maint: update helper scripts from gnulib
	* tests/ Update from gnulib.
	* bootstrap: Likewise.

	build: update gnulib submodule to latest

	maint: don't emit an extra newline in each of two diagnostics
	* src/main.c (context_length_arg, grepdir): Remove a stray \n in
	each of two diagnostics.

2010-05-24  Bruno Haible  <>

	search: Avoid out-of-bounds access.
	* src/dfasearch.c (EGexecute): Avoid access beyond end of buffer
	that could happen if start != beg - buf.

2010-05-23  Aharon Robbins  <>

	dfa: fix signedness warnings
	* src/dfa.c (dfaexec): Cast p when passing it to prepare_wc_buf.

2010-05-09  Jim Meyering  <>

	tests: update
	* tests/ Update from gnulib.

	tests: normalize code
	* tests/backref-multibyte-slow: Use one-line idiom.
	* tests/backref-word: Likewise.
	* tests/case-fold-backref: Likewise.
	* tests/case-fold-backslash-w: Likewise.
	* tests/case-fold-char-class: Likewise.
	* tests/case-fold-char-range: Likewise.
	* tests/case-fold-char-type: Likewise.
	* tests/char-class-multibyte: Likewise.
	* tests/dfaexec-multibyte: Likewise.
	* tests/empty: Likewise.
	* tests/euc-mb: Likewise.
	* tests/fedora: Likewise.
	* tests/fgrep-infloop: Likewise.
	* tests/fmbtest: Likewise.
	* tests/foad1: Likewise.
	* tests/ignore-mmap: Likewise.
	* tests/include-exclude: Likewise.
	* tests/max-count-vs-context: Likewise.
	* tests/pcre-z: Likewise.
	* tests/prefix-of-multibyte: Likewise.
	* tests/reversed-range-endpoints: Likewise.
	* tests/sjis-mb: Likewise.
	* tests/spencer1-locale: Likewise.
	* tests/word-delim-multibyte: Likewise.
	* tests/word-multi-file: Likewise.

	tests: update help-version
	* tests/help-version: Update from coreutils.

2010-05-06  Jim Meyering  <>

	tests: enable glibc's malloc-perturbing option
	* tests/ (MALLOC_PERTURB_): Define, in case it's not already
	set in your environment.
	(TESTS_ENVIRONMENT): Propagate MALLOC_PERTURB_ setting to test scripts.

2010-05-06  Paolo Bonzini  <>

	dfa: speed up [[:digit:]] and [[:xdigit:]]
	There's no "multibyte pain" in these two classes, since POSIX
	and ISO C99 mandate their contents.

	Time for "./grep -x '[[:digit:]]' /usr/share/dict/linux.words"
	Before: 1.5s, after: 0.07s.  (sed manages only 0.5s).

	* src/dfa.c (predicates): Declare struct dfa_ctype separately
	from definition.  Add sb_only.
	(find_pred): Return const struct dfa_ctype *.
	(parse_bracket_exp): Return const struct dfa_ctype *.  Do
	not fill MBCSET for sb_only character types.

2010-05-05  Jim Meyering  <>

	tests: readability: use awk rather than obfuscated sed
	* tests/backref-multibyte-slow: Generate input using an awk for-loop
	rather than expensive and harder-to-read sed pipes.
	Remove stray "set -x" and "wc -l in".

	dfa: avoid segfault when processing an invalid multi-byte sequence
	* src/dfa.c (dfaexec): Handle the cases in which mbrtowc returns
	(size_t)-1 or (size_t)-2, rather than setting mblen_buf[i] to an
	outrageously large value.

2010-05-05  Paolo Bonzini  <>

	grep: remove redundant syntax bit
	* grep.c (Gcompile): Remove RE_HAT_LISTS_NOT_NEWLINE.

	tests: add test for newly-fixed performance problem
	* tests/backref-multibyte-slow: New.
	* tests/ Add it.

2010-05-05  Paolo Bonzini  <>

	dfa: convert to wide character line-by-line
	This provides a nice speedup for -m in general, but especially
	it avoids quadratic complexity in case we have to go to glibc.

	* NEWS: Document change.
	* src/dfa.c (prepare_wc_buf): Extract out of dfaexec.  Convert
	only up to the next newline.
	(dfaexec): Exit multibyte processing loop if past buf_end.
	Call prepare_wc_buf again after processing a newline.

2010-05-01  Jim Meyering  <>

	maint: remove useless #if HAVE_STDLIB_H
	* src/mbsupport.h: Don't test HAVE_STDLIB_H.

2010-04-20  Jim Meyering  <>

	dfa: don't #ifdef-out member declarations
	* src/dfa.c (struct dfa): Remove "#if MBS_SUPPORT" guard that made
	several member declarations conditional on this cpp definition.
	(token): Likewise.
	Reported by Anders Wallin.

	tests: ensure that the --mmap option is ignored
	* tests/ignore-mmap: New file.
	* tests/ (TESTS): Add it.
	Reported by Jaroslav Škarvada in <>

2010-04-20  Paolo Bonzini  <>

	dfa: honor RE_DOT_NEWLINE and RE_DOT_NOT_NULL in UTF-8 period optimization
	* src/dfa.c (add_utf8_anychar): Check for RE_DOT_NEWLINE and

	grep: fix --mmap not being ignored
	* NEWS: Document bugfix.
	* main.c (main): Ignore MMAP_OPTION.

2010-04-19  Jim Meyering  <>

	maint: avoid syntax-check failure due to indentation via TABs
	* src/dfa.c (atom): Expand TABs in indentation.

	build: update gnulib submodule to latest

	maint: restrict scope of two globals to dfasearch.c
	* src/dfasearch.c (patterns, pcount): Declare these file-scoped
	globals to be static.

2010-04-19  Paolo Bonzini  <>

	dfa: optimize UTF-8 period
	* NEWS: Document improvement.
	* src/dfa.c (struct dfa): Add utf8_anychar_classes.
	(add_utf8_anychar): New.
	(atom): Simplify if/else nesting.  Call add_utf8_anychar for ANYCHAR
	in UTF-8 locales.
	(dfaoptimize): Abort on ANYCHAR.

	dfa: drop ORTOP
	* src/dfa.c (token, prtok, addtok_mb, nsubtoks, dfaanalyze, dfamust):
	Remove ORTOP.
	(regexp): Remove parameter, always add OR at the end, adjust callers.
	(atom): Adjust caller.
	(dfaparse): Adjust caller.  Always add OR at the end.

	dfa: fix {0,0}
	* NEWS: Document change.
	* src/dfa.c (struct dfa): Remove "broken" field.
	(lex): Do not set it.
	(closure): On {0,0}, backup and lex another closure without
	adding a CAT.
	(dfabroken): Remove.
	* src/dfa.h (dfabroken): Remove.
	* tests/spencer1.tests: Add testcases for {m,n}.

	dfa: simplify dfainit
	* src/dfa.c (dfainit): Use memset.

2010-04-17  Jim Meyering  <>

	doc: fix a nit in HACKING
	* HACKING: Correct size of .git/ dir: 9MB, not 30MB.

	tests: add an expected-to-fail test using \< in a multi-byte locale
	* tests/word-delim-multibyte: New test.  Currently failing.
	* tests/ (TESTS): Add it.
	(XFAIL_TESTS): Define, temporarily.
	Reported by Jaroslav Škarvada in

2010-04-16  Paolo Bonzini  <>

	test: cover just-fixed bug
	* tests/empty: Test -Fw too.

	grep: fix matching the empty string with grep -Fw
	* NEWS: Document fix.
	* src/kwsearch.c (Fexecute): The empty string is a valid match if it is
	a whole word.

2010-04-15  Jim Meyering  <>

	maint: update and HACKING
	* HACKING: Sync from coreutils.
	* tests/ Update from gnulib.

2010-04-13  Jim Meyering  <>

	build: update gnulib submodule to latest; adapt
	* COPYING: Remove empty line.
	* README: Likewise.
	* doc/fdl.texi: Likewise.
	* tests/backref-word: Likewise.

2010-04-11  Stefano Lattarini  <>

	tests: accept the Debian timeout program
	* tests/init.cfg: test timeout with `timeout 10s true'

2010-04-08  Jim Meyering  <>

	dfa: convert "cannot happen" code/comment to use assert
	* src/dfa.c (dfamust): There were numerous "cannot happen" comments,
	some associated with "if (expr) goto done;".  Replace each with an
	equivalent "assert (!expr);".

	build: use gnulib's isblank module
	* bootstrap.conf (gnulib_modules): Use gnulib's isblank module,
	now that we rely on the function by that name.

	maint: undo TAB-conversion change to gl/lib/*.c.diff
	This fixes a bootstrap failure due to the patches not applying.
	* .x-sc_prohibit_tab_based_indentation: Add ^gl/lib/.*\.c\.diff$
	* gl/lib/regcomp.c.diff: Revert today's TAB->space change.
	* gl/lib/regex_internal.c.diff: Likewise.
	* gl/lib/regexec.c.diff: Likewise.

2010-04-08  Arnold D. Robbins  <>

	dfa: fix declaration of dfabroken in dfa.h
	* dfa.h (dfabroken) [GAWK]: Fix declaration to match that in dfa.c.

2010-04-08  Jim Meyering  <>

	maint: add syntax-check rule to enforce the new no-leading-TABs policy
	* (sc_prohibit_tab_based_indentation): New rule, from coreutils.
	(sc_prohibit_emacs__indent_tabs_mode__setting): Likewise.
	(old_NEWS_hash): Update.
	* .x-sc_prohibit_tab_based_indentation: List exempt files.

2010-04-08  Jim Meyering  <>

	convert all TABs to equivalent spaces in indentation
	Using this file,

	cat > leading-blank.exempt <<\EOF

	run this command to convert all non-conforming leading white
	space to be all spaces:

	git ls-files \
	  | pcregrep -vf leading-blank.exempt \
	  | xargs pcregrep -l '^ *\t' \
	  | xargs perl -MText::Tabs -ni -le \
	    '$m=/^( *\t[ \t]*)(.*)/; print $m ? expand($1) . $2 : $_'

2010-04-08  Jim Meyering  <>

	build: include in the distribution tarball
	* (EXTRA_DIST): Add

2010-04-08  Jim Meyering  <>

	maint: tweak  (no semantic change)
	* (EXTRA_DIST): List one per line.  Sort.

	build: include in the distribution tarball
	* (EXTRA_DIST): Add

2010-04-08  Jim Meyering  <>

	dfa: move definition of __attribute__ back into dfa.h
	* src/dfa.c (__attribute__): Move definition back to...
	* src/dfa.h: ... this file.  It is essential for non-gcc compilers.
	Reported by Arnold Robbins.

2010-04-07  Arnold D. Robbins  <>

	dfa: move internals from dfa.h to dfa.c
	* src/dfa.h: Move internals into dfa.c.
	* src/dfa.c: The dfa internals are now totally local to this file.
	(dfaalloc, dfamusts, dfabroken): New functions to access features.
	* src/dfasearch.c (dfa): Change this global variable from struct to pointer.
	Adapt to that change, and use new functions, dfamusts and dfaalloc.

