# This is ChangeLog for changes before enca became a library.

 # The new ChangeLog starts after that.

 # Unfortunately, there's no ChangeLog for the transition.

 # Note: EVERYTHING as changed file means global change, i.e. `every file where

 # this change was applicable was changed'.

2002-06-23 Yeti <yeti@physics.muni.cz>

    * src/texts.c, src/common.h, src/options.c: define e_isatty() as portable

      isatty in common.h, print help instead of wainting for input when stdin

      is a tty

v0.9.4

2002-03-04 Yeti <yeti@physics.muni.cz>

    * src/printresult.c: fixed print_detailed_report() not to truncate charset

      names in the table

    * src/convert.c: put a no-op file_seek() after file_write() to make

      subsequent file_read() work correctly (an ISO C ,feature`, not Enca's)

2002-03-03 Yeti <yeti@physics.muni.cz>

    * src/convert.c: fixed convert_builtin() to convert whole files, not just

      first 64kB

2001-08-16 Yeti <yeti@physics.muni.cz>

    * src/options.c: removed strange `:' from start of short_options

2001-08-01 Yeti <yeti@physics.muni.cz>

    * src/common.h: defined e_free(x) to set x to NULL after freeing it

    * src/EVERYTHING: formal changes

v0.9.3

2001-07-16 Yeti <yeti@physics.muni.cz>

    * configure.in, m4/librecode.m4, m4/iconv.m4, src/Makefile.am,

      src/epress.c: conversion libraries are put into CONVERTER_LIBS, so

      epress doesn't link with them (they are included in enca_LDADD), LIBS

      is not modified by librecode and iconv tests

    * iconvcap.c: added program_name definition for the case it would link

      with librecode (shouldn't happen)

    * configure.in: added AC_AIX

    * Makefile.am: added forgotten m4/crash-me to EXTRA_DIST

    * src/encnames.c: --list=encodings prints only `known' encodings

    * Makefile.am: added topline.sh (needed by update target) to dist

    * updated docs

2001-07-15 Yeti <yeti@physics.muni.cz>

    * src/lang.c: fixed giving random results due to usage of unitialized

      memory in get_charsets()

    * src/printresult.c: fixed using double instead of int for ambiguous and

      multibyte flags in T_EncDetails structure

    * src/guess.c, src/printresult.c, src/printresult.h: added

      GUESS_QP_RESOLVED flag to be able to print logically consistent -d output

      in case of quoted printables

    * src/fileio.c, configure.in: include sys/stat.h unconditionally, we

      cannot compile w/o it anyway, warning added to configure

    * src/common.h: include both string[s].h and memory.h. instead of just one

      of them

    * src/common.h: define PACKAGE, VERSION, DEFAULT_EXTERNAL_CONVERTER and

      DEFAULT_CONVERTER_LIST when don't have config.h to compile w/o it

    * src/options.c: fixed too many \n's in feature list

    * configure.in, src/texts.c: added isatty() and ttyname() test, defined

      stdout_isatty(), pager is run only when we are able to positively say

      stdout is a tty

2001-07-14 Yeti <yeti@physics.muni.cz>

    * src/options.c, src/common.h, src/texts.c: made USE_PAGER a feature +pager

    * m4/recode-bugs.m4: added TeX/..ISO-8859-2 recode crash test

    * m4/iconv.m4, src/convert_iconv.c: added ICONV_ARG2_CONST test

    * m4/typevar.m4, src/common.h, configure.in: removed the no longer needed

      long long int test, mere long int is used for mathint

    * configure.in: prepend "-Wall -pedantic" before CFLAGS when compiles is

      GCC

    * Makefile.am: add a new debian changelog entry when version or release

      changes, otherwise update time of the current entry

2001-07-13 Yeti <yeti@physics.muni.cz>

    * src/fileio.c, src/convert_recode.c: removed the no-buffering tricks

      making it slightly faster on modern GNU/Linux, but much slower everywhere

      else

    * src/EVERYTHING: put back the #ifdef HAVE_CONFIG_H stuff

    * src/texts.c: fixed typo expand_char() -> fputc() in poor man's compress

    * src/common.c: added const to stpcpy() *p declaration to keep qualifiers

    * m4/pager.m4, Makefile.am, configure.in, src/texts.c: test whether less

      accepts -F, possibility to disable pager at all

v0.9.2

2001-07-13 Yeti <yeti@physics.muni.cz>

    * src/guess.c: fixed reporting `7bit ASCII characters' after a successfull

      detection of a 7bit encoding (TeX, UTF-7)

    * src/convert_recode.c: fixed typo HAVE_RECODEEXT_H -> HAVE_RECODEXT_H

2001-07-12 Yeti <yeti@physics.muni.cz>

    * src/filters.c: fixed hdata->eid's comparsion before their initialization

      in lang_hook_2cs() (so the hook was never run)

    * src/convert_recode.c: more correct handling of the situation when we

      realise we cannot seek in temporary file during recode failure recovery

    * debian/Makefile.am, m4/Makefile.am, Makefile.am, configure.in: removed

      the first two, debian/ generated by toplevel Makefile, m4/ is just

      distributed

    * Makefile.am, src/Makefile.am: replaced $< with literal file names, some

      `make's don't always substitute first dependence name

    * src/epress.c: added program_name definition to placate librecode

    * src/common.h: stdlib.h and unistd.h are included unconditionally (we

      depend on them anyway)

