Tue Jul 26 21:46:16 2005 Arnold D. Robbins * Release 3.1.5: Release tar file made. Tue Jul 26 21:44:54 2005 Arnold D. Robbins Copyright dates on all relevant files updated to 2005. Wed Jul 6 17:09:02 2005 Arnold D. Robbins Minor cleanups: * io.c (do_index): Remove unused variables `mbclen', `mbs1' and `mbs2'. * node.c (wstrstr): Remove unsed variable `j'. (dump_wstr): `#ifdef' out, not currently needed. * eval.c (op_assign): Move decl of `t1' and `t2' into a separate block for the `! HAVE_FMOD' case. Keeps the compiler quiet. Similar for `ltemp'. Wed Jul 6 16:51:31 2005 Arnold D. Robbins * io.c (close_io): Now takes pointer to boolean parameter indicating if there was a problem closing standard output or standard error. Update it in the right places. * awk.h (close_io): Update the declaration. * main.c (main): New variable `stdio_problem'. Pass it to `close_io'. Check the result and exit non-zero if there was a problem. (usage, version): Print warning message if problems with stdout. Unrelated: * main.c (main): For call to `setlocale' for LC_MESSAGES, just use `#ifdef LC_MESSAGES'. Per Bruno Haible . Wed Jul 6 16:44:58 2005 Jim Meyering * main.c (init_fds): If any of the STDIN_FILENO, STDOUT_FILENO, STDERR_FILENO are initially closed, reopen them with permissions contrary to common usage so that any reasonable attempt to use them will evoke the same sort of error as reading or writing to a closed file descriptor would. Mon Jul 4 09:38:29 2005 Arnold D. Robbins More multibyte fixes from Kimura Koichi, . * node.c (format_val, r_dupnode): Spell `wstptr' correctly. * regex_internal.c (build_wcs_upper_buffer): Label `offsets_needed' should not be inside `#ifdef _LIBC'. * regcomp.c (build_charclass): Fix declaration of `class_name' in prototype to not be unsigned. Thu Jun 30 11:52:34 2005 Arnold D. Robbins * profile.c (tree_eval): Node_not: Remember to print the exclamation point! Thanks to Dan Nielsen for the bug report. * mbsupport.h: Fix spelling of HAVE_ISWUPPER. Thanks to Kimura Koichi, . Sun Jun 26 16:37:59 2005 Arnold D. Robbins Unrelated changes: * builtin.c (do_length): Allow array argument to length(). Returns number of elements in array. * awkgram.y (yylex): Ignore carriage returns in source code. Sigh. Wed Jun 15 22:12:15 2005 Arnold D. Robbins * node.c (isnondecimal): Check loc.decimal_point before using it. Avoids problems with command line assignment when locale info may not be set up all the way yet. Wed Jun 15 21:59:54 2005 Arnold D. Robbins * node.c (make_str_node): If working with multibyte characters, while parsing string constants, keep multibyte characters together. This avoids problems in cases where one of the bytes is backslash. Initial patch supplied by Kimura Koichi, . Tue Jun 14 21:50:37 2005 Andrew J. Schorr Use Exponentiation By Squaring for integer powers for ^ and ^=. * eval.c (calc_exp, cal_exp_posint): New functions. (r_tree_eval): Use them. Fri Jun 3 12:15:54 2005 Arnold D. Robbins * configure.ac: Further change the hack at the end so that it works on Mac OS X `sed'. Sigh. Thu Jun 2 22:44:01 2005 Arnold D. Robbins * configure.ac (TYPE_SOCKLEN_T): Use `int' as default type if can't figure one out. * awkgram.y: Warn that `//' is not a C++ comment. (:-) Thu Jun 2 20:55:27 2005 Arnold D. Robbins From: Benno Schulenberg * eval.c (func_call): Take message out of gettext call since it's for debugging. * ext.c (get_actual_argument): Fix formatting of message. Wed May 25 09:19:37 2005 Arnold D. Robbins * configure.ac: Change hack at end that fixes Makefile to keep version.c to use `sed' and not `ed'. More portable to OS/2, probably other systems. Mon May 23 09:01:26 2005 Arnold D. Robbins Portability help from Jim Meyering. * io.c: Rework ifdefs for and . Test them individually. * configure.ac: Add AC_C_RESTRICT and code for socklen_t from rsync. Check for isascii and btowc for regex. Sat May 14 22:49:54 2005 Arnold D. Robbins * wait_any (errno): Remove decl. * gawk_popen: The pipe-simulated but not VMS or DOS version. Remove decl/use of `strdup' in favor of `emalloc' and `strcpy'. Wed May 11 18:33:30 2005 Arnold D. Robbins All files: Updated address of FSF to: 51 Franklin Street, Fifth Floor