2010-04-07  Jim Meyering  <>

	mbtolower: avoid potential NULL-dereference
	* src/searchutils.c: Include <assert.h>.
	(mbtolower): Assert that 0 < *n, to avoid possibility of NULL-deref.
	Remove dead increment.

	maint: tell git to ignore more build products
	* .gitignore: Also ignore results of "make ID" and "make tags".

	build: update gnulib submodule to latest

	tests: use consistently
	* tests/euc-mb: Call "path_prepend_ ." on a line by itself,
	and with a comment.  This makes it so all of the srcdir/
	lines are consistent, project-wide, and so that the addition of "."
	to PATH for this test is properly documented.
	* tests/sjis-mb: Likewise.

	maint: avoid new syntax-check failure, ... that the sole use of xmalloc no longer matches the
	regular expression used by the syntax-check rule.
	* .x-sc_prohibit_xalloc_without_use: Exempt src/kwset.c.

	grep: make kwset's obstack use xmalloc, not malloc
	This insidious bug could make grep fail to diagnose a failed malloc,
	and then proceed to dereference the resulting NULL pointer.
	Note that this bug was unlikely ever to cause real trouble; without
	the fix, grep would segfault upon OOM, now it exits with a diagnostic.
	* src/kwset.c (malloc) [GREP]: Define without the "(s)" macro
	parameter, so that unadorned uses of malloc are also mapped to xmalloc.
	One such use is in the expansion of obstack_init.
	Report and patch by Nelson H. F. Beebe, in

	tests: improve help-version (sync from gzip's version)
	* tests/help-version: Cross-check $VERSION and --version output.

2010-04-06  Jim Meyering  <>

	doc: update THANKS
	* THANKS: Update.

2010-04-06  Aharon Robbins  <>

	build: avoid conflict with WCHAR definition from Cygwin's <windows.h>
	* src/dfa.h (enum token): Remove the definition from this file.
	Replace with a declaration and typedef.  Moved to ...
	* src/dfa.c (enum token): ... here.
	Reported by Corinna Vinschen.

2010-04-06  Jim Meyering  <>

	doc: add HACKING
	* HACKING: New file.  Copied from coreutils, with s/coreutils/grep/
	and a few minor edits.

2010-04-05  Jim Meyering  <>

	tests: pull fixed from gnulib
	* tests/ Update from gnulib.

	maint: fix new argmatch-related syntax-check failures
	* (ARGMATCH_DIE): Use usage(EXIT_FAILURE), not exit(1).
	* po/ Add lib/argmatch.c.

	maint: update to work with gnulib's newer "make syntax-check"
	* Update to use new _sc_search_regexp interface.  Run this:
	perl -pi -e 's/\b_prohibit_regexp\b/_sc_search_regexp/;'
	-e 's/\bmsg=/halt=/; s/\bre=/prohibit=/;'
	and then adjust backslashes so they still line up.

	maint: update tests/ from gnulib
	This ensures that the explanation for any skipped or failed test
	is printed on stderr, not buried in each .log file.
	* tests/ Update from gnulib.
	* tests/init.cfg (stderr_fileno_): Define to 9, to match the
	literal 2>&9 in tests/

	build: update gnulib submodule to latest

2010-04-04  Jim Meyering  <>

	maint: use argmatch, for better --directories=INVAL diagnostics
	Before, you'd see this:
	  grep: unknown directories method

	Now, you'll see this:
	  grep: invalid argument `INVAL' for `--directories'
	  Valid arguments are:
	    - `read'
	    - `recurse'
	    - `skip'
	  Usage: src/grep [OPTION]... PATTERN [FILE]...
	  Try `src/grep --help' for more information.

	* bootstrap.conf: Add argmatch.
	* src/main.c (directories_type): Define.
	(directories_args, directories_types) Define.
	All of the above so we can...
	(main): Use XARGMATCH.
	(usage): Declare extern, now that argmatch calls it via ARGMATCH_DIE.

2010-04-04  Jim Meyering  <>

	dfa.c: const correctness; and remove useless casts of realloc and malloc
	* src/dfa.c (icatalloc, icpyalloc, istrstr, enlist): As above.
	(inboth, dfamust, comsubs): Likewise.

	dfa.c: use a better (unsigned) type for an index: int->unsigned int
	* src/dfa.c (dfaexec): Use "unsigned int" for a logically unsigned index.

	maint: style: use sizeof VAR, rather than sizeof TYPE, where possible
	* src/dfa.c (copyset, zeroset): Prefer sizeof EXPR, over sizeof TYPE,
	for improved readability/maintainability.
	(equal, parse_bracket_exp, addtok_wc, dfaparse, dfaexec): Likewise.

2010-04-02  Jim Meyering  <>

	dfa.c: use a better (unsigned) type for an index: int->size_t
	* src/dfa.c (parse_bracket_exp): Use size_t as type of index, not int.

	maint: const-correctness
	* src/dfa.c (tstbit, copyset, equal, charclass_index): Declare read-only
	"charclass" parameters to be "const".  No semantic change.

	maint: include <wchar.h> and <wctype.h> unconditionally
	* src/main.c: Include <wchar.h> and <wctype.h> unconditionally.
	Their presence/usefulness are assured by gnulib.
	* src/dfa.c: Likewise.
	* src/search.h: Likewise.

	maint: MBS_SUPPORT: define to 0/1, not undef/1
	Prepare to remove many of these #ifdefs.
	* src/mbsupport.h (MBS_SUPPORT): Define to 0/1, not undef/1.
	Change each "#ifdef MBS_SUPPORT" to "#if MBS_SUPPORT".  Use this:
	perl -pi -e 's/ifdef (MBS_SUPPORT)/if $1/' $(g grep -l ifdef.MBS_SUPPO)
	* src/dfa.c: s/#ifdef MBS_SUPPORT/#if MBS_SUPPORT/
	* src/dfa.h: Likewise.
	* src/dfasearch.c: Likewise.
	* src/kwsearch.c: Likewise.
	* src/main.c: Likewise.
	* src/search.h: Likewise.
	* src/searchutils.c: Likewise.

2010-04-02  Jim Meyering  <>

	maint: use STREQ in place of strcmp
	perl -pi -e 's/\bstrcmp *\((.*?)\) == 0/STREQ ($1)/' src/main.c
	perl -pi -e 's/\bstrcmp *\((.*?)\) != 0/!STREQ ($1)/' src/main.c

	* src/dfa.c (STREQ): Define.
	Use it instead of strcmp.
	* src/main.c (STREQ): Likewise.
	* (local-checks-to-skip): Remove sc_prohibit_strcmp,
	to enable the strcmp-prohibition.

2010-04-02  Jim Meyering  <>

	maint: enable the useless_cpp_parens syntax check
	* (local-checks-to-skip): Remove sc_useless_cpp_parens.
	* src/main.c (devices, fillbuf, exit_on_match): Remove useless parens.
	(print_line_head, grepfile, set_limits, main): Likewise.
	* src/vms_fab.h: Likewise.
	* vms/config_vms.h: Likewise.
	* src/mbsupport.h: Likewise.

	cleanup and improvement: parse command line arguments consistently
	* src/main.c: Include c-ctype.h, for this:
	(prepend_args): Use c_isspace, not ISSPACE.
	This is important so that we parse arguments consistently,
	and independently of the current locale.
	* bootstrap.conf (gnulib_modules): Add c-ctype.
	* src/system.h: Remove IS* definitions here, too.
	* src/dfasearch.c (WCHAR): Use isalnum, not ISALNUM.
	* src/kwsearch.c (WCHAR): Likewise.
	* src/searchutils.c (kwsinit): Use tolower, not TOLOWER.

	cleanup: rely on gnulib's ctype.h functions; remove IS* macros and is_*
	* src/dfa.c (setbit_case_fold, prednames): Use official names.
	(IS_WORD_CONSTITUENT, lex): Likewise.
	(is_alnum, is_alpha, is_blank, is_cntrl, is_digit, is_graph): Likewise.
	(is_lower, is_print, is_punct, is_space, is_upper, is_xdigit): Likewise.
	(isgraph): Likewise.

	build: update gnulib submodule to latest, and adjust
	* src/main.c (parse_grep_colors): Adjust diagnostics not to trigger
	the sc_error_message_period and sc_error_message_uppercase
	syntax-check rules.

	maint: remove all VMS-related code
	* (AC_CONFIG_FILES): Remove vms/Makefile
	* (SUBDIRS): Remove vms.
	* src/ (EXTRA_DIST): Remove vms_fab.c and vms_fab.h.
	* src/vms_fab.c, src/vms_fab.h, vms/ Remove files.
	* vms/, vms/README, vms/config_vms.h: Likewise.

	post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.6.3
	* NEWS: Record release date.

2010-04-02  Jim Meyering  <>

	grep: avoid used-undefined error with truncated multibyte input
	* src/dfa.c (addtok_wc): Don't use buf[0] (it's undefined) when
	wcrtomb returns <= 0.

	MBS_SUPPORT-removal: * src/dfa.c (dfastate):

2010-04-01  Jim Meyering  <>

	maint: avoid unnecessary 2nd getenv("TERM")
	* src/main.c (main): Don't call getenv("TERM") twice -- in the same
	expression, even.

	tests: remove all unportable uses of echo
	* src/main.c: Use printf rather than echo -ne in a comment.
	* tests/fedora: Use printf (not echo) also in ok/fail functions.
	* (sc_prohibit_echo_minus_en): New rule, to prohibit
	any future introduction.

	tests: add explicit requirement for en_US.UTF-8
	* tests/char-class-multibyte: Use require_en_utf8_locale_,
	rather than open-coding it.
	* tests/prefix-of-multibyte: Require the locale explicitly.
	* tests/fgrep-infloop: Likewise.
	This fixes test failures that would arise on systems without
	that particular locale.  Reported by Ludovic Courtès.

	tests: new function, to require an en_US UTF8 locale
	* tests/init.cfg (require_en_utf8_locale_): New function.

	tests: use printf, not echo -n, echo -e, or any combination
	* tests/fedora: Using printf is more portable.

	grep: remove unnecessary code
	* src/main.c (print_line_middle): Now that we use RE_ICASE
	(enabled in commit 70e23616, "dfa: rewrite handling of multibyte
	case_fold lexing"), this case-conversion code is useless and wasteful.
	Remove it.

	doc: fix typo: s/AM_V_AT/AM_V_at/
	* doc/ (egrep.1 fgrep.1): The former has case consistent
	with its sister variable, AM_V_GEN, but the latter is the one that
	actually works.

	doc: generated files are best made read-only, ... minimize risk of accidentally modifying the generated file
	rather than its template.  These are tiny, so no risk, but it's
	a good to be consistent, so generated files are easier to spot.
	* doc/ (egrep.1 fgrep.1): When generating these files,
	ensure that they too are created read-only.

	doc: generate grep.1 from template
	* doc/ (grep.1): New rule.
	(CLEANFILES): Add grep.1 to the list.
	* .gitignore: Add /doc/grep.1
	* doc/ Replace hard-coded "2.5.1-cvs" with @VERSION@.
	Update copyright year list.
	Omit the line-splitting \(co directive so that update-copyright
	will perform future updates automatically.
	Egmont Koblinger reported the outdated version string
	and copyright year list in the man page:

	doc: prepare to generate grep.1
	* doc/grep.1: Rename to...
	* doc/ ...this.

2010-03-31  Eric Blake  <>

	build: avoid another warning
	Noticed on cygwin:
	get-mb-cur-max.c: In function 'main':
	get-mb-cur-max.c:27: error: unused parameter 'argc' [-Wunused-parameter]

	* tests/get-mb-cur-max.c (main): Use argc.