    * src/texts.c: replaced the `is pager less?' test with a better one

    * configure.in, acconfig.h, m4/EVERYTHING: put descriptions directly to

      AC_DEFINE[_UNQUOTED]'s, almost get rid of acconfig.h

    * src/EVERYTHING: don't trust make passing `-DHAVE_CONFIG_H -I..' to

      compiler, #include "../config.h" unconditionally (we depend on it anyway)

2001-07-10 Yeti <yeti@physics.muni.cz>

    * debian/, Makefile.am: created debian/ and modified the Makefile.am to

      include it

2001-07-09 Yeti <yeti@physics.muni.cz>

    * configure.in, enca.spec.in: added MAINTAINER variable which can be used

      in rpm spec and debian/ files

    * src/texts.c: fixed wrong prototype putchar() -> expand_char() when

      no compressor is available

    * src/texts.c, acconfig.h, configure.in: compressed texts are displayed

      through a pager (if available)

    * autogen.sh, README.devel: added and make a note about it

v0.9.1

2001-06-25 Yeti <yeti@physics.muni.cz>

    * src/guess.c: fixed possible overflow in UCS-4 test (slightly changing

      what gets recognized as UCS-4)

    * src/guess.c: implemented unusual byteorders (3412 and 2143) tests using

      a single little-endian ucs-4 test what_if_it_was_ucs4() and shuffling

      bytes around shuffle_byte_order(), the same for UCS-4 EOL type tests

    * src/common.h: defined PVAR(f, v) [for debugging]

    * updated docs

    * packaged

2001-06-24 Yeti <yeti@physics.muni.cz>

    * src/encnames.h, src/encnames.c, src/guess.c: renamed SURF_PER_12[34] to

      more logical SURF_PER_[43]21

2001-06-02 Yeti <yeti@physics.muni.cz>

    * src/efilter.c, src/Makefile.am, src/texts.c, src/common.h: created

      efilter.c (filter making text files to compress better) and adding

      reverse filter to texts.c

    * m4/tolower.m4, configure.in, acconfig.h, src/common.h: define our own

      implementation of tolower and similar unconditionally

    * improved various docs

    * BUGS, Makefile.am: generated from manual page section of the same name

    * src/efilter.c, src/Makefile.am, src/texts.c, src/common.h: efilter was

      funny experiment but not much useful, removed again

2001-06-01 Yeti <yeti@physics.muni.cz>

    * src/getopt.c, src/getopt1.c, src/getopt_long.c, src/Makefile.am: removed

      the first two and made getopt_long.c non-generated, removed some unneeded

      stuff

    * src/common.c: put broken-{m,c,re}alloc fixes into conditionals (but no

      autoconf tests---I don't believe such a broken systems really do exist)

2001-05-28 Yeti <yeti@physics.muni.cz>

    * src/options.c: fixed prepend_env() segfault, improved diagnostics

    * revised temporary files usage, removed `/tmp bug' from docs

    * convert.c, convert_iconv.c, convert_recode.c: renamed tmpfile variable

      (synonymous to ISO function name) to tempfile, better not to tempt fate

    * script/b-map, script/b-cstocs, script/b-recode: added umask 077

2001-05-27 Yeti <yeti@physics.muni.cz>

    * src/encnames.c, src/encnames.h, man/enca.1: checks that all characters

      in charset/surface names are from some set of allowed characters

    * src/common.c: added hack for systems that fail on malloc(0) and similar

2001-05-20 Yeti <yeti@physics.muni.cz>

    * src/epress.c: forgotten [ISO C99] int16_t changed to int

    * src/filters.c, src/filters.h: added universal decide-between-2-cs hook

    * src/lang_cs.c, src/lang_sk.c: modified to use the universal hook,

      some formal changes

    * src/lang_ru.c: added maccyr/cp1251 hook (via the universal)

    * src/common.h, src/enca.c: sensible exit code (0, 1 or 2) returned

    * src/convert.c: convert() returns ERR_* error codes instead of just 0, 1

    * src/common.c, src/convert.c, src/convert_iconv.c, src/encnames.c,

      src/fileio.c, src/lang.c, src/locale_detect.c, src/options.c,

      src/texts.c: returned 2 on troubles

    * updated docs

2001-05-18 Yeti <yeti@physics.muni.cz>

    * src/license.c, src/license.h, src/texts.c, src/texts.h: renamed the

      former to the latter

    * src/epress.c, src/texts.c: implemented bzip2 and gz interface

    * m4/compress.m4, configure.in: added tests for libbz2 and libz, the best

      one found is used

    * configure.in, src/Makefile.am, src/options.c: COPYING.c and HELP.c are

      generated by epress and just linked with enca

    * src/Makefile.am: put getopt_long.c into BUILT_SOURCES

    * src/lang_ru.c, src/unicodemap.c, src/encnames.c: added maccyr charset

    * Makefile.am: added hook to delete BUILT_SOURCES before making dist

2001-05-17 Yeti <yeti@physics.muni.cz>

    * src/fileio.c: temporary files are created with umask 077

2001-05-06 Yeti <yeti@physics.muni.cz>

    * src/epress.c, src/COPYING.h, src/Makefile.am: created license compressor

    * src/license.c, src/options.c: license decompression and printing

2001-05-01 Yeti <yeti@physics.muni.cz>

    * m4/librecode.m4, m4/recode-bugs.m4, m4/long-text.l2: put test for bugs to

      recode-bugs.m4 (four bugs are checked, any will launch the warning)

2001-04-12 Yeti <yeti@physics.muni.cz>

    * m4/librecode.m4, configure.in: added a test for broken recode (no

      workaround [known], just prints a big warning message)

v0.9.0

2001-03-26 Yeti <yeti@physics.muni.cz>

    * just packaged

2001-03-25 Yeti <yeti@physics.muni.cz>

    * src/convert.c: fixed reading only first file block in copy_and_convert()