Cambridge, MA 02110-1301 Wed May 11 18:19:03 2005 Jim Meyering * configure.ac: Use AM_GNU_GETTEXT([external]). Reflect upgrade to gettext-0.14.4. Reflect renaming of `jm_'-prefixed macros. (AC_CONFIG_FILES): Remove intl/Makefile. * Makefile.am (SUBDIRS): Remove intl. (AM_CPPFLAGS): Remove -Iintl. Wed May 11 11:42:06 2005 Arnold D. Robbins Straighten out mess with `isblank' which is C99 function. * configure.ac: Remove check for `isblank' in call to AC_CHECK_FUNCS. * regex_internal.h: #ifdef out definition of `isblank' and provide `is_blank' function a la dfa.c. * field.c: Ditto. * regcomp.c: #ifdef use of `isblank' and add `is_blank' use instead. Mon May 9 08:29:37 2005 Arnold D. Robbins * configure.ac: Add type check for `socklen_t', fixes compile warning on AMD/64 Linux. * io.c (socketopen): Change type of socket lenght variables to `socklen_t' from `size_t'. Thu May 5 22:00:03 2005 John E. Haque * io.c (iop_alloc): Let a input processor hook installed via `register_open_hooks' open its own fd in case gawk does not know how to open it. (iop_open): Call `os_close_on_exec' after `iop_alloc'. ADR: If `iop_alloc' returns NULL but the fd is valid, close the fd to avoid an fd leak. Mon May 2 08:05:59 2005 Arnold D. Robbins * eval.c (update_ERRNO): Don't use `return' in a `void' function. * awk.h (AWKNUM): Back out use of `long double' based on LDBL_MANT_DIG. * builtin.c (tmp_integer): Back out extra ifdefs. Fri Apr 29 13:01:05 2005 Arnold D. Robbins * configure.ac: Look for `isblank' function. * field.c: Add define for `isblank' if we don't have it. Fri Apr 29 12:01:33 2005 Julian Foad From grep. Doesn't seem to affect awk. * dfa.c (lex): Fix bug #9519: "echo do^re | grep do^re" was failing to find a match. [Towards end, set `lasttok' before returning `c'.] Fri Apr 29 00:28:46 2005 Arnold D. Robbins * configure.ac: Jump through an amazingly convoluted hoop to get config.status to keep version.c upon `make distclean'. Seems to work though. Thu Apr 28 23:40:02 2005 Stepan Kasal * configure.ac (PRINTF_HAS_F_FORMAT): Some cosmetic changes. (custom.h): Don't cat custom.h at the end of config.h; instead, use AH_BOTTOM([#include "custom.h"]) * awklib/Makefile.am (AM_CPPFLAGS): Add $(top_srcdir) so that custom.h can be found. Thu Apr 28 23:21:22 2005 Jim Meyering * field.c (set_FIELDWIDTHS): Tighten up the code to accept FIELDWIDTHS values in [1..INT_MAX], e.g., detect overflow and invalid strings, and reject strings starting with `-'. Thu Apr 28 23:05:33 2005 Stepan Kasal * dfa.c (parse_bracket_exp_mb): Shorten one part of the code, to get closer to grep's copy. Thu Apr 28 23:00:58 2005 Pat Rankin * builtin.c (format_tree) [#if VAXCRTL]: For floating point formatting, reject zero_flag if using old VAXCRTL run-time library to avoid getting erroneous results which appear as if numerically incorrect (due to an embedded space in some cases, extra trailing zeroes in others) rather than just misformatted. `hsprint' test still fails, but not as badly. Thu Apr 28 19:12:03 2005 Arnold D. Robbins * re.c (reflags2str): Add three new RE_ flags from current regex.h to bring the table up to date. * builtin.c (format_tree): Save 2 spare bytes instead of one. Suggested by Stepan Kasal. Thu Apr 28 18:16:09 2005 Andrew J. Schorr * awk.h (IOBUF): Add new fields `opaque', `get_record', and `close_func', to support insertion of an alternate input processor. This is used by the XML extension. (register_deferred_variable, register_open_hook, update_ERRNO_saved): Declare new functions. (load_environ, load_procinfo): Remove declarations -- these functions are no longer global, since we use register_deferred_variable instead. * awkgram.y (register_deferred_variable): New function to allow calling code to register special variable names that trigger a callback upon the first reference. This is now used to implement ENVIRON and PROCINFO. (variable): Search the list of deferred variables instead of hardcoded tests for ENVIRON and PROCINFO. * eval.c (set_BINMODE): Fix spelling of "arbitrary" in warning message. (update_ERRNO_saved): New function that allows the caller to specify the errno value instead of using the current value. (update_ERRNO): Implement by