2010-03-31  Paolo Bonzini  <>

	tests: fix on systems with broken sh
	* tests/ (TESTS_ENVIRONMENT): Adjust coreutils remnants.
	* tests/ Invoke script with $SHELL if defined.
	* tests/ Likewise.
	* tests/spencer1-locale: Likewise.
	* tests/ Likewise.

	tests: improve empty test
	* tests/empty: Add more tests, note expected failure.

	tests: improve empty test with respect to locales
	* tests/empty: Add tests for multiple locales.

	grep: fix grep -F against empty string
	* src/searchutils.c (is_mb_middle): Do not return true for empty matches
	when p == buf.

	tests: rename to empty
	* tests/ Rename to...
	* tests/empty: ... this.
	* tests/ (TESTS): Adjust.

	tests: convert to new style
	* tests/ Convert to, add 10-second timeout.

	tests: use get-mb-cur-max in char-class-multibyte
	* tests/char-class-multibyte: Use get-mb-cur-max to detect UTF-8 support.
	Rewrite previous locale detection code as a grep test.

	tests: fix -Wformat failure
	* tests/get-mb-cur-max (main): Cast MB_CUR_MAX to int.

2010-03-30  Jim Meyering  <>

	doc: add a "Reply-To" to the suggested announcement mail header
	* README-release: Add "Reply-To" with the list address,
	to minimize risk of replies to the other announcement recipients.
	Suggestion from Eric Blake.

2010-03-29  Jim Meyering  <>

	build: avoid compiler warning when building test program
	so that all the usual C compile-and-link machinery comes into play.
	* tests/get-mb-cur-max.c: Include "progname.h".
	Remove unnecessary inclusion of <ctype.h>.
	Mike Frysinger reported the "implicit decl of set_program_name" warning.

	build: detect PCRE support also when <pcre/pcre.h> is the header
	* m4/pcre.m4: Also check for <pcre/pcre.h>.
	* src/pcresearch.c: Include <pcre/pcre.h>, if needed.
	Guard inclusions with HAVE_PCRE_H and HAVE_PCRE_PCRE_H, not HAVE_LIBPCRE.
	* NEWS (Bug fixes): Mention it.
	Dmitry V. Levin reported that PCRE support was not detected
	on systems with <pcre.h> not in the default include path.

	post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.6.2
	* NEWS: Record release date.

2010-03-29  Eric Blake  <>

	build: avoid warnings on cygwin
	* lib/savedir.c (isdir): Avoid shadowing a declaration.
	* src/main.c (get_nondigit_option): Cast away const to avoid
	compiler warning.

	maint: ignore new test executable
	* .gitignore: Enhance.

2010-03-29  Jim Meyering  <>

	doc: consolidate redundant-looking entries
	* NEWS: Consolidate the two --include/exclude-related entries.
	Suggested by Eric Blake.

2010-03-29  Paolo Bonzini  <>

	tests: use $(...) consistently
	* tests/ Use `...' instead of ``...'' in comments.
	* tests/bre.awk: Use $(...) instead of `...`.
	* tests/ere.awk: Use $(...) instead of `...`.
	* tests/euc-mb: Use $(...) instead of `...`.
	* tests/fmbtest: Use $(...) instead of `...`.
	* tests/foad1: Use $(...) instead of `...`.
	* tests/pcre-z: Use $(...) instead of `...`.  Quote output of grep.
	* tests/spencer1-locale.awk: Use $(...) instead of `...`.
	* tests/spencer1.awk: Use $(...) instead of `...`.
	* tests/ Use $(...) instead of `...`.

2010-03-29  Jim Meyering  <>

	build: make doc/ cleaner and more robust
	* doc/ (egrep.1 fgrep.1): Generate robustly, i.e.,
	do not redirect directly to $@.
	Use $(AM_V_GEN).
	Do not distribute intermediate files like and
	Likewise, do not use them to generate their %.1 images.
	Instead, generate the .1 files directly.

2010-03-29  Paolo Bonzini  <>

	tests: add program to detect locales
	* tests/ (check_PROGRAMS): Add get-mb-cur-max.
	* tests/get-mb-cur-max.c: New.
	* tests/euc-mb: Use it.  Fail if the former detection test fails.
	* tests/sjis-mb: Use it.  Fail if the former detection test fails.  Expand

2010-03-29  Paolo Bonzini  <>

	tests: add tests for SJIS character sets
	The attached test will be skipped unless (on a glibc system) you run
	something like

	  mkdir /usr/lib/locale/ja_JP.SHIFT_JIS
	  zcat /usr/share/i18n/charmaps/SHIFT_JIS.gz | \
	    localedef \
	      -f - \
	      -i /usr/share/i18n/locales/ja_JP \

	* tests/ Add sjis-mb.
	* tests/sjis-mb: New.

2010-03-29  Paolo Bonzini  <>

	grep -F: fix a bug with SJIS character sets
	Commit db9d6 would erroneously skip matches in SJIS character sets.  In
	this character set low bytes (i.e. ASCII bytes) are also valid second
	bytes in a double-byte character, so you have to continue looking for
	a match, even if you match in the middle of a double-byte character.

	* src/kwsearch.c: Ensure that beg is advanced by at least one byte,
	but do not fail immediately after matching in the middle of a double-byte

2010-03-28  Bruno Haible  <>

	build: update after change in gnulib's lib-ignore module
	* src/ (AM_LDFLAGS): Define.  Use gnulib's new

2010-03-28  Jim Meyering  <>

	tests: disable new texinfo-acronym syntax-check from gnulib
	* (local-checks-to-skip): Add new sc_texinfo_acronym, to skip it.

2010-03-28  Norihiro Tanaka  <>

	tests: exercise fix for improper match of incomplete MB char prefix
	* tests/prefix-of-multibyte: New file.
	* tests/ (TESTS): Add it.

2010-03-28  Jim Meyering  <>

	grep -F: fix a multi-byte erroneous-match-in-middle bug
	Just as Perl prints nothing in this case,
	  printf '\357\274\241\n' | perl -CIO -lne '/\357/ and print'

	grep should also print nothing when used as follows.
	However, these would mistakenly match with grep prior to 2.6.2:
	  printf '\357\274\241\n' | LC_ALL=en_US.UTF-8 src/grep -F $'\357'
	  printf '\357\274\241\n' | LC_ALL=en_US.UTF-8 src/grep -F $'\357\274'

	* src/searchutils.c (is_mb_middle): New parameter: the length of the
	match, in bytes, as determined by kwsexec.  Use this to detect when
	the nominal match found by kwsexec must be skipped because it is for
	an incomplete multi-byte character that is a prefix of a character
	in the input.
	* src/dfasearch.c (EGexecute): Update caller.
	* src/kwsearch.c (Fexecute): Likewise.
	* src/search.h: Update prototype.
	* NEWS (Bug fixes): Mention it.
	Report and analysis by Norihiro Tanaka.

2010-03-28  Norihiro Tanaka  <>

	tests: add tests for the fgrep-infloop bug
	* tests/init.cfg (require_timeout_): New function.
	* tests/fgrep-infloop: New file.  Test for the above fix.
	* tests/ (TESTS): Add it.

2010-03-28  Jim Meyering  <>

	grep -F: avoid infinite loop when searching for incomplete MB character
	Searching for an incomplete non-prefix of a multi-byte character
	should find no match.

	  Just as these print nothing,
	    printf '\357\274\241\357\274\241\n' \
	      | perl -CIO -ne '/\241\357/ and print'
	    printf '\357\274\241\n' | perl -CIO -ne '/\274\241/ and print'
	    printf '\357\274\241\n' | perl -CIO -ne '/\241/ and print'
	    printf '\357\274\241\n' | perl -CIO -ne '/\274/ and print'

	  These should also print nothing, but with grep-2.6 and grep-2.6.1,
	  they would infloop:
	    printf '\357\274\241\n' | LC_ALL=en_US.UTF-8 src/grep -F $'\241'
	    printf '\357\274\241\n' | LC_ALL=en_US.UTF-8 src/grep -F $'\274'
	    printf '\357\274\241\n' | LC_ALL=en_US.UTF-8 src/grep -F $'\274\241'

	* src/kwsearch.c (Fexecute): Don't infloop when searching for
	an incomplete non-prefix part of a multi-byte character.
	* NEWS (Bug fixes): Mention it.
	Reported and diagnosed by Norihiro Tanaka.

2010-03-28  Jim Meyering  <>

	tests: rename: -> fmbtest
	* tests/ Rename to ...
	* tests/fmbtest: ...this, dropping the .sh suffix.
	* tests/ (TESTS): Reflect renaming.

	tests: convert to use
	* tests/ Use and adapt accordingly:
	Use "grep", not ${GREP}.  Use Exit, not exit.

	tests: also exercise the --include + glob path
	* tests/include-exclude: Exercise Javier's fix.

2010-03-28  Javier Villavicencio  <>

	grep -r: fix --include with globs, too
	The previous fix addressed only the non-glob case.
	* src/main.c (main): Use add_exclude's EXCLUDE_WILDCARDS option,
	to enable the use of fnmatch with --include=GLOB.
	gnulib: Update to latest, for the fixed exclude.c.

2010-03-28  Jim Meyering  <>

	grep -r: fix --include with non-globs
	* lib/savedir.c (savedir): Fix logic error.  Introduced by commit
	bf3bd92c, "build: adapt to the newer exclude API we now get from gnulib"
	* tests/include-exclude: Test for this bug by exercising --include, too.
	* NEWS (Bug fixes): Mention it.
	Reported by Philipp Kohlbecher in

2010-03-27  Jim Meyering  <>

	kwset: correct comments; require non-NULL kwsmatch argument
	* src/kwset.c (kwsexec): Correct comments.  This function has been
	returning an offset, not a pointer, for 9 years.
	Do not test for kwsmatch == NULL.  All callers pass non-NULL.
	(cwexec): Likewise.
	* src/kwset.h (kwsexec): Mark the 4th parameter, kwsmatch, as non-NULL.
	Include "arg-nonnull.h".

	build: add -I$(top_builddir)/lib so we also find generated .h files
	* src/ (AM_CPPFLAGS): Rename from INCLUDES to avoid
	warning from automake -Wall.
	Add -I$(top_builddir)/lib, so we find generated .h files like
	getopt.h in a non-srcdir build.

	build: remove superfluous LOCALEDIR definition
	* src/ (INCLUDES): Remove unnecessary definition of
	LOCALEDIR here.  Now, it's defined via gnulib's configmake.h.
	* src/system.h: Include "configmake.h" for its LOCALEDIR definition.

	grep: don't segfault upon use of --include or --exclude* options
	* lib/savedir.c (isdir1): Fix fatal typo: deref "dir" argument,
	not the global (initially-NULL) "path".  Reported by Standish Parsley.
	* tests/include-exclude: New file.
	* tests/ (TESTS): Add it.
	* NEWS (Bug fixes): Mention it.

2010-03-26  Jim Meyering  <>

	tests: rename: -> foad1
	* tests/ Rename to ...
	* tests/foad1: ...this, dropping the .sh suffix.
	* tests/ (TESTS): Reflect renaming.

	tests: convert to use
	This fixes a spurious test failure when "make check" is run with
	certain envvars set, e.g., "make check GREP_COLOR=always"
	* tests/ Use and adapt accordingly:
	Use "grep", not ${GREP}.  Test VERBOSE against "yes", not "1",
	to be consistent with
	Use Exit, not exit.
	Reported by Nelson H. F. Beebe.

	tests: insulate tests from envvar settings
	* tests/init.cfg (vars_): Unset each envvar that can affect how
	grep works.  This protects only those tests that have been
	converted to use

2010-03-25  Eric Blake  <>

	maint: ignore 'make dist pdf' droppings
	* .gitignore: Add more exemptions.