    * configure.in, src/EVERYTHING: memory.h test added, string/strings/memory

      header file inclusion put into common.h

    * src/fileio.c: fixed file_setvbuf() and file_open() (not enough magic)

    * src/convert_recode.c: fixed not opening original file

    * configure.in, src/convert_recode.c: added recodext.h test, is used when

      available for setting diacritics_only and ascii_graphics flags

    * src/convert_iconv.c: fixed not opening original file

    * src/convert_iconv.c: fixed writing no output in iconv_one_step()

    * src/convert.c: fixed intermixed child and parent process in

      convert_external()

    * src/fileio.c: added file_getline() requiring much less system calls

      than GNU libc's fgets()

    * src/locale_detect.c: fixed segfault when language detection failed

    * m4/tools.m4, configure.in, acconfig.h, src/options.c: consolidated

      autoconf macros regarding external converters

    * man/enca.1: minor corrections

2001-03-24 Yeti <yeti@physics.muni.cz>

    * src/encnames.c: added quoted printable surface

    * src/guess.c: integrated quoted printable into the guessing process

    * src/encnames.c, src/options.c: added print_public_surfaces() wrapper to

      allow printing --human-readable list of surfaces

    * src/lang_cs.c, src/lang_ru.c, src/lang_sk.c: replaced slovak and russian

      statistical data with some new (hope better), some hardcoded array sizes

      now computed from sizeof()

    * iconvcap.c: cosmetical changes

2001-03-20 Yeti <yeti@physics.muni.cz>

    * src/fileio.c, src/fileio.h, src/common.c, src/common.h: T_Buffer type

      moved from fileio to common

    * src/lang.c, src/lang.h, src/lang_cs.c, src/lang_ru.c, src/lang_sk.c:

      use the new T_Buffer type

    * configure.in: removed now-unused tests

    * src/EVERYTHING: removed redundant header inclusion, ensured satisfying

      header dependencies, try to use strings.h when string.h is not avaialble

    * src/guess.c: fixed not initializing memory buffer

    * src/options.c: exchagned -E and -C option letters (incompatibility!)

    * src/guess.c: implemented quoted printable recognition (unused now)

2001-03-19 Yeti <yeti@physics.muni.cz>

    * src/fileio.c, src/fileio.h: some more magic employed, only initiated

      can use them now

    * src/convert.c, src/convert_iconv.c, src/convert_recode.c, src/lang.c:

      changed to use the new file i/o interface

    * src/convert_iconv.c: fixed a memory leak

2001-03-18 Yeti <yeti@physics.muni.cz>

    * src/fileio.c, src/fileio.h: finished implementation

    * src/common.c, src/common.h: removed stuff belonging to fileio

    * src/locale_detect.c, src/guess.c, src.guess.h, src/lang.h, src/enca.c,

      src/convert.h: changed to use the new file i/o interface

    * src/guess.c: UCS tests check remainder of file length instead of buffer

      position (which is still used as a fallback in case of stdin)

2001-03-17 Yeti <yeti@physics.muni.cz>

    * configure.in, script/b-cstocs, script/b-cstocs.in, script/b-map,

      script/b-map.in, script/b-recode, script/b-recode.in, script/Makefile.am:

      (I'm stupid) reverted the last change

    * src/fileio.c, src/fileio.h: created (unique file i/o interface, at last)

2001-03-12 Yeti <yeti@physics.muni.cz>

    * m4/tools.m4: AC_MSG_WARN is used to print warnings

    * m4/iconv.m4: iconvenc.h is always created (the dark side is that iconv

      usability is no longer cached)

    * m4/tools.m4, configure.in, script/b-cstocs, script/b-cstocs.in,

      script/b-map, script/b-map.in, script/b-recode, script/b-recode.in,

      script/Makefile.am: external converters are located including path and

      it's then used for configure-time substitution in the scripts

v0.9.0pre5

2001-03-12 Yeti <yeti@physics.muni.cz>

    * src/common.h, src/common.c, src/encnames.h, src/guess.c, src/options.c,

      src/printresult.c, src/convert.c: eliminated rest of Settings, replaced

      by module methods, except read_limit which is still global

    * enca.spec.in: various improvements (see the spec file changelog)

2001-03-11 Yeti <yeti@physics.muni.cz>

    * src/EVERYTHING: Settings.ProgName and Settings.Verbose made globals:

      program_name and verbosity_level

    * src/common.h, src/options.c, src/convert.c: Settings.ExtConverter made

      convert.c module global extern_converter, set by set_external_converter()

    * src/common.h, src/options.c, src/printresult.c: Settings.PreFName made

      printresult.c module global extern_converter, set by

      print_set_prefix_filename()

2001-03-10 Yeti <yeti@physics.muni.cz>

    * enca.spec.in: use global cache file

2001-03-09 Yeti <yeti@physics.muni.cz>

    * Makefile.am: (generated!) enca.spec included to distribution (rpm -tb)