calling update_ERRNO_saved(errno). * io.c (iop_close): Call `iop->close_func' if non-NULL. (close_redir): Should save `errno' value, otherwise `lintwarn' messages might update it. Then use `update_ERRNO_saved' to set ERRNO. (do_getline): Call `update_ERRNO_saved' to set ERRNO based on the error code returned by the redirect function (instead of the current value of errno). Similarly, use `update_ERRNO_saved' to set ERRNO based on the value returned by `get_a_record'. But add a special check to avoid updating ERRNO if `get_a_record' returns an error code value of -1 (this is used by the XML extension which already sets ERRNO before returning). (register_open_hook): New function to register a function to be called whenever a new data file is opened. This can be used to install a special input processor (as in the XML extension). (iop_alloc): Call registered open hook. (get_a_record): If a `get_record' method has been set, call that instead. * main.c (init_vars): Use `register_deferred_variable' to implement ENVIRON and PROCINFO. (load_environ, load_procinfo): Now static instead of global. * doc/gawk.texi: Document new internal functions `update_ERRNO_saved', `register_deferred_variable', and `register_open_hook'. Thu Apr 28 10:50:10 2005 Arnold D. Robbins * array.c (assoc_find, do_delete): Change incorrect uses of STREQN to memcmp. * builtin.c (do_index): Same. * field.c (set_FS): Same. * io.c (redirect, getredirect, do_close, set_RS): Same. * re.c (reisstring): Same. Wed Apr 27 21:35:57 2005 Arnold D. Robbins Allow for long double. Initial changes from Jean-Marc Saffroy . * awk.h (AWKNUM): If have long doubles (LDBL_MANT_DIG), define AWKNUM as long double, otherwise just use double. * builtin.c (format_tree): Change type of tmpval to double. (do_strtonum): Same for `d' and types used in casts. (tmp_integer): Don't do bit shifting if have long doubles. Unrelated, from Andrew J. Schorr: * io.c (close_one): Check for RED_FILE|RED_WRITE, not just RED_FILE. Mon Apr 25 12:23:18 2005 Andrew J. Schorr * eval.c (r_tree_eval): In Node_assign_concat case, when copying string constants, include the terminating zero byte. Fri Apr 1 06:26:31 2005 Arnold D. Robbins Update to Automake 1.9.5. * INSTALL, aclocal.m4, depcomp, install-sh, missing, mkinstalldirs, ylwrap: Updated. Unrelated: * builtin.c (do_tolower, do_toupper): Remove old code based on 8-bit character table. Wed Feb 23 08:23:22 2005 Arnold D. Robbins * bisonfix.awk: New file, fixes continued #ifdef for dumb compilers. * Makefile.am (awkgram.c): Fix rule to use it. (EXTRA_DIST): Include bisonfix.awk. Tue Feb 22 21:18:50 2005 Arnold D. Robbins * random.h: Remove include of config.h and move it to ... * random.c: Here. Move include of random.h back to where it was. * regcomp.c, regex.c, regexec.c: NUKED all use of alloca not inside `_LIBC' ifdef. Hooray! Sat Feb 19 20:13:28 2005 Pat Rankin Workarounds for bugs and missing C89 features in old VAX C compiler. * regex_internal.h "mbsupport.h": Suppress inclusion if NO_MBSUPPORT is defined. [MB_CUR_MAX]: Define as 1 if mbsupport.h hasn't defined it. [ER_ERRMSG, ERRMSG_TYPE, ERRMSG_OFFSET, ERRMSG_SEPARATOR]: New macros conditionalized upon gawk's NO_TOKEN_PASTING macro. * regcomp.c: Use them. (parse_dup_op): Use alternate initialization of start_token if RE_TOKEN_INIT_BUG is defined. * regexec.c (proceed_next_node): Compare push_fail_stack() result explicitly against REG_NOERROR rather than implicitly against 0. Sat Feb 19 20:05:50 2005 Pat Rankin * dfa.c "mbsupport.h": Suppress inclusion if NO_MBSUPPORT is defined. Wed Feb 16 20:43:07 2005 Pat Rankin * awk.h "mbsupport.h": Suppress inclusion if NO_MBSUPPORT is defined. * regex.h : Guard inclusion with HAVE_SYS_TYPES_H. * regex.c : Likewise. * random.c "random.h": include this first to get config.h setup. : Guard inclusion with HAVE_FCNTL_H. : Guard inclusion with HAVE_UNISTD_H. * io.c [#if defined(MSDOS) ||... defined(__CYGWIN__)]: Splice the backslash continuation back into one long line. Wed Feb 16 10:11:21 2005 Arnold D. Robbins * node.c (unref, format_val): Add assertions checking that both `tmp->wstptr != NULL' and `(tmp->flags & WSTRCUR) != 0' before freeing `tmp->wstptr'. Thanks