2010-03-25  Jim Meyering  <>

	tests: avoid spurious test failure due to lack of a French UTF8 locale
	* tests/init.cfg: New file.  If either $LOCALE_FR or $LOCALE_FR_UTF8
	is set to "none", reset it to the empty string.
	Reported by Mike Frysinger and Sven Joachim.
	* tests/ (EXTRA_DIST): Add init.cfg.

	build: do not use pkg-config to test for PCRE support
	Do not modify CPPFLAGS; that belongs to those who invoke make.
	Instead, use autoconf's AC_CHECK_HEADERS and AC_SEARCH_LIBS via the
	new macro, gl_FUNC_PCRE, defined in...
	* m4/pcre.m4 (gl_FUNC_PCRE): New macro, to handle pcre-related
	configure-time tests.
	* src/ (grep_LDADD): Use LIB_PCRE, not PCRE_LIBS.
	* src/pcresearch.c: Test HAVE_LIBPCRE via "#if", not "#ifdef".
	All other cpp tests of this symbol used "#if".
	Prompted by a suggestion from Bruno Haible.
	* NEWS (Build-related): Mention this.

	doc: correct and amend NEWS entries for 2.6.1
	* NEWS (Bug fixes): Correct character ranges bug description.
	Add an example from Dmitry V. Levin.
	Add that the word-with-backref bug was introduced in 2.5.1.
	* (old_NEWS_hash): Update to match.

	post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.6.1
	* NEWS: Record release date.

2010-03-25  Tony Abou-Assaleh  <>

	tests: use awk's -v option more portably
	* tests/spencer1-locale: Add a space between awk's "-v" option and
	the following VAR=value string, to avoid test failure on Mac OS X.

2010-03-25  Norihirio Tanaka  <>

	dfa/grep: fix compilation with MBS_SUPPORT
	* src/dfa.c (cur_mb_len): Initialize to 1 and always make it available.
	(setbit_case_fold): Do not use wint_t in prototype if !MBS_SUPPORT.
	(parse_bracket_exp): Fix compilation with !MBS_SUPPORT.
	* src/kwsearch.c (kwsinit): Do not use mbtolower and MB_CUR_MAX
	* src/searchutils.c (kwsinit): Do not refer to MB_CUR_MAX if !MBS_SUPPORT.

	* tests/char-class-multibyte: Skip if UTF-8 matching does not work.
	* tests/ Likewise.

2010-03-25  Jim Meyering  <>

	build: avoid warnings about unnecessary use of "return"
	* src/grep.c (Gcompile, Ecompile, Acompile): Do not "return X"
	from a function returning void, not even when X itself is a
	function returning void.  This avoids warnings from Sun Studio 11
	reported by Dagobert Michelsen.
	* src/egrep.c (Ecompile): Likewise.

2010-03-25  Norihirio Tanaka  <>

	grep: fix printing when -w is used and regex is needed for matching
	* NEWS: Document bugfix.
	* src/dfasearch.c (EGexecute): After assess_pattern_match len, is either
	invalid or end-beg; jump to success.
	* tests/ (TESTS): Add new test.
	* tests/backref-word: New.

2010-03-25  Paolo Bonzini  <>

	dfa: fix single byte character ranges
	* src/dfa.c (in_coll_range): Fix ordering for second strcoll.  Reported
	by Dmitry V. Levin.
	* tests/spencer1-locale.awk: Also test single-byte character sets.
	* NEWS: Add a note about this bugfix.
	* THANKS: Add Dmitry.

2010-03-25  Norihirio Tanaka  <>

	grep: reset state after truncated or invalid multibyte sequences
	* src/searchutils.c (is_mb_middle): When treating an invalid sequence
	or a truncated multibyte character as a single byte character, reset

	grep: do lowercase conversion in print_line_middle only for single-byte case
	* src/main.c (print_line_middle): Restrict match_icase code
	to MB_CUR_MAX == 1.  Adjust comments.

2010-03-25  Jim Meyering  <>

	tests: provide framework_failure_ function
	The shell function "framework_failure" was called in the unusual
	event that some fundamental test set-up operation would fail.
	However it was not defined.  Define it, but with a trailing underscore
	to impinge less on the test writer's name space.  Adjust all uses.
	* tests/ (framework_failure_): New function.
	* tests/case-fold-backref: s/framework_failure/framework_failure_/
	* tests/case-fold-char-class: Likewise.
	* tests/case-fold-char-range: Likewise.
	* tests/case-fold-char-type: Likewise.
	* tests/char-class-multibyte: Likewise.
	* tests/dfaexec-multibyte: Likewise.
	* tests/max-count-vs-context: Likewise.
	* tests/word-multi-file: Likewise.

2010-03-24  Jim Meyering  <>

	doc: tweak THANKS
	* THANKS: Update Arnold's name and address, per request.

	portability: use gnulib's lseek wrapper
	* bootstrap.conf (gnulib_modules): Use gnulib's lseek wrapper,
	for improved portability.  lseek does not fail with ESPIPE on
	pipes on some systems.

	build: avoid link failure on Solaris 8
	* bootstrap.conf (gnulib_modules): Add wctob.
	* NEWS (Portability): Mention this.
	Reported by Dagobert Michelsen in <>.

2010-03-24  Petr Písař  <>

	doc: translate new --help message
	* src/main.c: Translate "after_options".

2010-03-24  Jim Meyering  <>

	doc: NEWS make it clear that the bug was introduced in 2.6
	* NEWS: Clarify.

2010-03-24  Paolo Bonzini  <>

	tests: fix char-class-multibyte
	* tests/char-class-multibyte: Make it pass.

2010-03-23  Jim Meyering  <>

	build: avoid compilation failure when MBS_SUPPORT not defined
	* src/dfa.c (setbit_case_fold) [!MBS_SUPPORT]: Fix curly brace mismatch.

2010-03-23  Paolo Bonzini  <>

	dfa: fix sigsegv on multibyte character classes
	Reported by Jaroslav Škarvada <>.  This is
	unfortunate.  grep needs an automatic testcase generator.

	* NEWS: Document bug.
	* THANKS: Mention reporter.
	* src/dfa.c (set_bit_casefold): Change type of first argument for
	(parse_bracket_exp): Fix call.
	* tests/ Add new testcase.
	* tests/char-class-multibyte: New testcase.

2010-03-23  Jim Meyering  <>

	post-release administrivia
	* NEWS: Add header line for next release.
	* .prev-version: Record previous version.
	* (old_NEWS_hash): Auto-update.

	version 2.6
	* NEWS: Record release date.

	build: avoid warnings: tell gcc and clang that dfaerror never returns
	* src/dfa.h (__attribute__): Define.
	(dfaerror): Declare with the "noreturn" attribute.
	* src/dfasearch.c (dfaerror): Add an unreachable use of abort.

2010-03-22  Eric Blake  <>

	build: fix cygwin build
	Portions of gnulib depend on -lintl, and cygwin does not allow
	lazy linking.

	* src/ (LDADD): Include libraries in correct order.

2010-03-22  Paolo Bonzini  <>

	grep: remove --mmap
	mmap is a bad idea for sequentially accessed file because it will cause
	a page fault for every read page.  Just consider it a failed experiment,
	and ignore --mmap while accepting it for backwards compatibility.

	* (AC_FUNC_MMAP): Remove.
	* doc/grep.texi (Other options): Say --mmap is ignored.
	* src/grep.c (mmap_option): Remove.
	(long_options): Do not reference it.
	(bufmapped, initial_bufoffset): Remove.
	(reset, fillbuf): Remove HAVE_MMAP code.
	(grepfile): Remove bufmapped reference.
	(usage): Say --mmap is ignored.

2010-03-22  Paolo Bonzini  <>

	grep: rename files for intuitiveness
	* (libgrep_a_SOURCES, grep_SOURCES, egrep_SOURCES,
	fgrep_SOURCES): Adjust.
	* grep.c: Rename to main.c.
	* esearch.c: Rename to egrep.c.
	* fsearch.c: Rename to fgrep.c.
	* gsearch.c: Rename to grep.c.

	* NEWS: Document slight semantic change.
	* TODO: #ifdefs are gone.
	* po/ Update.
	* src/ (grep_SOURCES, egrep_SOURCES, fgrep_SOURCES): Remove
	(noinst_LIBRARIES): Change libsearch.a to libgrep.a.
	(libsearch_a_SOURCES): Rename to libgrep_a_SOURCES, add grep.c
	(LDADD): Change libsearch.a to libgrep.a.
	* src/esearch.c: Add before_options and after_options.
	* src/fsearch.c: Likewise.
	* src/gsearch.c: Likewise.
	* src/grep.c (short_options, long_options): Remove GREP_PROGRAM
	(usage): Use before_options and after_options, look at matchers.
	(setmatcher): Merge with install_matcher.
	(main): Call setmatcher (NULL) instead of install_matcher.
	* src/grep.h (GREP_PROGRAM): Remove.
	(before_options, after_options): Add.

	thank Eric Blake
	* THANKS: Add Eric Blake, who reported the warning fixed by 774d0ee.

	grep: libify *search.c
	* src/ (libsearch_a_SOURCES): Add dfasearch.c, kwsearch.c,
	* src/esearch.c, src/fsearch.c, * src/gsearch.c: Only include search.h.
	* src/dfasearch.c (GEAcompile, EGexecute): Export.
	* src/kwsearch.c (Fcompile, Fexecute): Export.
	* src/pcresearch.c (Pcompile, Pexecute): Export.
	* src/search.h: Add new exported functions.

	grep: prepare for libification of *search.c
	* src/dfasearch.c (Ecompile): Remove.
	* src/esearch.c: Place it here...
	* src/gsearch.c: ... and here.

	grep: split search.c
	* po/ Update.
	* src/ (grep_SOURCES, egrep_SOURCES, fgrep_SOURCES): Move
	kwset.c and dfa.c to libsearch.a.  Add searchutils.c there too.
	* src/search.h, src/dfasearch.c, src/pcresearch.c, src/kwsearch.c,
	src/searchutils.c: New files, split out of src/search.c.
	* src/esearch.c, src/fsearch.c: Include the new files instead of search.c.
	* src/gsearch.c: Likewise, plus move Gcompile/Acompile here.

	grep: remove one #ifdef
	* search.c (GEAcompile) [EGREP_PROGRAM]: Use common code.  Inline IF_BK.

2010-03-22  Paolo Bonzini  <>

	grep: eliminate {COMPILE,EXECUTE}_{RET,ARGS,FCT}
	Modern compilers warn about type mismatches.

	* src/grep.c (do_execute): Write full declaration.
	(compile_fp_t, execute_fp_t): Write full declaration.
	* src/search.c (GEAcompile, Gcompile, Acompile, Ecompile, EGexecute,
	Fcompile, Fexecute, Pcompile, Pexecute): Write full declaration.