2001-03-02 Yeti <yeti@physics.muni.cz>

    * man/enca.1: spell-checked

    * src/options.c: corrected available lists in --list= help

2001-02-28 Yeti <yeti@physics.muni.cz>

    * src/lang.c: language is printed in lang_init() when verbose

    * src/guess.c: all Latin, Cyrillic and Greek letters have equal weight in

      UCS-2 and UCS-4 tests

    * src/guess.c: fixed bad default endianess and initial makes-sense-check in

      UCS-4 test

    * src/printresult.c: UCS-* rating is printed, too

v0.9.0pre4

2001-02-26 Yeti <yeti@physics.muni.cz>

    * just packaged

2001-02-25 Yeti <yeti@physics.muni.cz>

    * src/options.c, src/encnames.c, src/printresult.c: conversion from output

      type to what-name is done in encnames.c, introduced new output types

      and what-names ALIASES and NONE, --name=aliases prints list of aliases,

      removed --list=aliases, --list=encodings is sensitive to value of --name=

      instead, --list=names lists all valid --name= values, default output type

      is now NONE (initialized after option processing when needed)

    * updated rest of docs

    * configure.in, m4/libm.m4: fixed `make install' fail due to dependance on

      libtool: get rid of libtool dependance, created our own math library

      test

2001-02-24 Yeti <yeti@physics.muni.cz>

    * configure.in, acinclude.m4, m4/: removed acinclude.m4, created m4 dir and

      put the definitions there

v0.9.0pre3

2001-02-23 Yeti <yeti@physics.muni.cz>

    * src/options.c: added --list=lists listing

    * NEWS: updated

    * man/enca.1: partly updated

2001-02-22 Yeti <yeti@physics.muni.cz>

    * configure.in: added feature and failure lists

    * acinclude.m4: cleanup, defined some *_ok variables for feature list,

      fixed some `! test ... = ...' to `test ... != ...'

    * srcipt/b-cstocs, srcipt/b-map, srcipt/b-recode: `#!/bin/bash' changed

      to `#! /bin/sh' (even if I'm not sure whether the scripts are 100%

      sh-compatible)

    * iconvcap.c: initial ASCII test changed back to ISO-8859-1 since the

      former is not supported on some systems (SunOS, IRIX, ...) having

      otherwise-usable iconv

    * src/getopt_long.c: generated from getopt.c and getopt1.c, but distributed

      (mainly because automake is unable to understand something we have in

      LIBOBJS can be generated...)

v0.9.0pre2

2001-02-22 Yeti <yeti@physics.muni.cz>

    * configure.in, src/Makefile.am: merged src/getopt.c and src/getopt1.c to

      src/getopt_long.c to allow simple AC_REPLACE_FUNC() be used for

      getopt_long optional build

    * acinclude.m4: simplified iconv test and added check for libiconv

2001-02-21 Yeti <yeti@physics.muni.cz>

    * Makefile.am: corrected iconvconf.h to iconvenc.h

    * src/options.c: fixed undefined and misused ENCA_ENV_VAR when we don't

      have wordexp() to parse it

    * acinclude.m4: iconvenc.h is always created since I'm not able to explain

      automake we don't need it when we don't have iconv

v0.9.0pre1

2001-02-20 Yeti <yeti@physics.muni.cz>

    * src/convert.c: trivial conversions from ASCII are carried out in built-in

      converter

    * src/EVERYTHING: cleaned terminology encoding vs. charset; encoding means

      (charset,surface) pair

    * src/options.c: converter names are case sensitive again

    * config.h is included only when HAVE_CONFIG_H

    * iconvenc.nul.h, acinclude.c, src/encnames.c: iconvenc.nul.h merged into

      encnames.c (neede only there when !HAVE_ICONV)

    * configure.in, src/getopt.c, src/getopt.h, src/getopt1.c, src/options.c:

      added GNU getopt so long options can be used everywhere (needs testing)

    * src/options.c: -m as complement to -M resets affected options to defaults

2001-02-19 Yeti <yeti@physics.muni.cz>

    * src/common.h, src/options.c, src/printresult.c: added iconv name output

      type, `-i' (only when HAVE_ICONV)

    * src/options.c: `-n' changed to `-e'

    * src/options.c: output types (except `-x') can be specified as -n NAME

    * iconvcap.c: updated for the new encodings and changed ICONV_ prefix to

      ICONV_NAME_, also changed the suffixes to pure alphanumeric, ISO-8859-1

      test changed to ASCII

    * src/convert_iconv.c: updated to the new encoding/surface model

    * src/guess.c: try to detect swapped UCS-2 from the first byte pair, too

    * src/options.c: pointer woodoo to comply ISO C in print_some_list(): one

      more level of indirection in abbreviation table data

    * src/options.c, src/encnames.c: iconv name output type defined always, but

      when iconv names are not available error message is printed instead

    * iconvconf.h, iconvenc.h: the former renamed to the latter

    * iconvenc.h.nul: created. contains all iconv names defined to NULL

    * acinclude.m4, src/encnames.c, src/convert_iconv.c: iconvenc.h is either

      successfully generated by iconvcap or copied from iconvenc.h.nul so it

      always exists and is always #included

    * configure.in, src/options.c: getopt.h test (defines getopt_long(), etc.)

    * src/common.c: expand_abbreviation() returns pointer to whole abbreviation

      structure so it's possible to fetch the expanded name, too; all callers

      changed

    * script/bcstocs, script/brecode, script/b-cstocs, script/b-recode: the

      first two renamed to the second two

    * script/b-map: created. perl `map' wrapper

2001-02-18 Yeti <yeti@physics.muni.cz>

    * src/common.c: e_tmpfd() now returns empty string a file name when fails

    * src/options.c: max read limit increased to 16MB

    * src/options.c: default converter list changed, now is

      "built-in,librecode" for +librecode-interface, [+-]iconv-interface

      "built-in,iconv" for -librecode-interface, +iconv-interface

      "built-in" for -librecode-interface, -iconv-interface

    * src/convert_recode.c: cleaned-up to use the new encoding/surface model

    * src/convert.c, src/convert_common.h, src/convert_recode.c: request

      formatting (either for printing or for recode requests) function

      format_request_string()

    * src/convert_recode.c: temporary copy of original file is not deleted

      when rescue of damaged original fails (what a paranoia!)

