# From arnold@f7.net Wed Apr 27 09:48:37 2005 # Return-Path: # Received: from localhost (skeeve [127.0.0.1]) # by skeeve.com (8.12.11/8.12.11) with ESMTP id j3R6mZVm015791 # for ; Wed, 27 Apr 2005 09:48:37 +0300 # Received: from pop.012.net.il [84.95.5.221] # by localhost with POP3 (fetchmail-6.2.5) # for arnold@localhost (single-drop); Wed, 27 Apr 2005 09:48:37 +0300 (IDT) # Received: from mtain3.012.net.il ([10.220.5.7]) # by i_mss3.012.net.il (HyperSendmail v2004.12) # with ESMTP id <0IFK00L1DIZ02530@i_mss3.012.net.il> for arobbins@012.net.il; # Tue, 26 Apr 2005 22:18:36 +0300 (IDT) # Received: from VScan3 ([10.220.20.3]) # by i_mtain3.012.net.il (HyperSendmail v2004.12) # with ESMTP id <0IFK007U1IZ0U980@i_mtain3.012.net.il> for arobbins@012.net.il # (ORCPT arobbins@012.net.il); Tue, 26 Apr 2005 22:18:36 +0300 (IDT) # Received: from i_mtain1.012.net.il ([10.220.5.1]) # by VScan3 with InterScan Messaging Security Suite; Tue, # 26 Apr 2005 22:15:22 +0300 # Received: from f7.net ([209.61.216.22]) # by i_mtain1.012.net.il (HyperSendmail v2004.12) # with ESMTP id <0IFK009SIIYRN7G0@i_mtain1.012.net.il> for arobbins@012.net.il; # Tue, 26 Apr 2005 22:18:33 +0300 (IDT) # Received: (from arnold@localhost) by f7.net (8.11.7-20030920/8.11.7) # id j3QJFAg18376 for arobbins@012.net.il; Tue, 26 Apr 2005 15:15:10 -0400 # Received: from fencepost.gnu.org (fencepost.gnu.org [199.232.76.164]) # by f7.net (8.11.7-20030920/8.11.7) with ESMTP id j3QJF5J18304 for # ; Tue, 26 Apr 2005 15:15:06 -0400 # Received: from monty-python.gnu.org ([199.232.76.173]) # by fencepost.gnu.org with esmtp (Exim 4.34) # id 1DQVVh-0004gD-CH for bug-gawk@gnu.org; Tue, 26 Apr 2005 15:14:17 -0400 # Received: from Debian-exim by monty-python.gnu.org with spam-scanned # (Exim 4.34) id 1DQVYa-0002PR-2b for bug-gawk@gnu.org; Tue, # 26 Apr 2005 15:17:56 -0400 # Received: from [129.183.4.8] (helo=ecfrec.frec.bull.fr) # by monty-python.gnu.org with esmtp (Exim 4.34) # id 1DQVYZ-0002Lr-EF for bug-gawk@gnu.org; Tue, 26 Apr 2005 15:17:15 -0400 # Received: from localhost (localhost [127.0.0.1]) # by ecfrec.frec.bull.fr (Postfix) with ESMTP id 5782819D907 for # ; Tue, 26 Apr 2005 21:12:53 +0200 (CEST) # Received: from ecfrec.frec.bull.fr ([127.0.0.1]) # by localhost (ecfrec.frec.bull.fr [127.0.0.1]) (amavisd-new, port 10024) # with ESMTP id 06763-10 for ; Tue, # 26 Apr 2005 21:12:51 +0200 (CEST) # Received: from ecn002.frec.bull.fr (ecn002.frec.bull.fr [129.183.4.6]) # by ecfrec.frec.bull.fr (Postfix) with ESMTP id 4488B19D906 for # ; Tue, 26 Apr 2005 21:12:51 +0200 (CEST) # Received: from daphne ([129.183.192.6]) # by ecn002.frec.bull.fr (Lotus Domino Release 5.0.12) # with ESMTP id 2005042621231613:3312 ; Tue, 26 Apr 2005 21:23:16 +0200 # Date: Tue, 26 Apr 2005 21:12:49 +0200 (CEST) # From: Jean-Marc Saffroy # Subject: GNU awk unable to handle 64-bit ints on IA64 # X-X-Sender: saffroyj@daphne.frec.bull.fr # To: bug-gawk@gnu.org # Message-id: # MIME-version: 1.0 # Content-type: TEXT/PLAIN; charset=US-ASCII; format=flowed # X-MIMETrack: Itemize by SMTP Server on ECN002/FR/BULL(Release 5.0.12 |February # 13, 