2010-03-22  Paolo Bonzini  <>

	grep: make egrep/fgrep use struct matcher
	* (grep_SOURCES): Add gsearch.c.
	(EXTRA_DIST): Add search.c.
	* esearch.c (matchers): New.
	* fsearch.c (matchers): New.
	* gsearch.c: New.
	* search.c (matchers): Remove.
	* grep.c: Always compile most !GREP_PROGRAM sections.
	(main): Use first matcher if none is explicitly provided.  Remove
	"default" matcher.
	* grep.h (struct matcher): Adjust comments.

	grep: change struct matcher termination
	* src/grep.c (setmatcher): Look for NULL matchers[i].name.
	* src/grep.h (struct matcher): Change name to pointer.  Adjust comments.
	* src/search.c (matchers): Terminate with three NULLs.

	grep: remove one #ifdef
	* search.c (Ecompile): Always go through GEAcompile to use same code path
	for both grep and egrep.

	grep: remove getpagesize.h
	* src/getpagesize.h: Remove.
	* src/ (noinst_HEADERS): Remove getpagesize.h.

2010-03-21  Jim Meyering  <>

	build: use the fcntl-h module, not "fcntl"
	* bootstrap.conf (gnulib_modules): We might need fcntl.h somewhere,
	but don't use the fcntl function.  Reported by Bruno Haible.

	build: avoid link failure on systems using gnulib's fcntl but not open
	* bootstrap.conf (gnulib_modules): Using gnulib's fcntl module
	and including <fcntl.h>, but not also using gnulib's "open" module
	would result in link failure due to references to rpl_open
	on systems requiring the replacement (e.g., Cygwin and Darwin).

	build: avoid compilation failure on systems using rpl_open
	This new build failure has arisen as a result of using gnulib's
	"fcntl" module.  Now that an inadequate "open" syscall is replace
	by gnulib's wrapper, it is essential to include <fcntl.h>.
	* src/grep.c: Include <fcntl.h>.
	This is required, for grepfile's use of open, at least on
	Cygwin and Darwin.

	maint: use gnulib's fcntl module, just in case
	* bootstrap.conf (gnulib_modules): Add fcntl.
	Grep uses at least O_BINARY, which may be defined therein.

	maint: remove TYPE_* definitions from src/system.h
	definitions.  They are provided by intprops.h.
	* src/grep.c: Include "intprops.h"
	* bootstrap.conf (gnulib_modules): Add intprops.

	maint: alphabetize #include directives
	* src/grep.c: Alphabetize #include directives.

2010-03-20  Jim Meyering  <>

	build: stop using gnulib's memmove module
	* bootstrap.conf (gnulib_modules): Remove obsolete module: memmove

	build: reinstate gnulib's fcntl-h-tests
	* bootstrap.conf (gnulib_tool_option_extras): Do not avoid
	the fcntl-h-tests.  I cannot reproduce the failure.

2010-03-20  Eric Blake  <>

	build: allow compilation on cygwin
	Gnulib is incompatible with -Wunused-macros.  Addtionally,
	cygwin 1.7.1 coupled with --enable-gcc-warnings tripped on:

	grep.c: In function 'print_line_middle':
	grep.c:805: error: array subscript has type 'char' [-Wchar-subscripts]
	grep.c: In function 'main':
	grep.c:1833: error: 'optarg' redeclared without dllimport attribute: previous dllimport ignored [-Wattributes]
	grep.c:1834: error: 'optind' redeclared without dllimport attribute after being referenced with dll linkage

	* (GNULIB_WARN_FLAGS): Disable -Wunused-macros.
	* src/grep.c (print_line_middle): Use correct type to tolower.
	(main): Drop useless redeclarations.
	* .gitignore: Ignore more built files.

2010-03-20  Jim Meyering  <>

	tests: ensure that all programs handle [b-a] consistently
	* tests/reversed-range-endpoints: New test.
	* tests/ (TESTS): Add it.

2010-03-20  Jim Meyering  <>

	build: update gnulib submodule to latest
	This pulls in the latest regex module from gnulib, including a fix
	to make it honor the RE_NO_EMPTY_RANGES syntax bit.

	tests: temporarily disable irrelevant-to-grep failing C++ fcntl-h-tests
	* bootstrap.conf (gnulib_tool_option_extras): Temporarily add
	--avoid=fcntl-h-tests, until the C++ part of that test is fixed.

2010-03-20  Jim Meyering  <>

	reject reversed-endpoint ranges, with all regex variants
	* src/search.c: Add RE_NO_EMPTY_RANGES to the syntax bits
	in three places, so that all of grep, egrep, and grep -E reject
	a range with reversed endpoints like '[b-a]'.  This is required,
	when using the latest version of gnulib's regex module, since it
	now honors the RE_NO_EMPTY_RANGES flag, rather than acting as if
	it were always set.
	Based on a change by Matthew Burgess.

2010-03-19  Jim Meyering  <>

	maint: correct macro parameter parentheses
	* src/dfa.c (FETCH_WC, FETCH): Parenthesize macro parameters.

2010-03-19  Paolo Bonzini  <>

	tests: change help-version to per-program functions
	* help-version: Change each *_args variable to a *_setup function.

	dfa: fix wchar_t/wint_t type mismatch
	* src/dfa.c (FETCH_WC): Pass a local wchar_t variable to mbrtowc.
	(FETCH): Rename temporary second argument to FETCH_WC.
	(parse_bracket_exp): Always use FETCH_WC.

2010-03-19  Jim Meyering  <>

	doc: add README-prereq, referenced from README-hacking
	* README-prereq: New file.  Cloned from coreutils, s/coreutils/grep/
	Reported by Tony Abou-Assaleh.

2010-03-19  Arnold Robbins  <>

	maint: sync dfa comments from gawk
	* src/dfa.h (struct dfa) [newlines]: Amend comment.
	* src/dfa.c: Update copyright year list to include gawk's.

2010-03-17  Jim Meyering  <>

	maint: remove obsolete "cvs-clean" make target
	* (cvs-clean): Remove obsolete target.

2010-03-17  Paolo Bonzini  <>

	dfa: initialize struct mbcset using memset
	* src/dfa.c (parse_bracket_exp): Use memset to initialize workmbc.

	dfa: spell out "unsigned int"
	* dfa.c (setbit, tstbit, clrbit, setbit_case_fold, lex, dfaoptimize,
	free_mbdata): Put "int" after unsigned.
	* dfa.h (struct position, struct dfa): Likewise.

2010-03-17  Paolo Bonzini  <>

	dfa: optimize simple character sets under UTF-8 charsets
	Only use a bitset when possible without involving MBCSET.  Testcase:
	   yes 'the quick brown fox jumps over the lazy dog' | sed 100000q | \

	Before: 51ms (best of three runs); after: 16ms(best of three runs).

	* src/dfa.c (parse_bracket_exp): For simple bracket expressions
	under UTF-8, use a CSET.

2010-03-17  Paolo Bonzini  <>

	dfa: speed up handling of brackets
	This patch has two sides.  One is to fold the parsing of brackets in the
	single- and multi-byte cases.  The second is to leverage this change,
	and use a bitset to test for single-byte characters in the charset.
	Splitting the two would be very hard.

	   yes 'the quick brown fox jumps over the lazy dog' | sed 100000q | \

	Before: 59ms (best of three runs); after: 51ms (best of three runs).
	Nice, but mostly providing infrastructure for the next patch.

	* src/dfa.c (setbit_case_fold): Try applying towlower/towupper.
	(looking_at): Remove.
	(FETCH_WC): New.
	(fetch_wc): Merge into FETCH_WC [MBS_SUPPORT].
	(prednames, find_pred, is_blank and other predicates): Move above,
	remove K&R syntax support.
	(parse_bracket_exp): New name of parse_bracket_exp_mb, rewritten to
	include single-byte character set parsing of brackets.
	(lex): Adjust for fetch_wc->FETCH_WC change, remove single-byte
	character set parsing of brackets.
	(match_mb_charset): Test against work_mbc->cset.
	* src/dfa.h (struct mb_char_classes): Add cset.

2010-03-17  Paolo Bonzini  <>

	syntax-check: remove space-tab exception
	* .x-sc_space_tab: Remove.
	* src/dfa.c: Fix space-tab occurrence.

	THANKS: fix Jim Meyering's email address
	* THANKS: Jim is now with Red Hat.

	dfa: add missing function
	* src/dfa.c (using_utf8): New.
	(addtok_wc, free_mbdata, dfaoptimize) [!MBS_SUPPORT]: Do not define.
	(dfacomp) [!MBS_SUPPORT]: Do not call dfaoptimize.

	tests: fix typo
	* fedora: Fix typo.

	tests: use Exit
	* euc-mb: exit with "Exit 0".

	grep: remove more register keywords
	* dosbuf.c: Remove register keywords.
	* grep.c: Remove register keywords.
	* kwset.c: Remove register keywords.
	* search.c: Remove register keywords.

2010-03-17  Paolo Bonzini  <>

	dfa: run simple UTF-8 regexps as a single-byte character set
	This provides a speedup whenever fgrep is "almost" sufficient but
	not quite (e.g. grep ^abc).  This affects test cases such as, which are already worked around
	by the line-by-line matching patch c32c04; without that patch the
	speedup can reach 1000x even on non-contrived testcases.

	* src/dfa.c (dfaoptimize): New.
	(dfacomp): Call it.

2010-03-17  Paolo Bonzini  <>

	tests: fix syntax-check failures
	* tests/case-fold-backref: Use "foo" instead of "the".
	* tests/dfaexec-multibyte: Remove trailing blanks.

2010-03-17  Paolo Bonzini  <>

	grep: remove check_multibyte_string, fix non-UTF8 missed match
	Avoid computing ahead something that can be computed lazily as efficiently
	(or more efficiently in the case of UTF-8, though this is left as TODO).
	At the same time, "soften" the rejection condition for matching in the
	middle of a multibyte sequence to fix bug 23814.

	Multibyte "grep -i" would still be very slow if it wasn't for the workaround
	patch c32c042 (grep: match multibyte charsets line-by-line when using -i,

	* NEWS: Document bugfix.
	* src/search.c (check_multibyte_string): Rewrite as...
	(is_mb_middle): ... this.
	(EGexecute, Fexecute): Adjust.
	* tests/ (TESTS): Add euc-mb.
	* tests/euc-mb: New testcase.

2010-03-17  Paolo Bonzini  <>

	dfa: cache MB_CUR_MAX for dfaexec
	* src/dfa.c (state_index, dfaexec): Use d->mb_cur_max.
	(dfainit): Initialize it.
	(free_mbdata): New, extracted out of dfafree.
	(dfafree): Use it.

	dfa: improve documentation of struct dfa
	* src/dfa.h (struct dfa): Reword some comments.

	tests: factor name of output files into a variable
	* tests/case-fold-backref, tests/case-fold-char-class,
	tests/case-fold-char-range, tests/case-fold-char-type,
	tests/dfaexec-multibyte: Use a variable for the output filename,
	as it is common to the grep and compare invocations.

	tests: use different output files to simplify reading failed .log files
	* tests/case-fold-backref, tests/case-fold-char-class,
	tests/case-fold-char-range, tests/case-fold-char-type: Use a different
	name for each output file from grep.
	* tests/dfaexec-multibyte: Likewise, and merge some grep invocations.

	tests: add another grep -i testcase, from bug 16179
	* tests/case-fold-backref: New.
	* tests/ (TESTS): Add it.

2010-03-16  Paolo Bonzini  <>

	dfa: rewrite handling of multibyte case_fold lexing
	Let dfacomp do the folding to lowercase of multibyte input strings,
	and remove it from grep.c.  Input strings to kwset.c are still folded
	outside kwset.c, so we still need to do mbtolower in search.c.