    * src/common.c: e_fopen() and e_fclose(), fopen() and fclose() wrappers,

      unsed in various places

    * src/EVERYTHING: perror(NULL) changed to perror("") since some C libraries

      don't understand the former to print empty prefix string

    * src/encnames.c, src/options.c: added `surfaces' to --list option

    * src/convert.c: valid converter names are printed one per line

    * src/encnames.c: added iconv name to T_EncInfo, all defined to NULL when

      iconv interface is not build

    * src/common.c: implemented abbreviation searching expand_abbreviation()

    * src/options.c: print_some_list() uses the new abbreviation engine

    * src/convert.c: add_converter() uses the new abbreviation engine (needed

      some data types shake-up)

    * src/lang.c: got rid of `initialization discards qualifier...' warnings

      by shaking up with the consts (no functionality changed)

2001-02-17 Yeti <yeti@physics.muni.cz>

    * src/common.h: surface type formalized to surfint (instead of uint16)

    * src/encnames.c: implemented surface <-> name conversion

    * src/encnames.c, src/guess.c: changed UCS surfaces to recode style

    * src/options.c, src/enca.c: updated parse_x_arg(), removed

      dull-converter-like beahviour

    * src/locale_detect.c: get rid of str[c]spn() (they are ISO C, but autoscan

      tries to convince me I should check for them), flag-table approach is

      more efficient anyway

    * src/common.c: commented out strstrcount() (not needed)

    * src/encnames.c, src/encnames.h: empty surface SURF_REMOVE (to get only

      "/" as surface name) (dirty)

2001-02-14 Yeti <yeti@physics.muni.cz>

    * src/options.c: option letter space polluted by various listing options

      cleaned by introducing --list=WORD (incompatible with previous -l usage)

    * src/encnames.c: print_encoding_aliases() prints list of accepted encoding

      aliases (`recode -l' style)

    * src/encnames.c: encodings sorted alphabetically (by canonical name)

    * src/lang.c, src/lang_cs.c, src/lang_sk.c, src/lang_ru.c: ability to

      print what regular encodings belong to this particular language

2001-02-12 Yeti <yeti@physics.muni.cz>

    * src/lang.c: fixed not initializing language filter report for language

      `none'

    * src/encnames.h, src/encnames.c: introduced ENCF_MULTI flag marking

      multibyte encodings

    * src/printresults: heavily improved details (but still misses surfaces)

    * src/options.c: file name is printed with details except when user

      specifically asks not to print it

    * src/options.c: help texts updated

    * src/encnames.c, src/printresults.c: cstocs names of encodings not known

      to it changed to ???, but still printed the same in details

    * src/defaults.h: removed. all definitions moved to appropriate C files

      (or common.h)

    * src/encnames.c, src/options.c, src/prinresult.c: removed option `-g'

    * src/encnames.c: presqueezed aliases replaced by normal, we now squeeze

      names when needed (with hashing it's OK), also changes canonical names

    * config.h.top, configure.in: removed the former and add -D_GNU_SOURCE

      directly to CFLAGS

    * configure.in: removed warnings

    * configure.in: integer type sizes tested only when system doesn't provide

      stdint.h (and int's default changed to ISO C minimum)

    * configure.in, acinclude.h: moved my tests to the latter (newly created)

      to make the former more readable

    * guess.c: TEX_* and UTF_* defines moved to appropriate functions (as const)

    * guess.c: fixed missing print_flags_or() for language hooks (from ????)

2001-02-11 Yeti <yeti@physics.muni.cz>

    * src/encnames.c: hashing is used to find encoding names

    * src/options.c: fixed not recognizing output encoding (from 2000-02-02)

    * src/unicodemap.c: `-l' prints each group on separate line

2001-02-10 Yeti <yeti@physics.muni.cz>

    * src/guess.c: handle correctly case when a language has only one encoding

    * src/guess.c: multibyte tests use count table for fast rejecting

    * src/guess.c: surface detection, eol_surface()

2001-02-08 Yeti <yeti@physics.muni.cz>

    * src/lang_ru.c, src/encnames.c, src/unicodemap.c: added IBM 866 charset

    * src/filters.h: made ibm866 filter alias to keybcs2 filter (they're

      identical)

    * src/guess.c: saved stdin is not restored and `up' is not recomputed when

      nothing was filtered out

2001-02-06 Yeti <yeti@physics.muni.cz>

    * src/detect_lang.c, src/locale_detect.c: renamed the former to the latter

    * src/filters.c, src/filters.h, src/lang_cs.c: first two created (language

      filter repository) and all box-drawing filters moved there

    * src/filters.c: added bow-draing filters for more encodings

    * src/common.h, src/options.c, src/printresult.c: implemented -n output

      type (prints `canonical' encoding name used internally in enca)

    * src/lang_sk.c, src/lang_sk.h, src/lang_ru.c, src/lang_ru.h, src/lang.c,

      src/encnames.c, src/unicodemap.c: added Slovak and Russian languages and

      appropriate encodings

    * src/unicodemap.c: introduced the idea of compatible encodings; LATIN2

      and CYRILLIC groups defined

    * all headers: revised header file dependencies

    * src/lang.c, src/locale_detect.c: implemented the notion of no

      language---when user sets language to `none' no regular encodings are

      processed (so when you say `-UL none' only pure ascii gets recognized)