to kimura.koichi@canon.co.jp. * random.c (HAVE_UNISTD_H): Conditionalize include of . Thanks to Scott Deifik . Sun Feb 13 18:24:50 2005 Arnold D. Robbins * io.c (socketopen): Move `#ifdef MSG_PEEK' up to above declarations too. Thanks to Michal Jaegermann. * config.guess, config.sub: Updated from Savannah. Thu Feb 10 15:48:48 2005 Arnold D. Robbins * regex_internal.c (re_dfa_add_node): Remove variable `type' and just use `token.type' directly in RE_ENABLE_I18N code below. Saves a compiler warning, and a good compiler will handle it anyway. * regexec.c (check_arrival_add_next_nodes): Move decl of `err' inside #ifdef RE_ENABLE_I18N code where it's used. * awkgram.y (yylex): Add casts to int before use of `strlen' results for printf-style precision. Avoid a compiler warning. * io.c (redirect, do_close): Same for use of tmp->stlen. Thanks to Michal Jaegermann . Wed Feb 9 10:19:15 2005 Stepan Kasal * Makefile.am (datadir, libexecdir): Removed. (awkdatadir): Renamed to pkgdatadir. (pkgdatadir, LDADD): Use the make syntax to refer to other variables, not @...@. Wed Feb 9 10:05:46 2005 Arnold D. Robbins * node.c (r_dupnode): Assign NULL to r->wstr after `getnode'. Fix count of bytes to copy in call to `memcpy'. Thanks to Kimura Koichi, . Tue Feb 8 19:26:22 2005 Pat Rankin * regcomp.c (init_dfa): Avoid strcasecmp() since regex.c doesn't use awk.h and none of the assorted other included header files are guaranteed to declare it. (parse_expression): Modify casts for the string arguments passed to build_charclass_op() to fix char * vs unsigned char * mismatch. (parse_bracket_exp): Likewise add cast for the string argument passed to build_charclass(). Mon Feb 7 15:04:09 2005 Arnold D. Robbins * eval.c (make_scalar): Don't use P() macro in definition. Thanks to Juergen Kahrs . Wed Feb 2 16:36:19 2005 Arnold D. Robbins * main.c (main): Call `close_io', without its result affecting the exit status. Super small, super dark corner. See test/exitval2.awk. Tue Feb 1 11:58:29 2005 Arnold D. Robbins * regex.h (__APPLE_CC__): Removed test and definition of __restrict. Not needed for current MacOS X compiler. Sun Jan 30 13:56:37 2005 Arnold D. Robbins Fresh merge with CVS regex routines. Fixes handling of \B. See tests/gnureop3.awk and also http://sources.redhat.com/bugzilla/show_bug.cgi?id=693. CVS base versions: * regcomp.c: Version 1.92, Thu Jan 27 19:05:20 2005. * regexec.c: Version 1.77, Thu Jan 27 19:06:34 2005. * regex_internal.c: Version 1.49, Thu Jan 27 19:07:15 2005. * regex_internal.h: Version 1.60, Wed Jan 26 22:40:50 2005. * regexec.c: Version 1.77, Thu Jan 27 19:06:34 2005. Sat Jan 22 22:30:40 2005 Arnold D. Robbins Reinstate patch of 18 Nov 2001, for VMS, at least: * random.c (srandomdev): ifdef-out. Lots of compile time problems on multiple platforms, and gawk doesn't even use the routine. The heck with fine-grained solutions. Thu Jan 20 14:15:32 2005 Arnold D. Robbins * awkgram.y (LEX_FOR): Free NAME tokens in transformation of `for (iggy in foo) delete foo[iggy]' into `delete foo'. Thanks and a tip of the hatlo to Valgrind. * dfa.c (_): Clean up stuff here by just including "gettext.h". Per Bruno Haible. Wed Jan 19 18:29:23 2005 Arnold D. Robbins * awkgram.y (yylex): Improve parsing of numeric constants and hex values, via a push from Paul Eggert. See test/hex.awk. * regex_internal.c (re_node_set_alloc): If `size' is 0, just zero out the structure. From valgrind. Tue Jan 18 17:23:25 2005 Arnold D. Robbins Make gawk multibyte aware. This means that index(), length(), substr() and match() all work in terms of characters, not bytes. * awk.h (NODE): Add `wsp' and `wslen' elements to value for wide string. (WSTRCUR, wstptr, wstlen, force_wstring): New macros. (str2wstr, wstrstr, wcasestrstr): New declarations. * builtin.c (do_index, do_length, do_substr, do_match): Handle wide strings. * eval.c (flags2str): Add WSTRCUR. * node.c (format_val, r_dupnode, mk_number, make_str_node, unref): Add code to deal with wide strings. (str2wstr, dump_wstr, wstrstr, wcasestrstr): New functions. Sun Jan 16 15:10:35 2005 Arnold D. Robbins * random.h (int32_t): Define this type. Thu Jan 13 14:38:13 2005 Arnold D. Robbins Cause `configure --disable-nls' to still allow locale-correct formating of numeric values. * builtin.c (format_tree): Change #ifdefs to only test HAVE_LOCALE_H. Improve code for ' flag so that extraneous separator is not included if number of digits is multiple of locale separater count (3, 6, 9, etc.) * dfa.c (dfaparse): Change ifdef to only test #ifdef LC_COLLATE. * eval.c (fmt_ok): Remove ENABLE_NLS from #ifdef test. * gettext.h: Include on both sides of test. Should really be factored out. * main.c (loc): Remove ENABLE_NLS from #ifdef test. (main): Same in call to localeconv(). * node.c (isnondecimal): Remove ENABLE_NLS from #ifdef test. Unrelated: * regcomp.c (init_dfa): Change `codeset' to `codeset_name' in two places. Mon Jan 10 11:49:56 2005 Arnold D. Robbins Annual sync with glibc. * getopt_int.h: New file. * Makefile.am (base_sources): Add it. * getopt.h, getopt.c, getopt1.c: Updated. * regcomp.c, regex.c, regex.h, regex_internal.c, regex_internal.h, regexec.c: Updated. Original versions: getopt_int.h, 1.1, Tue Mar 9 10:31:19 2004 getopt1.c, 1.10, Tue Mar 9 10:35:37 2004 getopt.h, 1.21, Fri Mar 19 00:19:32 2004 getopt.c, 1.53, Wed Mar 10 23:13:26 2004 regcomp.c, 1.87, Mon Dec 6 02:56:42 2004 regex.c, 1.126, Fri Jan 30 05:19:58 2004 regex.h, 1.33, Thu Nov 18 23:50:57 2004 regex_internal.c, 1.46, Thu Jan 6 20:59:49 2005 regex_internal.h, 1.57, Mon Dec 27 16:29:05 2004 regexec.c, 1.75, Mon Dec 27 16:29:52 2004 * regex.h: Add check for __APPLE_CC__ and definition of __restrict. * regex.c: Add check for _MSC_VER and include . * regex_internal.h (_RE_ENABLE_I18N): Change test. (re_realloc): Add check/fix for SunOS 4.1.x. * regex_internal.c (build_wcs_upper_buffer): ifdef label `offsets_needed', add cast in call to `wcrtomb'. * regcomp.c (build_charclass, build_charclass_op): Remove `unsigned' from declarations of `char *' params. (regerror): Remove use of mempcpy. (peek_token): Disallow \s and \S for gawk. (build_charclass): Change decl of `class_name' and use it directly. Nuke variable `name'. (build_charclass_op): Change decl of `class_name' and `extra'. Thu Jan 6 16:44:32 2005 Arnold D. Robbins Improve autoconfiscation stuff for wide character use. * builtin.c (do_tolower, do_toupper): Conditionally compile call to `wide_tolower_toupper'. (wide_tolower_toupper): Conditionally compile typedefs and function. * mbsupport.h: Add check for having `wint_t', and `iswlower', `iswupper', `towlower' and `towupper'. * configure.ac (HAVE_WINT_T): Add test. (AC_CHECK_FUNCS): Add `wint_t', `iswlower', `iswupper', `towlower' and `towupper'. Unrelated change: * hard-locale.h (hard_locale): Add decl of `xmalloc' to prevent redeclaration problems on some compilers. Wed Jan 5 10:20:17 2005 Arnold D. Robbins Update to Bison 2.0. * bisonfix.sed: Removed, no longer needed. * Makefile.am (EXTRA_DIST): Removed bisonfix.sed. (awkgram.c): Fix build rule. * awkgram.c: Regenerated. Tue Jan 4 18:47:56 2005 Arnold D. Robbins Update to Automake 1.9.4. * alocal.m4, config.guess, config.sub, install-sh: Updated. Mon Jan 3 14:08:27 2005 Arnold D. Robbins Update to Automake 1.9.3. * INSTALL, alocal.m4, config.guess, config.sub, depcomp, install-sh, missing, ylwrap: Updated. Mon Jan 3 11:23:36 2005 Arnold D. Robbins Fix obscure issue. ^ in RS should only match at the very beginning of the input. Essentially, the file is one long string. To do this, use the `not_bol' flag in the `struct pattern_buffer'. Thanks to Stepan Kasal for pointing out the problem and to Andreas Schwab for pointing out the mechanism for a solution. * awk.h (RE_NEED_START, RE_NO_BOL): New flags for `research'. (IOP_AT_START): New flag for IOBUF. (research): Last parameter is now `flags'. * builtin.c (do_match, sub_common): Change calls to `research'. * eval.c (interpret, match_op): Same. * field.c (re_parse_field): Same. * io.c (spec_setup): Add IOP_AT_START flag. (iop_alloc): Same. (rsrescan): Modify logic to check IOP_AT_START and if not on to add RE_NO_BOL to flags value in call to `research'. (get_a_record): Clear IOP_AT_START upon return from `*matchrec'. (iopflags2str): Add IOP_AT_START to table. Also IOP_CLOSED, which was missing. (Ooops.) * re.c (research): Last paramater is now