	* NEWS: Document bugfixes.
	* .x-sc_cast_of_argument_to_free: Remove.
	* src/dfa.c (wctok, addtok_wc): New.
	(cur_mb_index, update_mb_len_index): Remove.
	(FETCH): Do not call it.
	(parse_bracket_exp_mb) [GREP]: Disable case-folding of ranges and
	(addtok): Extract part to...
	(addtok_mb): ... this new function.
	(lex): Call fetch_wc in the main loop for MB_CUR_MAX > 1.  Return WCHAR
	for normal characters if MB_CUR_MAX > 1.
	(atom): Handle WCHAR instead of treating multibyte characters specially.
	Do case folding of multibyte characters here.
	(dfacomp): Remove case_fold special casing.
	* src/dfa.h (WCHAR): New.
	* src/grep.c (mb_icase_keys): Remove.
	(main): Do not call it.
	* src/search.c (kwsinit): Init transition table only for MB_CUR_MAX == 1.
	(mbtolower): New.
	(kwsincr_case): New.
	(kwsmusts): Call it instead of kwsincr.
	(check_multibyte_string): Remove.
	(check_multibyte_string_no_icase): Rename to check_multibyte_string.
	(GEAcompile, EGexecute, Fcompile): Use mbtolower instead of the old
	* tests/ (TESTS): Add case-fold-backslash-w.
	* tests/ Enable fixed tests.
	* tests/case-fold-backslash-w: New.

2010-03-16  Paolo Bonzini  <>

	grep: match multibyte charsets line-by-line when using -i
	The turtle combination -i + MB_CUR_MAX>1 requires case conversion ahead
	of time.  Avoid doing this repeatedly when many matches succeed.  Together
	with the previous changes, this fixes

	* NEWS: Document new speedup.
	* src/grep.c (do_execute): New.
	(grepbuf): Use it.

2010-03-15  Paolo Bonzini  <>

	dfa: fix handling of ranges in multibyte character sets
	* src/dfa.c (parse_bracket_exp_mb): Add separate ranges for
	lowercase and uppercase endpoints if folding case.
	* tests/ (TESTS): Add case-fold-char-range.
	* tests/case-fold-char-range: New.

	tests: add more UTF-8 test cases
	* tests/ (TESTS): Add spencer1-locale.
	(EXTRA_DIST): Add spencer1-locale.awk.
	* tests/spencer1-locale.awk: New.
	* tests/spencer1-locale: New.

2010-03-15  Jim Meyering  <>

	tests: complete the renaming -> fedora
	* tests/ (TESTS): Rename -> fedora here, too.

2010-03-15  Jim Meyering  <>

	* tests/ Rename to...
	* tests/fedora: ...this, to reflect new convention:
	Use the lack of a suffix to indicate we've converted to the new test framework.

	tests: adjust to handle traps more portably

2010-03-15  Jim Meyering  <>

	tests: adjust to handle traps more portably
	* tests/ Use "Exit", not "exit".

	tests: for each test, set an envvar to its name
	each test.  This is used to help make the output of hundreds of
	independent, often-parallel valgrind runs more manageable.

2010-03-14  Jim Meyering  <>

	tests: clean up
	* tests/ Use "grep", not ${GREP}.
	Use timeout 10, not sleep 1 (three times).
	The latter would always sleep for 3 seconds, and the test would
	fail with a false positive on a slow system or with a heavily
	instrumented (valgrind) executable.

2010-03-12  Jim Meyering  <>

	build: avoid build failure with --enable-gcc-warnings
	* src/dfa.c: Don't include <assert.h>, now that it is not used.
	[DEBUG]: Remove #ifdef block.

2010-03-12  Paolo Bonzini  <>

	syntax-check: enable space-tab
	* (local-checks-to-skip): Enable space-tab.
	* .x-sc_space_tab: Add exceptions.
	* tests/ Fix occurrence.

	syntax-check: enable m4-quote-check
	* (local-checks-to-skip): Enable m4-quote-check.
	* Fix occurrence.

	syntax-check: enable makefile-TAB-only-indentation
	* (local-checks-to-skip): Enable makefile-TAB-only-indentation.
	* Fix only occurrence.

	grep: fix error-message-uppercase
	* (local-checks-to-skip): Enable error-message-uppercase.
	* src/dfa.c (parse_bracket_exp_mb, lex, dfaparse): Fix occurrences.
	* src/search.c (Pcompile, Pexecute): Fix occurrences.

	dfa, grep: cleanup if-before-free and cast-of-argument-to-free
	* .x-sc_avoid_if_before_free: Remove.
	* .x-sc_cast_of_alloca_return_value: Remove.
	* .x-sc_cast_of_x_alloc_return_value: Remove.
	* .x-sc_cast_of_argument_to_free: Temporarily add src/search.c.
	* (local-checks-to-skip): Remove sc_cast_of_argument_to_free.
	* src/dfa.c (ifree): Remove.
	(dfamust, build_state, transit_state, dfafree): Do not do if-before-free,
	do not cast free argument to ptr_t or char *.
	(freelist): Call free instead of ifree.
	* src/dfa.h (ptr_t): Remove.

2010-03-12  Paolo Bonzini  <>

	dfa: remove CRANGE dead code
	The only use of CRANGE was removed by commit 193830d.  In theory it is
	more correct to do what CRANGE did, but in practice it seems like it did
	not work.

	* src/dfa.h (token): Remove CRANGE.
	* src/dfa.c (atom): Do not handle CRANGE.
	(prtok): Likewise.

2010-03-12  Paolo Bonzini  <>

	dfa: get rid of x*alloc
	* src/dfa.c: Include xalloc.h.
	(xmalloc, xrealloc, xcalloc): Remove.

	grep: cleanup one const cast
	* src/search.c (GEAcompile): Do not reuse motif when operating on the
	(const) pattern, so we can make it non-const.  Remove cast from free.

	kwset/system: remove ptr_t
	* src/kwset.h: Declare kwset using an incomplete struct type.
	* src/system.h (ptr_t): Remove.

2010-03-12  Jim Meyering  <>

	tests: add test cases for dfaexec bug
	* tests/dfaexec-multibyte: New test.
	* tests/ (TESTS): Add it.
	Reported by Paolo Bonzini in
	and .

2010-03-12  Jim Meyering  <>

	dfa: manually merge gawk's dfaexec
	* src/dfa.c (dfaexec): Adjust API: return pointer, not offset, and
	take an "end" pointer parameter, rather than integral "size".
	Adjust comment accordingly.
	(build_state): Maintain d->newlines.
	(copytoks): Update multibyte_prop indices.
	Return NULL, rather than (size_t) -1.
	(realloc_trans_if_necessary): Realloc d->newlines.
	* src/dfa.h (struct dfa): New member, "newlines".
	(struct dfa) [GAWK]: New member, "broken".
	(dfaexec): Update prototype and copy the new comment from dfa.c.

	dfa: make search.c use the new dfaexec API

	* src/search.c: Adjust to new dfaexec API.
	Now, dfaexec returns a pointer, not an integer,
	and the third parameter is END, not buffer size.
	* src/dfa.c (dfaexec): Rewrite the function's comment.
	Don't just clobber *END.  While doing that happens to be
	fine for gawk's usage, in grep, *END usually points to the
	first byte of the next buffer.  Save the initial value,
	and restore it just before returning.
	* src/dfa.h (dfaexec): Update comment; include parameter names.

2010-03-12  Jim Meyering  <>

	dfa: appease static analyzers
	* src/dfa.c (transit_state_singlebyte): Call abort rather
	than returning in a "can't happen" scenario.
	This stops clang from emitting a false-positive report (I think it
	was used-uninitialized) about a caller.

2010-03-11  Jim Meyering  <>

	dfa: do not accept [[:UPPER:]] or [[:LOWER:]] internally
	* src/dfa.c (parse_bracket_exp_mb): Those class names are not
	valid, and rejected elsewhere, so there is no point in allowing
	upper or mixed-case versions here.

2010-03-11  Jim Meyering  <>

	maint: remove a trailing space
	* src/search.c (EXECUTE_FCT): Remove trailing space.

	maint: remove all uses of PARAMS
	Remove most with this:
	git grep -lw PARAMS |xargs perl -pi -e 's/\bPARAMS *\((.*)\);/$1;/'
	Remove the remainder manually.

2010-03-11  Jim Meyering  <>

	maint: remove all uses of PARAMS
	* lib/savedir.h (PARAMS): Remove definitions manually.
	Remove the remaining ones via this command:
	  git grep -l define.PARAMS |xargs perl -ni -e '/define PARAMS/ or print'
	* src/dfa.h (PARAMS): Remove definitions.
	* src/system.h (PARAMS): Likewise.
	Remove most uses with this:
	  git grep -lw PARAMS |xargs perl -pi -e 's/\bPARAMS *\((.*)\);/$1;/'
	Remove the remainder manually.

	maint: remove now-useless prototypes
	* src/dfa.c: Remove the prototype of each static, non-recursive
	function whose definition precedes first use.

	grep: plug an inconsequential leak
	* src/grep.c (main): Plug a leak: free "keys".

	grep: avoid useless allocations for empty GREP_OPTIONS
	* src/grep.c (prepend_default_options): Ignore GREP_OPTIONS
	when it's empty, not just when it's undefined.
	There are still relatively harmless leaks when GREP_OPTIONS
	is set and non-empty.  We'll address those, eventually.

2010-03-09  Jim Meyering  <>

	build: record build-from-clone tool requirements
	* bootstrap.conf (buildreq): This makes bootstrap fail with
	a clear explanation of the problem.  Otherwise, you'd get into
	the build process and fail with something far more cryptic.

	dfa: remove a trailing blank
	* src/dfa.c (dfaexec): No trailing blanks allowed.

	dfa: sync a tiny change from gawk
	* src/dfa.c (state_index) [MBS_SUPPORT]: Initialize .mpbs.nelem member
	unconditionally.  Also initialize .mbps.elems.

	dfa: avoid a leak (work_mbc->chars)
	* src/dfa.c (parse_bracket_exp_mb): Remove useless (and leaked MALLOC).

	doc+bootstrap: document build-from-git-clone process
	* bootstrap: Update from coreutils/gnulib.
	* README-hacking: New file, nearly identical to the one in coreutils.

2010-03-08  Paolo Bonzini  <>

	more work on TODO
	* TODO: More work on the first section.  Use clearer section headers.

2010-03-08  Reuben Thomas  <>

	bring TODO up-to-date
	* TODO: merge with TODO section of
	and remove done items. Some small bits of tidying also.

2010-03-07  Paolo Bonzini  <>

	simplify parsing of [a-z]
	* src/dfa.c (in_coll_range): New.
	(lex): Use it instead of regcomp/regexec.

	Small refactoring in src/dfa.c
	* src/dfa.c (parse_bracket_exp_mb): Return MBCSET.
	(lex): Assign return value of parse_bracket_exp_mb to lasttok, return it.

	use do...while(0) idiom
	* dfa.c (FETCH): Wrap with do...while(0).