    * src/options.c: fixed broken `-P' (an old bug)

2001-02-05 Yeti <yeti@physics.muni.cz>

    * src/lang.c, src/lang.h, src/lang_cs.c, src/lang_cs.h: created the first

      two and moved all regular encoding routines to them

    * src/options.c: implemented language settings (added Language to Settings)

2001-02-04 Yeti <yeti@physics.muni.cz>

    * src/unicodemap.c: is_subset_consistent() now uses translation table for

      checking

    * src/guess.c: incorporated multibyte encodings into the guessing process

    * src/guess.c: implemented an absolute likehood test of the applied to the

      relative winner

2001-02-03 Yeti <yeti@physics.muni.cz>

    * src/common.c, src/common.h: introduced type flagint for tables of flags

      (defined as short int), all callers changed

    * src/options.c: fixed accepting invalid option values, ReadLimit must be

      a multiple of 4 (to make UCS tests more reliable)

    * src/guess.c, src/printresult.c: added number of 8bit's (up) to details

    * src/guess.c: finished is_valid_utf8(), is_valid_utf7(), looks_like_tex(),

      looks_like_ucs2() and looks_like_ucs4() tests (except surfaces)

2001-02-02 Yeti <yeti@physics.muni.cz>

    * src/unicodemap.c: added many missing characters to maps

    * src/unicodemap.c: shortened maps by starting them from first character

      that doesn't map to itself

    * src/options.c: finally removed old dull-conversion syntax

    * src/options.c, configure.in: long `GNU style' options---when

      getopt_long() function is available, configure test added

    * src/options.c, configure.in: program_invocation_short_name is used when

      offered by system, otherwise strip_path() is used to make it

    * src/options.c, configure.in: value of environment variable ENCAOPT is

      prepended before command line options---when wordexp() function is

      available, configure test added

2001-02-01 Yeti <yeti@physics.muni.cz>

    * src/EVERYTHING: encoding are no longer identified by name, but integer

      eid is used; T_Encoding contains---beside eid---surface, but it's not

      used for anything yet

    * implemented name squeezing and alias recognition---almost any sensible

      (and a whole bunch of stupid) encoding identifier is recognized

    * introduced new, less stupid, `canonical' names---passed to external

      converter and looked-up faster than the others

    * src/guess.c, src/enca.c, src/printresult.c: result is passed as

      T_Encoding type

    * src/guess.c, src/printresults.c, src/lang_cs.c: print implemented as

      `object', guess details are fed by different functions when needed

      (no globals and circular dependencies anymore)

    * src/printresults.c: mapping from output type to encoding name is used

      instead of ugly switch ()

    * src/unicodemap.c: character 0xa4 from koi8cs2 converted to tilde (0x7f)

    * src/options.c: refuse to serve as a dull converter when input encoding

      is not known to us

    * src/common.c, configure.in: implemented stpcpy() when not provided by

      system (configure test added)

    * src/guess.c, src/enca.c: 7bit ascii is no longer discriminated and no

      longer causes enca return nonzero error code

2001-01-30 Yeti <yeti@physics.muni.cz>

    * src/langdata_cs.c, src/langhook_cs.c, src/lang_cs.c: the first two merged

      to the third containing [almost?] all language specific stuff

2001-01-29 Yeti <yeti@physics.muni.cz>

    * src/encnames.c: created. (language independent encoding name handling)

2001-01-28 Yeti <yeti@physics.muni.cz>

    * src/common.h, configure.in: tests for stdint.h and integer type sizes

      used to define uint16 and uint32

    * src/unicodemap.c: changed map storage type to uint16 thus saving several

      kilobytes, removed need for an empty last table

    * src/guess.c: added looks_like_ucs2(), looks_like_ucs4() and partially

      looks_like_utf7()

2001-01-27 Yeti <yeti@physics.muni.cz>

    * src/guess:c reimplemented look_like_TeX() in a more efficient way

    * src/EVERYTHING: static file globals used only in one function moved into

      the appropriate function

    * man/enca.1: made more human readable, different macros are used

2001-01-26 Yeti <yeti@physics.muni.cz>

    * src/detect_lang.c, src/detect_lang.h: detect_lang() accepts string

      parameter now

    * src/guess.c: implemented utf-8 parse test is_valid_utf8() and

      (La)TeX-encoded accents test look_like_TeX()

2001-01-25 Yeti <yeti@physics.muni.cz>

    * src/detect_lang.c, src/detect_lang.h, src/Makefile.am, src/defaults.h:

      guessing user's preferred language from locale (not used yet)

    * configure.in, acconfig.h: implemented check for locale.alias

v0.7.7

2001-01-01 Yeti <yeti@physics.muni.cz>

    * e_read4() now consistently sets number of bytes in buffer to zero even

      when reading of zero bytes is requested (seems to break nothing, but...)