flags. Modify logic to handle RE_NO_BOL case by setting the right bit initially. Clean up control flow so that it's cleared before returning. If RE_NO_BOL, don't bother with the dfa matcher, as it doesn't have an analogous capability. Wed Dec 22 12:33:48 2004 Arnold D. Robbins For --exec, don't allow x=y assignments where filenames would be. Do allow -v. This is because we assume --exec is used mainly for CGI stuff and we don't want var assigns to affect the code. Suggested by Stepan Kasal; motivated by reading about web security. * main.c (disallow_var_assigns): New variable. (main): Set the var for --exec. (arg_assign): Check it appropriately. Sun Dec 19 17:27:09 2004 Arnold D. Robbins * builtin.c (do_gensub): Make `global' flag smarter, such that a string numeric constant (e.g., "3") acts like a numeric constant. * node.c (r_force_number): Not really related: Only set NUMCUR if we actually convert some digits. Sun Dec 19 16:08:50 2004 Arnold D. Robbins * dfa.h, dfa.c: Synchronize with what's happening in GNU grep development. Effectively only minor whitespace changes and some slight code motion of ifdefs and includes. * hard-locale.h: New file, extracted from old dfa.c. * Makefile.am (base_sources): Add hard-locale.h. Sun Dec 19 11:13:45 2004 Arnold D. Robbins * io.c (socketopen): Change type of `readle' and `namelen' variables to size_t. For QNX, but a good idea anyway. Thanks to `Anthony' (rz1a@mail.ru). Mon Dec 6 11:11:22 2004 Arnold D. Robbins Undid change of Mar 9 2004, to add gofast patch. It gets things wrong for gawk. This removes the bandaid of ifdef-ing out the main check. Eventually this'll all get straightened out in the GNU grep code. * dfa.c (buf_offset): Removed. (SKIP_REMAINS_MB_IF_INITIAL_STATE): Removed use of buf_offset, do free `mblen_buf', `inputwcs'. (match_anychar, match_mb_charset, transit_state_consume_1char, transit_state): Remove use of buf_offset in mblen_buf. (dfaexec): Use `free' and `malloc', not `realloc'. Mon Dec 6 10:55:37 2004 Fumitoshi UKAI Forwarded from james@nocrew.org, the Debian contact. * dfa.c (parse_bracket_exp_mb): 1. Build range correctly when IGNORECASE for [a-a] to also get 'A'. 2. For [:lower:] and [:upper:], if ignoring case, set type string to "alpha". This parallels code in the regex routines. 3. Reset wc1 to EOF when parsing bracket expressions. Mon Nov 29 18:36:25 2004 Arnold D. Robbins * io.c (redirect): When allocating a new struct redirect, set rp->pid to -1, not 0, so that code checking for EOF on an input pipe works correctly. Thu Nov 25 14:22:41 2004 Stepan Kasal * Makefile.am (MAINTAINERCLEANFILES): Add. * version.in (version_string): Use PACKAGE_STRING. Tue Nov 23 17:27:38 2004 Stepan Kasal * re.c: Fix a typo in a comment. Mon Nov 22 16:47:00 2004 Arnold D. Robbins * awkgram.y (yylex): Add lint check for tawk style modifiers on regexes, /.../i and /.../s. Not that it'll help anyone. Wed Oct 27 14:25:18 2004 Stepan Kasal * builtin.c (do_tolower, do_toupper): Fix the wide char handling, especially when the lowercased char doesn't ocuppy the same number of bytes as its uppercase equivalent. Make use of ... (wide_tolower_toupper): ... this new static function. Mon Oct 25 11:51:14 2004 Arnold D. Robbins * builtin.c (bchunk, bchunk_one, cksize): Change type of `olen' to size_t from long. It is the 21st century now, after all... Thanks to Stepan Kasal. Mon Oct 11 10:49:09 2004 Arnold D. Robbins * awkgram.y (yylex): Improve lint warnings for non-decimal constants. * node.c (isnondecimal): Made a little smarter, thanks to Stepan Kasal. Thu Oct 7 21:59:38 2004 Arnold D. Robbins * dfa.c (dfamust): Redo fix of 22 Sep to match code from current GNU grep. Sun Oct 3 23:06:00 2004 Arnold D. Robbins * node.c (isnondecimal): Made smarter, so that 0xEE does register as non-decimal. Added parameter to indicate use of locale's decimal point and changed declaration and callers. Tue Sep 28 18:38:17 2004 Arnold D. Robbins * node.c (isnondecimal): New function, now smarter. * awk.h (isnondecimal): Changed from macro to function. Wed Sep 22 11:24:46 2004 Arnold D. Robbins * dfa.c (dfamust): At end, check results of `malloc'. Based on bug report from Sorav Bansal for grep. Mon