2010-03-06  Paolo Bonzini  <>

	extract common code from if/else
	* dfa.c (dfaexec): Simplify logic for MB_CUR_MAX > 1 case.

	remove register variable hacks
	* dfa.c (dfaexec): We can extract the address of a variable without fearing
	performance problems, modern compilers know better.

	remove register keywords
	* dfa.c (dfaexec): Modern compilers just ignore it.

	allow grep -Pz
	* NEWS: Document grep -P improvements.
	* src/search.c (Pcompile): Remove restriction on grep -Pz.
	* tests/pcre-z: New.
	* tests/ (TESTS): Add pcre-z.

	fix cross-line matching in PCRE backend
	* search.c (Pexecute): Split the buffer in lines and match each line
	* tests/ Add regression testsuite.

	fix formatting of NEWS
	* NEWS: fix formatting of 2.6 entries.

	fix a bug in handling of -i and character type
	* dfa.c (parse_bracket_exp_mb): Convert [[:lower:]] and [[:upper]] to
	[[:alpha:]] when folding case.
	* tests/case-fold-char-type: New file.  Test for the bug.
	* tests/ (TESTS): Add it.
	* NEWS (Bug fixes): Mention it.

	fix previous test case change
	* tests/case-fold-char-class: Do not reset fail to 0 after first test.

2010-03-06  Mike Frysinger  <>

	grep(1) man page: touchup --label option
	* doc/grep.1 (--label): Don't italicize ending period.  Point to -H

2010-03-06  Paolo Bonzini  <>

	augment case-fold-char-class test case
	* tests/case-fold-char-class: Test matching lowercase against uppercase
	as well as vice versa.

2010-03-05  Reuben Thomas  <>

	doc: improve the discussion of PCRE
	* doc/grep.1: Add a sentence about Perl regular expressions,
	and point to pcresyntax(3) and pcrepattern(3).
	* doc/grep.texi: Likewise.

2010-03-05  Jim Meyering  <>

	maint: dfa-sync: comment and dead-to-grep code: no semantic change
	* src/dfa.c: Sync a comment and some #ifdef GAWK code.

	maint: dfa-sync: don't malloc zero
	* src/dfa.c (dfacomp): Skip case_fold logic when length is zero.
	This probably "no semantic change", but does improve efficiency in
	a degenerate case.

	maint: dfa-sync: use CALLOC rather than equiv. MALLOC+initialize-loop
	* src/dfa.c (dfaanalyze): Sync from gawk.  No semantic change.

	dfa.c: add support for \s and \S
	* src/dfa.c (lex): Sync from gawk's dfa.c.

	maint: dfa-sync: add omitted array initializer
	* src/dfa.c (prednames): Add a "0" to final initializer.
	No semantic change.

	fix a bug in handling of -i and character classes
	* dfa.c (parse_bracket_exp_mb): Sync one part of this function
	from gawk's dfa.c, which was patched by Arnold D. Robbins.
	* tests/case-fold-char-class: New file.  Test for the bug.
	* tests/ (TESTS): Add it.
	definitions into tests.
	* NEWS (Bug fixes): Mention it.

2010-03-05  Paolo Bonzini  <>

	Fedora Grep regression test suite
	* tests/ (TESTS): Add
	(CLEANFILES): Add several new files.
	* tests/ New file, originally by Lubomir Rintel but somewhat
	rewritten to avoid bashisms.

2010-03-05  Paolo Bonzini  <>

	convert AUTHORS file to UTF-8
	* AUTHORS: Convert to UTF-8.

	eliminate invalid "ptr += (ptr2 - ptr1)"
	* lib/savedir.c (savedir): new_name_space and name_space do not point into
	the same object, so computing their difference is invalid.  Similarly,
	summing the difference to namep is invalid because namep and the result
	point into different objects.  Avoid this.

	fix for bug 21276
	* lib/savedir.c (isdir1): Use realloc instead of calloc.  Remove
	dead code.
	(savedir): Do not leak name_space if allocation of new_name_space fails.

2010-03-04  Jim Meyering  <>

	tests: add a test based on an example from Paolo Bonzini
	* tests/word-multi-file: New test.
	* tests/ (TESTS): Add it.

	doc: document release procedure
	* README-release: New file.

	build: update gnulib submodule to latest

2010-02-22  Paolo Bonzini  <>

	add --group-separator=FOO and --no-group-separator
	* src/grep.c (group_separator): New.
	(long_options): Add --group-separator=FOO and --no-group-separator.
	(prtext): Print group_separator instead of SEP_STR_GROUP.  Optionally
	suppress the separator altogether.
	* doc/grep.texi (Context control): Document it.
	* NEWS: Mention it.
	* tests/ Add testcases.

2010-02-21  Jim Meyering  <>

	tests: don't use "echo -n"
	* tests/ Use printf, not echo -n.  The latter is not portable.
	Reported by Daniel Richman.

2010-02-08  Jim Meyering  <>

	remove useless DJGPP-specific code
	* src/grep.c (grepfile): Remove now-useless DJGPP-specific code.
	Now, all S_IS* macros are guaranteed to be defined via gnulib.

2010-02-07  Jim Meyering  <>

	tests: add help-version sanity tests from coreutils
	* tests/help-version: New test, from coreutils.
	* tests/ (TESTS): Add it.
	(TESTS_ENVIRONMENT) [built_programs]: Define it.

	tests: correct TESTS_ENVIRONMENT's PATH setting
	* tests/ (TESTS_ENVIRONMENT): Set PATH to start with
	$(abs_top_builddir)/src, so that we test the programs we've just built.

	grep: use the correct exit status (2) upon write failure, not 1
	* src/grep.c (main): Initialize exit_failure to EXIT_TROUBLE.
	* NEWS (Bug fixes): Mention this fix.

	maint: enable the prohibit_magic_number_exit syntax check
	* (local-checks-to-skip): Remove sc_prohibit_magic_number_exit,
	to enable that check.
	* src/system.h (EXIT_TROUBLE): Define.
	* src/grep.c: Use symbolic names, EXIT_SUCCESS, EXIT_FAILURE, and
	EXIT_TROUBLE, not 0, 1, 2.
	* src/search.c: Likewise.
	* src/vms_fab.c (string): Likewise.

2010-02-04  Jim Meyering  <>

	doc: adjust NEWS item
	* NEWS: Correct a description.

2010-02-03  Jim Meyering  <>

	tests: exercise surprising -m1 vs. --context behavior
	* tests/max-count-vs-context: New test.  Exercise the surprising,
	but documented, behavior reported by Markus Jochim in
	* tests/ (TESTS): Add it.

	tests: use from gnulib
	* tests/ New file, from gnulib.
	* tests/ (EXTRA_DIST): Add it.
	(TESTS_ENVIRONMENT): Add variables and features.
	(VERBOSE): Define.

	maint: remove unused Makefile rule
	* tests/ (dist-hook): Remove rule.  No longer needed.

	maint: adjust formatting in tests/
	* tests/ (TESTS, CLEANFILES): Align and sort.

	build: avoid warnings in gnulib-supplied regex files
	Now that we enable more warnings in lib/, we choose
	to avoid some via patches applied by bootstrap, using
	files in the gl/ hierarchy.  Other, less-important
	warnings are avoided simply by turning off the
	-Wold-style-definition option and using a slightly
	relaxed set of warnings $(GNULIB_WARN_CFLAGS) in lib/.
	* gl/lib/regcomp.c.diff: Avoid warnings.
	* gl/lib/regex_internal.c.diff: Likewise.
	* gl/lib/regex_internal.h.diff: Likewise.
	* gl/lib/regexec.c.diff: Likewise.
	* (GNULIB_PORTCHECK): Disable only -Wold-style-definition.
	* lib/ (AM_CFLAGS): Use $(GNULIB_WARN_CFLAGS) rather
	than the slightly more strict $(WARN_CFLAGS).

	tests: adjust spencer #37 to pass with gnulib's regex code
	* tests/spencer1.tests: Change #37 to expect an exit status of 2, not 1.
	grep 'a[b-a]' reports "Invalid range end".

	maint: use regex from gnulib, rather than our bit-rotting one
	* bootstrap.conf (gnulib_modules): Add regex.
	* Don't use jm_INCLUDED_REGEX.
	Update use of cache variable.
	* lib/regex.c: Remove file.
	* lib/regex.h: Likewise.
	* m4/regex.m4: Likewise.
	* Update to match.

	build: update gnulib submodule to latest

2010-01-28  Jim Meyering  <>

	maint: update to latest gnulib; adjust
	* gnulib: Update submodule to latest.
	* (old_NEWS_hash): Update to reflect NEWS Copyright line change.

2010-01-06  Jim Meyering  <>

	maint: avoid old jm_* macros
	There were jm_* macros here, until very recently.
	* (sc_prohibit_jm_in_m4): New rule, from coreutils.

	maint: remove decl.m4
	* m4/decl.m4: Remove unused file.

	maint: rely on gnulib's new isdir.h
	* src/grep.c: Include "isdir.h".
	* src/system.h: Remove declaration of isdir.

	build: rename local to avoid shadowing global, dfa
	* src/dfa.c (dfamust): Rename parameter: s/dfa/d/.

	build: avoid warning from -Wmissing-prototypes
	* src/dfa.c (match_mb_charset): Declare to be static.

	build: avoid shadowing warning for "link"
	* src/kwset.c (link): Define to kwset_link, to avoid shadowing
	the function.

	build: avoid shadowing warning for unused "rs"
	* src/dfa.c (transit_state): Remove dead stores;
	move a declaration "down".
	Ignore transit_state_consume_1char return value.

	build: avoid shadowing warnings
	* src/dfa.c (match_mb_charset): Rename parameter: s/index/idx/.
	(check_matching_with_multibyte_ops, match_anychar): Likewise.

	build: avoid warning about unused definition of N_
	* src/dfa.c (N_): Remove unused definition.

	build: avoid format-string warnings
	* src/search.c (dfaerror): Use literal "%s" as format string.
	(kwsmusts, GEAcompile): Likewise.
	(Pcompile): Likewise.

	build: add configure-time --enable-gcc-warnings option; avoid warnings
	* bootstrap.conf (gnulib_modules): Add "manywarnings" module.
	* Add --enable-gcc-warnings, derived from code in bison.
	* lib/ (AM_CFLAGS): Likewise, but append.

	build: remove now-useless -I../intl option
	* src/ (INCLUDES): Remove -I../intl, now that intl is gone.

	maint: avoid more warnings
	* src/grep.c (MAX): Remove definition of unused macro.
	(usage): Declare with __attribute__ ((noreturn)).
	Split long strings into chunks of length < 509.

	fix a possible bug: remove errant semicolon
	* src/grep.c (prline): Remove erroneous semicolon-after-if-expr.

	maint: avoid compilation warnings
	* bootstrap.conf (gnulib_modules): Add ignore-value.
	* src/search.c (check_multibyte_string_no_icase): A variant of
	check_multibyte_string that does *not* convert case, and hence
	does not modify its BUF parameter.
	(check_multibyte_string): Use xcalloc in place of xmalloc+memset.
	Use ignore_value to ignore the return value from wcrtomb.  This is
	ok, since we know the input is a valid upper case wide character.
	(Fexecute, EGexecute): Update callers of check_multibyte_string
	to use both it and check_multibyte_string_no_icase.

	maint: avoid warnings about unused fwrite return value
	* bootstrap.conf (gnulib_modules): Add unlocked-io.
	* src/system.h: Include "unlocked-io.h".

	maint: remove {m4,lib}/.gitignore; they were undergoing too much churn
	* .gitignore: Ignore all of m4/* except m4/djgpp.m4
	and all of lib/* except, savedir.c and savedir.h.
	* m4/.gitignore: Remove file.
	* lib/.gitignore: Remove file.