    * documentation synchronized

2000-12-31 Yeti <yeti@physics.muni.cz>

    * `none' is accepted as converter name and causes clearing the converter

      list

    * finished iconv interface

    * changed some messages, hope no one parses them

    * ssize_t availability checked by configure

2000-11-29 Yeti <yeti@physics.muni.cz>

    * long long int availability is now explicitely checked by configure

    * unsigned char -> byte

    * implemented e_write4() and e_read4() that allow to specify buffer address

      and size (e_read_with_limit replaced by e_read4())

    * fixed files closed twice in copy_and_convert()

    * fixed bad return value tests for e_read() and e_write() in

      copy_and_convert()

    * implemented iconv_one_step() and enc_trans() for iconv, so it is almost

      usable on iconv-transitive (GNU) systems now

2000-11-28 Yeti <yeti@physics.muni.cz>

    * iconv transitivity is now explicitely checked by iconvcap instead of

      checking for gconv

2000-11-26 Yeti <yeti@physics.muni.cz>

    * misspelled `Europian' corrected to `European' in macce long name

2000-11-21 Yeti <yeti@physics.muni.cz>

    * fixed some nonsubstantial stuff, redundant #includes, docs and comments

      forgotten in too hastily released v0.7.6

    * convert_iconv.c separated from convert.c

v0.7.6

2000-11-20 Yeti <yeti@physics.muni.cz>

    * unicode mapping data made NULL terminated instead of fixed-length

    * discovered bug in gcc :-(

    * all upper-half-of-ascii structures promoted to 256 characters (too many

      changes to record here)

    * support for cork encoding ceased

    * help text updated and divided to sections (thus fixing warning about

      too long, ISO C violating, string)

    * get_in_enc_list() and get_converters() changed to print_* printing the

      lists, called directly from process_opt()

    * convert_recode.c separated from convert.c to make the amount of #if's

      bearable

    * encoding list made NULL terminated instead of fixed-length

    * reversed order of generating recoding table so, character with lowest

      8bit code is always outputed for synonyms (instead of highest)

    * implemented converter flags CONV_EXTERN, we do not try to recover after

      external converter failure, since it's impossible by definition

    * xlat table cache bubble sorting on use

    * man page and other documentation synchronized

2000-11-19 Yeti <yeti@physics.muni.cz>

    * a simple test suite introduced in test/ (not distributed)

    * fixed stupid || -> && bug in e_close() causing failure in pipes

    * copy_and_convert() now processes what is saved in io_buffer always when

      called on stdin (not only when recoding is done too), this fixes bug in

      conversion to the same encoding in pipe

    * fixed comparing return value of e_tmpfd() with zero (instead of -1)

      causing not calling external converter in redirection

    * stdin is always copied to stdout when we are not able to perform

      conversion, irregardles of reason

    * some error messages improved

2000-11-18 Yeti <yeti@physics.muni.cz>

    * it's up to every converter to translate cstocs encoding names to its

      native names---implemented recode's enc_trans(), others use cstocs

    * rearranged converters to work in any order as expected, implemented ERR_*

      (internal) error codes in convert.c

    * implemented -c printing list of all valid converter names

      (get_converters())

    * implemented e_tolower that cannot fail (and appropriate autoconf test)

    * split process_opt() to make less monster from it

    * error messages begin with enca file name without path (strip_path())

    * some preliminary iconv support, the converter doesn't actually exist yet

    * errorneous values of most commandline parametres no longer cause enca to

      abort

    * request cache bubble sorting on use

2000-11-16 Yeti <yeti@physics.muni.cz>

    * fixed destroying outer and request after every use

    * implemented librecode request chache optimization

    * implemented e_unlink()

    * any converter combination/order can be specified on command line with

      -E option (meaning changed!)

    * external converter failure made non-fatal

2000-11-14 Yeti <yeti@physics.muni.cz>

    * bcstocs and brecode return exit status and message is printed when it

      fails

2000-11-05 Yeti <yeti@physics.muni.cz>

    * made iconv configure test, all configure results are cached

2000-10-31 Yeti <yeti@physics.muni.cz>

    * config.h is always included first (even before system headers)

2000-10-30 Yeti <yeti@physics.muni.cz>

    * fixed freeing not allocated request in convert_recode()

    * implemented some more verbosity in converter

    * strstr() alternative moved to common.c

    * implemented strstrcount() conting occurences of needle

    * '/' is added after both in_enc and out_enc when creating recode request

      string, partially helps with touchy librecode, but sometimes it fails

      anyway

    * monstrous convert_recode() split to several functions, nevertheless

      remains monstrous

2000-10-25 Yeti <yeti@physics.muni.cz>

    * more autoconf madness

v0.7.5

2000-10-25 Yeti <yeti@physics.muni.cz>

    * strstr() defined in options.c when not provided by system library

2000-10-22 Yeti <yeti@physics.muni.cz>

    * copying policy changed to GNU/GPL version 2, explicitely, instead of

      version 2 or any later version

    * librecode autoabort feature finally disabled

    * fixed request cache initialization bug

    * fixed O_RDONLY zero/nonzero portability problem

    * convert_recode() restores file from temporary copy when librecode

      converter fails

    * encoding names are not duplicated but assigned in get_encodings()

    * heavy use of const modifier (too many changes to record here)

    * solved cyclic dependancy between langdata and langhook headers by making

      hookdata plain pointer (i.e. void*) and moving T_HookData definition

      to .c file

2000-10-18 Yeti <yeti@physics.muni.cz>

    * removed ffname_[rw]() filename wrapper in cases we cannot get stdin/stdout

      (like e_lseek())

    * stdin/stdout is internally passed as NULL (instead of empty string) so

      one can no longer call enca '' '' to make it wait for several stdin's

    * return values of recode_new_outer() and recode_new_request() are checked

    * implemented librecode request caching, outer is now global and new

      request_cache too, neverhteless caching strategy is poor

    * implemented e_strdup() (strdup()) is not ISO/POSIX) and heavily used

      instead of strlen/malloc/strcpy sequences

2000-10-15 Yeti <yeti@physics.muni.cz>

    * fixed closing stderr in e_close() (though calling it with filedescriptor 2

      would be bug anyway)

    * cleaned types of variables library functions are called with

      (unsigned long int -> size_t, etc.)