Sep 20 13:18:18 2004 Arnold D. Robbins New --exec option. Needed for gawk CGI scripts to prevent arbitrary options and/or source getting passed in from the web. * main.c (optab): New long option, --exec. (main): Catch it. Like -f but end option processing. (usage): Add it to the usage message. Thanks to John DuBois and Don Stokes for their input. Unrelated: * dfa.c (dfaexec): Disabled caching into buffer that bypasses multibyte initialization, since it can get things wrong. Thanks to Andreas Schwab . Mon Sep 20 12:59:42 2004 Andreas Schwab * awkgram.y (nextc): Check for end of lexer buffer before advancing ring buffer index. Wed Sep 8 09:54:53 2004 Arnold D. Robbins * main.c (main): Force LC_NUMERIC locale to "C" before parsing the program, since a variable assignment with -v can leave the locale set incorrectly. Thanks to Sirix for reporting the problem. Wed Aug 25 18:55:30 2004 Arnold D. Robbins * main.c (UPDATE_YEAR): New constant at top of file, where we won't miss it. (copyleft): Use it. Sun Aug 22 17:26:39 2004 Stepan Kasal Define gawk_mb_cur_max even if there is no mbs support, as ``const int'' and assign 1 to it. This fixes a bug in re.c where #ifdef MBS_SUPPORT was missing. * awk.h (gawk_mb_cur_max): Declare. * main.c (gawk_mb_cur_max): Define. * awkgram.y (nextc_is_1stbyte): Without mbs support, define to 1. * builtin.c (index_multibyte_buffer): Define a dummy function when there is no mbs support. * awkgram.y, builtin.c, re.c: Remove some `#ifdef MBS_SUPPORT'. Sun Aug 15 22:08:04 2004 Arnold D. Robbins Import current FreeBSD random.c. Make it work for gawk. Needed for cases where long is more than 32 bits. * random.c: Imported from FreeBSD. Header includes tweaked. * random.h: Typdef gawk_uint32_t appropriately and #define uint32_t to it. * configure.ac: Add calls to AC_CHECK_SIZEOF for unsigned int and unsigned long. Started with http://www.freebsd.org/cgi/cvsweb.cgi/src/lib/libc/stdlib/random.c Thanks to Andreas Schwab for the pointer. Thu Aug 12 13:09:53 2004 Arnold D. Robbins * main.c (copyleft): Fix copyright year. Mon Aug 2 12:18:15 2004 Arnold D. Robbins * Release 3.1.4: Release tar file made. Mon Aug 2 12:17:40 2004 Arnold D. Robbins * configure.ac: Fix the hack. Do the sed on `Makefile', not `Makefile.in'. Sigh. Sun Aug 1 14:48:30 2004 Arnold D. Robbins * configure.ac: MAJOR HACK: At end, remove version.c from Makefile.in variable `CONFIG_CLEAN_FILES' so that `make distclean' doesn't remove version.c. Mon Jul 19 17:07:27 2004 Arnold D. Robbins * config.guess, config.sub: Updated from Savannah CVS. Fri Jul 16 10:59:07 2004 Arnold D. Robbins * re.c (make_regexp): Bracket code using `gawk_mb_cur_max' inside `#ifdef MBS_SUPPORT'. Thu Jul 15 12:36:25 2004 Arnold D. Robbins * dfa.c (parse_bracket_exp_mb): If doing case folding, include the other case for regular characters inside [...]. * re.c (make_regexp): Smarten up handling of IGNORECASE, particularly for multibyte character sets. Sigh. Wed Jul 14 16:25:23 2004 John Haque * eval.c (interpret): For `Node_K_return', use `copynode' and not `dupnode' for non-PERM, non-TEMP values. (func_call): Don't add TEMP flag to returned value. These two fix a problem uncovered by the July 8 change in `assoc_lookup'. Wed Jul 14 16:14:09 2004 Arnold D. Robbins * awkgram.y (node_common): Add check `lexeme >= lexptr_begin', from valgrind run. Wed Jul 14 16:00:51 2004 John Haque * io.c (rsrescan): Fix off by one error at end of record. Thu Jul 8 16:59:51 2004 Stepan Kasal * awkgram.y (output_redir): Make sure not to dereference NULL pointer. The bug was triggered by the following code: gawk 'BEGIN{print "date" |& getline}' No test case created, beacuse of the following: Correct interpretation involves executing "1" or "0" -- as the user may have defined this, we would have to override this in the test script. It's not worth the hassle. Thu Jul 8 12:59:49 2004 Arnold D. Robbins * awk.h (load_casetable): Name changed from `load_ignorecase'. * eval.c (load_casetable): Name changed from `load_ignorecase'. Fix all uses. Thu Jul 8 12:32:13 2004 John Haque * awkgram.y (get_src_buf): Fix off-by-one error to avoid "does not end in newline" messages. * array.c (assoc_lookup): Small