2010-01-05  Jim Meyering  <>

	build: run gnulib's tests, too
	* (SUBDIRS): Add gnulib-tests.
	* gnulib-tests/ New file.
	* bootstrap.conf (bootstrap_epilogue): New function, from coreutils.
	(gnulib_tool_option_extras): Define.
	* Add gnulib-tests/Makefile.

2010-01-03  Jim Meyering  <>

	maint: record update-copyright options for this package
	* Next time, just run "make update-copyright".

2010-01-01  Jim Meyering  <>

	maint: update all FSF copyright year lists to include 2010
	Use this command:
	git ls-files |grep -vE '^(\..*|COPYING|gnulib)$' |xargs \
	env UPDATE_COPYRIGHT_USE_INTERVALS=1 build-aux/update-copyright

2009-12-23  Jim Meyering  <>

	fix multi-byte-locale read-beyond-end-of-buffer error
	Avoid read-beyond-end-of-buffer errors, evoked by running this:
	LC_ALL=en_US.UTF-8 valgrind src/grep -f <(printf 'a\nb\n') <(echo c)

	Conditional jump or move depends on uninitialised value(s)
	   at 0x78136D: __gconv_transform_utf8_internal (in /lib/
	   by 0x7E7232: mbrtowc (in /lib/
	   by 0x8055773: dfaexec (dfa.c:2816)
	   by 0x804D7B0: EGexecute (search.c:353)
	   by 0x804ACD8: grepbuf (grep.c:1036)
	   by 0x804B023: grep (grep.c:1156)
	   by 0x804B460: grepfile (grep.c:1287)
	   by 0x804CF0D: main (grep.c:2282)

	Conditional jump or move depends on uninitialised value(s)
	   at 0x7E7248: mbrtowc (in /lib/
	   by 0x8055773: dfaexec (dfa.c:2816)
	   by 0x804D7B0: EGexecute (search.c:353)
	   by 0x804ACD8: grepbuf (grep.c:1036)
	   by 0x804B023: grep (grep.c:1156)
	   by 0x804B460: grepfile (grep.c:1287)
	   by 0x804CF0D: main (grep.c:2282)

	* src/dfa.c (dfaexec) [MBS_SUPPORT]: Do not access one byte beyond
	end of buffer.

2009-12-23  Jim Meyering  <>

	build: update gnulib submodule to latest

2009-12-23  Paolo Bonzini  <>

	Speed up insert.
	Suggested by Johan Walles <> (bug 23354).

	* src/dfa.c (insert): Use binary search.

2009-12-23  Johan Walles  <>

	Decrease epsclosure memory usage
	Fixes bug 23321.

	* src/dfa.c (epsclosure): Make visited an array of char.

2009-12-22  Paolo Bonzini  <>

	Make 'grep -1 -2' and 'grep -1v2' equivalent to grep -2
	Fixes bug 12128.

	* src/grep.c (get_nondigit_option): Reset the buffer every time
	a non-digit option is found or a new argument is started.

2009-12-22  Paolo Bonzini  <>

	Improve description of --label
	Fixes bug 22681.

	* doc/grep.1 (--label): Use -H in the example, improve wording.
	* doc/grep.texi (Output Line Prefix Control): Likewise.

2009-12-22  Paolo Bonzini  <>

	Avoid using an invalid memchr result.
	Related to bug 13161.  I cannot find a testcase, but it is better to be
	defensive considering that these bug were found in the past.

	* src/search.c (EGexecute, Fexecute): Check for memchr return values.

2009-12-11  Jim Meyering  <>

	build: update gnulib submodule to latest

2009-12-04  Jim Meyering  <>

	maint: enable prohibit_have_config_h check
	* (local-checks-to-skip): Enable sc_prohibit_have_config_h
	* lib/regex.c: Remove useless cpp test of HAVE_CONFIG_H.
	* lib/savedir.c: Likewise.
	* src/grep.c: Likewise.
	* src/kwset.c: Likewise.
	* src/search.c: Likewise.

	maint: enable cast_of_x_alloc_return_value check
	* (local-checks-to-skip): Enable sc_cast_of_x_alloc_return_value.
	* .x-sc_cast_of_x_alloc_return_value:
	* src/dfa.c (CALLOC, MALLOC, REALLOC): Remove casts.
	* src/dosbuf.c (undossify_input): Likewise.
	* src/grep.c (print_line_middle, prepend_default_options): Likewise.

	maint: enable cast_of_alloca_return_value check
	* (local-checks-to-skip): Enable sc_cast_of_alloca_return_value.
	* .x-sc_cast_of_alloca_return_value: New file.

2009-12-04  Paolo Bonzini  <>

	fix "grep -Ff" on CRLF-terminated files
	* src/search.c (Fcompile) [HAVE_DOS_FILE_CONTENTS]: Recognize \r\n as
	a line terminator.

	fix compilation with included regex
	* (libgreputils_a_DEPENDENCIES): New.

	switch to pkg-config for PCRE detection
	* use pkg-config to detect PCRE
	* src/ (grep_LDADD): link grep with PCRE_LIBS

2009-12-04  Jim Meyering  <>

	maint: remove "missing" script
	* missing: Remove now-unused file.

	maint: make .gitignore ignore more
	* .gitignore: Ignore more.

	maint: enable useless-if-before-free check
	* (local-checks-to-skip): Enable sc_avoid_if_before_free.
	* .x-sc_avoid_if_before_free: New file.  Exempt regex.c and dfa.c,
	in case anyone ever tries to merge their contents with other versions.
	* src/grep.c (print_line_middle, grepdir): Remove useless if-before-free.
	* src/search.c (IF_BK, EXECUTE_FCT): Likewise.

	maint: enable po-check
	* (local-checks-to-skip): Enable sc_po_check.
	* po/ Sort and update.

2009-12-03  Paolo Bonzini  <>

	update gnulib, fixing missing inclusion of stdbool.h
	* gnulib: Update.

2009-11-30  Jim Meyering  <>

	maint: enable two checks
	* (local-checks-to-skip): Enable two:
	sc_prohibit_xalloc_without_use sc_two_space_separator_in_usage
	* src/grep.c (usage): Conform: use two spaces, not 1.
	* src/kwset.c (malloc): Define as a function-macro so that the
	syntax-check rule sees that we are indeed using xmalloc here.

	maint: enable makefile_path_separator check
	* (local-checks-to-skip): Enable sc_makefile_path_separator_check,
	now that the sole offender, an old po/, is gone.

	maint: remove now-generated file: po/
	* po/ Remove file, now generated via bootstrap.

	maint: enable makefile @...@ check
	* (local-checks-to-skip): Enable sc_makefile_check.
	* lib/ (libgreputils_a_LIBADD): Use $(...), rather than
	anachronistic @...@ notation.
	* src/ (LDADD): Likewise.
	* tests/ (AWK): Remove definition.

	maint: enable trailing_blank check
	* (local-checks-to-skip): Enable sc_trailing_blank.
	* AUTHORS: Remove trailing blanks.
	* COPYING: Likewise.
	* README: Likewise.
	* README-alpha: Likewise.
	* README-boot: Likewise.
	* THANKS: Likewise.
	* TODO: Likewise.
	* src/dfa.c: Likewise.
	* src/mbsupport.h: Likewise.
	* tests/ Likewise.
	* tests/ Likewise.
	* tests/ Likewise.
	* tests/tests: Likewise.
	* vms/README: Likewise.
	* vms/ Likewise.

	maint: enable unmarked_diagnostics check
	* (local-checks-to-skip): Enable sc_unmarked_diagnostics
	* src/grep.c (fillbuf): Mark a diagnostic for translation.
	(reset): Likewise.

	maint: enable require_config_h checks
	* (local-checks-to-skip): Enable sc_require_config_h
	and sc_require_config_h_first.
	* src/dosbuf.c: Include <config.h>.
	* src/vms_fab.c: Likewise.
	* .x-sc_require_config_h: New file: list the exceptions.
	* .x-sc_require_config_h_first: Likewise.

	maint: use gnulib's progname module; enable set_program_name check
	* bootstrap.conf (gnulib_modules): Add progname.
	* src/grep.c: Include "progname.h".
	(program_name): Remove declaration.
	(main): Call set_program_name.
	* (local-checks-to-skip): Add sc_program_name.

	maint: enable "file system" check
	* (local-checks-to-skip): Enable sc_file_system.
	* lib/savedir.c (savedir): Tweak spelling.  Remove trailing blanks.

	maint: enable immutable_NEWS check
	* NEWS: Move copyright to the bottom.
	Use the format required by release-related tools.
	* .prev-version: New file.
	* (old_NEWS_hash): Define.
	(local-checks-to-skip): Enable check: sc_immutable_NEWS.

	maint: disable the many failing syntax-checks
	* New file.
	(local-checks-to-skip): Define to the list of disabled rules.
	Subsequent change-sets will enable them, one by one.

	build: require automake-1.11, enable silent-rules, parallel tests, xz
	* (AM_INIT_AUTOMAKE): Create xz-compressed tarballs,
	not bzip2-compressed ones.  Enable automake's silent-rules,
	parallel tests, and test PASS/FAIL coloring options.
	Use AC_CONFIG_HEADERS, not AM_CONFIG_HEADER.  Quote the argument.

	build: use git-version-gen for inter-release version strings
	* (AC_INIT): Use git-version-gen.

	build: add several build- and release-related gnulib modules
	* bootstrap.conf (gnulib_modules): Add announce-gen update-copyright
	do-release-commit-and-tag git-version-gen gnu-web-doc-update
	gnupload maintainer-makefile useless-if-before-free

	build: adapt to the newer closeout module from gnulib
	* src/grep.c: Include "exitfail.h".
	(main) [-q]: Set the global variable, exit_failure, rather than
	calling the now-removed close_stdout_set_file_name function.

	build: adapt to the newer exclude API we now get from gnulib
	* src/grep.c (main): Adapt to newer exclude.c: add EXCLUDE_WILDCARDS as
	the new "option" argument in calls to add_exclude and add_exclude_file.

	build: get more lib/* files from gnulib, adjust savedir
	* bootstrap.conf (gnulib_modules): Add the following:
	closeout exclude hard-locale isdir strtoumax.
	* lib/.gitignore, m4/.gitignore: Update.
	* lib/closeout.c, lib/closeout.h: Remove.
	* lib/exclude.c, lib/exclude.h: Remove.
	* lib/hard-locale.c, lib/hard-locale.h: Remove.
	* lib/strtoumax.c: Remove.
	* lib/isdir.c: Remove.
	* lib/ Remove here, too.
	* lib/savedir.c: Adapt to new exclude module:
	s/excluded_filename/excluded_file_name/ and remove 3rd argument.

	build: update gnulib submodule to latest

	maint: generate ChangeLog from git logs
	* (dist-hook, gen-ChangeLog): New rules.
	* bootstrap.conf (gnulib_modules): Add gitlog-to-changelog.
	Ensure that ChangeLog exists.
	* ChangeLog-2009: Rename from ChangeLog
	* ChangeLog: Remove file.
	* .gitignore: Add ChangeLog.

	maint: list gnulib modules one per line
	* bootstrap.conf (gnulib_modules): List them one per line.

2009-11-29  Tony Abou-Assaleh  <>

	Acknowledge new maintainers, update README-alpha
	* AUTHORS: new maintainers added
	* THANKS: same
	* README-alpha: change CVS references to Git