2003) at 26/04/2005 21:23:16, # Serialize by Router on ECN002/FR/BULL(Release 5.0.12 |February 13, # 2003) at 26/04/2005 21:23:16, Serialize complete at 26/04/2005 21:23:16 # X-Virus-Scanned: by amavisd-new at frec.bull.fr # Original-recipient: rfc822;arobbins@012.net.il # X-Spam-Checker-Version: SpamAssassin 2.63 (2004-01-11) on skeeve.com # X-Spam-Level: # X-Spam-Status: No, hits=-4.9 required=5.0 tests=BAYES_00 autolearn=ham # version=2.63 # Status: RO # # # Hello, # # I have rounding problems when manipulating 64-bit ints (actually they are # addresses) on Linux/IA64: # # $ echo 0xa000000100000813|./gawk '{printf("0x%lx\n",strtonum($1));}' # 0xa000000100000800 # $ echo 0xffffffffffffffff|./gawk '{printf("0x%lx\n",strtonum($1));}' # 0x8000000000000000 # $ ./gawk --version|head -1 # GNU Awk 3.1.4 # # The problem seems to be that AWKNUM is defined to be a double, which has a # 53-bit mantissa. On IA64 with gcc 3.2.3 (maybe other compilers as well) # there is a long double type with a larger mantissa: # # $ grep define.*LDBL_MANT_DIG /usr/lib/gcc-lib/ia64-redhat-linux/3.2.3/include/float.h # #define LDBL_MANT_DIG 64 # # So I changed AWKNUM to be a long double; this does not seem to be # sufficient, because of some dubious casts to double (there may be others # left, I didn't check), see patch below. Now it's much nicer: # # $ echo 0xa000000100000813|./gawk '{printf("0x%lx\n",strtonum($1));}' # 0xa000000100000813 # $ echo 0xffffffffffffffff|./gawk '{printf("0x%lx\n",strtonum($1));}' # 0xffffffffffffffff # # Maybe the gawk configure script should set AWKNUM to be a long double on # Linux/IA64? # # # Regards, # # -- # Jean-Marc Saffroy - jean-marc.saffroy@ext.bull.net # # # diff -ru gawk-3.1.4/awk.h gawk/awk.h # --- gawk-3.1.4/awk.h 2004-07-26 16:11:05.000000000 +0200 # +++ gawk/awk.h 2005-04-26 19:19:10.545419273 +0200 # @@ -273,7 +273,7 @@ # /* ------------------ Constants, Structures, Typedefs ------------------ */ # # #ifndef AWKNUM # -#define AWKNUM double # +#define AWKNUM long double # #endif # # #ifndef TRUE # diff -ru gawk-3.1.4/builtin.c gawk/builtin.c # --- gawk-3.1.4/builtin.c 2004-07-13 09:55:28.000000000 +0200 # +++ gawk/builtin.c 2005-04-26 20:53:41.211365432 +0200 # @@ -578,7 +578,7 @@ # char *cend = &cpbuf[30];/* chars, we lose, but seems unlikely */ # char *cp; # const char *fill; # - double tmpval; # + AWKNUM tmpval; # char signchar = FALSE; # size_t len; # int zero_flag = FALSE; # @@ -2773,16 +2773,16 @@ # do_strtonum(NODE *tree) # { # NODE *tmp; # - double d; # + AWKNUM d; # # tmp = tree_eval(tree->lnode); # # if ((tmp->flags & (NUMBER|NUMCUR)) != 0) # - d = (double) force_number(tmp); # + d = (AWKNUM) force_number(tmp); # else if (isnondecimal(tmp->stptr)) # d = nondec2awknum(tmp->stptr, tmp->stlen); # else # - d = (double) force_number(tmp); # + d = (AWKNUM) force_number(tmp); # # free_temp(tmp); # return tmp_number((AWKNUM) d); # # # ##################################################################################### # This Mail Was Scanned by 012.net Anti Virus Service - Powered by TrendMicro Interscan # { printf("0x%lx\n",strtonum($1)); }