    * unsigned long long int is used when available for weight/occurence

      computations, mathint type introduced

    * cleaned names possibly conflicting with POSIX reserved names and libc

      header reserved names

    * recode_scan_request() return value is now checked for success

    * fails back to external converter when librecode cannot convert due to

      errorneous request (as one would understand from man page)

    * prints error message when conversion is impossible

    * merged all file-copying code in convert.c to copy_and_convert(),

      convert_internal_stdin() optimized out of existence

    * return codes in convert.c changed, -1 is now returned as error code only

      by low level i/o from common.c

    * get/generate_xtable() are no longer able to generate identity (not needed)

2000-10-14 Yeti <yeti@physics.muni.cz>

    * cleaned some T_GResult.boxout residua

    * addresses of language filter/hook reports printing functions passed in

      T_GResult

    * some #include <...> corrected to "..." for local files

    * corrected filter reports in SMS mode

2000-10-13 Yeti <yeti@physics.muni.cz>

    * binary filter is more drastic, requires BIN_MIN_TEXT_CHAR letters

      (instead of non-binary characters) to switch back to text mode

v0.7.4

2000-10-11 Yeti <yeti@physics.muni.cz>

    * fixed behaviour in SMS mode (mismatched if([!]Settings.SMSMode); might

      even cause coredumps?)

    * implemented box-drawing character filter filter_boxdraw_out() and put it

      into langdata_cz.c, again with interface lang_filter(); not run in SMS

      mode (printresults.c updated accordingly)

v0.7.3

2000-10-08/09 Yeti <yeti@physics.muni.cz>

    * fixed reading only 4096 bytes from pipe (by a crude way, even if trying to

      lower number of system calls as much as possible) from 0.7.1

    * regenerated data (hope last time) and realized pair/... based guessing

      is nonsense

    * implemented filter that filters out blocks of probably binary data (stdin

      is saved when conversion is required), filter_binary_out() number of

      filtered characters is printed by -d

    * stdin is copied to stdout when its encoding is unknown (much more logical,

      but can break existing scripts)

v0.7.2

2000-10-05 Yeti <yeti@physics.muni.cz>

    * corrected F_EMPTY message (F_EMPTY now really means file is empty)

    * second best is marked in -d output by `-' (added p_esec to T_GResult)

2000-10-04 Yeti <yeti@physics.muni.cz>

    * *_cs filenames corrected to *_cz (Czechoslovakian -> Czech)

    * lang_hook() divided to interface lang_hook() and lang_hook_stz() doing

      the real work (preparation for more langhooks)

    * information about active language hooks saved in active_hooks (for

      printing)

    * perhaps solved language hook info printing dependency (when -d): new

      function print_lang_hook_data() is defined directly in langhook_cz and

      called in print_results

2000-10-02/03 Yeti <yeti@physics.muni.cz>

    * recoding chains made possible by starting -x parameter with `..'

    * BSD [s]random() changed to ISO [s]rand()

    * fixed all remaining warnings except 1.4 kB long string containing help

      text (compile tried with -Werror -Wall -pedantic -ansi

      -Wstrict-prototypes -Wmissing-prototypes -Wmissing-declarations

      -Wnested-externs)

v0.7.1

2000-10-01 Yeti <yeti@physics.muni.cz>

    * moved to low level i/o: e_fopen() changed to e_open(), etc., e_write(),

      etc. introduced, convert_die() abadoned

    * return codes slightly changed to distinguish between guessing failure

      and i/o failure

    * get_xtable() and generate_xtable() changed to be able to generate identity

      (that is then used in convert_internal_stdin())

    * not all i/o failures cause aborting

    * documentation improvements and corrections, as usual

2000-09-30 Yeti <yeti@physics.muni.cz>

    * temporary file creation is tried 3x and existing temporary files are not

      overwirtten

v0.7.0

2000-09-25/26 Yeti <yeti@physics.muni.cz>

    * removed option `disable external converter' (-e), empty -C parameter

      can be used instead

    * e_tmpf() divided to e_tmpf() creating the file and e_tmpfname() generating

      temporary file name

    * fixed memory leak (not freeing tmpfname) in convert_recode() and

      convert_external()

    * man page updated

2000-09-24/25 Yeti <yeti@physics.muni.cz>

    * fixed io_pos == 0 bug when enca converts stdin as dull converter

    * unicode.* renamed to unicodemap.*

    * -x encoding separator changed to .. to be compatible with recode and

      comply RFC 1345 (allowing comma in encoding names)

    * multiple .. allowed in -x argument to make possible specify recode chains

      (but only recode understand them)

    * ffname() divided to ffname_r() and ffname_w(), one returns stdin name one

      stdout name

    * implemented convert_recode() converting via recode library

    * conditionaly added program_name and other stuff requred by librecode

    * == changed to = in bcstocs and brecode to make them work in older bash

    * conversion to the same encoding no longer causes warnings

    * conversion to the same encoding works correctly even for stdin

    * ascii no longer considered unknown encoding

    * introduced verbose option (-V), converter now prints what is doing on

      -V not on -p

    * librecode interface _disabled_ by default in configure.in

2000-09-23/24 Yeti <yeti@physics.muni.cz>

    * two underscores removed from begin (and end) of #defines to comply ISO C

    * guess read buffer and convert read/write buffer merged (the same applies

      to BUFFER_SIZE and Settings.ReadLimit) and this io_buffer made persistent

      (created by new functions enca_init(), destroyed by enca_done()) so

      information from stdin does not get lost

    * user specified read limit is rounded up to nearest multiple of 16