performance hack: for TEMP subs nodes, use its string memory for ahname. * ext.c (get_actual_argument): Minor code cleanup. * builtin.c (do_lshift, do_rshift, do_and, do_or, do_xor, do_compl): fixed to issue "non-numeric argument" lint warnings before using `force_number'. Mon Jun 21 16:53:35 2004 Arnold D. Robbins More changes from John Haque to rationalize extension functions. * awk.h (get_curfunc_arg_count): Name changed from `get_curfunc_parm_count'. * eval.c (get_curfunc_arg_count): Ditto, body redone to count actual args passed at call time. * ext.c (get_argument): Update range check. (get_actual_argument): Simplify the code. Mon Jun 14 14:01:16 2004 Arnold D. Robbins Changes from John Haque and ADR to rationalize extension functions. * awk.h (check_special, get_curfunc_parm_count, get_actual_argument): new function declarations. (get_scalar_argument, get_array_argument): new macros. * awkgram.y (check_special): new function. (yylex): Use `check_special' to search `tokentab'. (dump_funcs): Always count functions, in order to get dynamic ones. Removed bogus use of `static' on `tab' variable. * eval.c (struct fcall): Change type of `count' to `size_t'. (get_curfunc_parm_count): New function. (push_args): Set `r->rnode' to NULL for local variable. * ext.c (make_builtin): Add sanity checking for presence and name of new function, and that it's not a redefinition. (get_argument): Check that requested arg is within range of actual number of parameters. Also clean up logic for Node_var_new, Node_var_array, Node_array_ref. (get_actual_argument): New function. * profile.c (pp_builtin): Better handling of dynamic extension function. Sun Jun 13 14:32:22 2004 Arnold D. Robbins * io.c (redirect): Conditionalize checking for process recovery on `#ifdef PIPES_SIMULATED'. Needed for MS-DOS and VMS. * builtin.c (tmp_integer): Change bracketing of magic test to `#ifdef HAVE_UINTMAX_T' which is more general and more correct. Wed Jun 9 21:36:01 2004 Arnold D. Robbins * re.c (make_regexp): Add dfa matching into IGNORECASE handling. Tue Jun 8 15:38:56 2004 Arnold D. Robbins * awk.h (casetable): Remove `const'. * eval.c (casetable): Remove `const'. (load_ignorecase): New function. Loads locale-correct values in upper 128 bytes. (set_IGNORECASE): Call `load_ignorecase'. Tue Jun 8 14:04:19 2004 Arnold D. Robbins * awkgram.y (get_src_buf): Make sure that value from `optimal_bufsize' is reasonable. Increase it if not. Tue Jun 8 13:54:28 2004 John E. Haque * awkgram.y (statement:LEX_FOR): Fix bug in loop to `delete a' optimization. * io.c (format_tree): Check for out of range values for positional specifiers. Mon Jun 7 17:02:48 2004 Arnold D. Robbins * builtin.c (tmp_integer): Bracket the magic test inside `#ifndef VMS'. * awk.h (child_catcher): Remove declaration. * main.c (main): Remove `signal' calls for SIGCLD, SIGCHLD. * io.c (child_died, child_signo, child_catcher): Removed. (get_a_record): Remove code checking for death of child. (redirect): If `rp' matches and is at EOF and type is input pipe, and `rp->pid' is not -1, call `wait_any' to reap the child. This is a heuristic, but it works pretty well. Sun Jun 6 18:35:17 2004 Arnold D. Robbins * io.c (get_a_record): Restore use of `memmove' instead of memcpy. Otherwise some tests break on some systems. We think. (child_died): Don't reset signal handler; breaks on some S5 systems. (get_a_record): Reset handler if child_died. Still flaky on Solaris. * configure.ac (version.c): Made from version.in again, for non-Unix systems. * Makefile.am (base_sources): Add version.c back. * version.in: Include config.h for definition of const. * main.c (version_string): Add back declaration, don't include "version.i". (main): Don't install child_catcher on Sun. (HACK) Thu Jun 3 14:06:06 2004 Arnold D. Robbins * main.c (version_string): Removed declaration, since version.i is included directly. * version.in: Removed test for definition of const. Tue Jun 1 19:23:53 2004 Stepan Kasal * Makefile.am (base_sources): Don't mention version.c, so that it doesn't get distributed. * po/POTFILES.in: remove version.c * configure.ac: Create version.i from version.in. * main.c: include version.i. Tue Jun 1 18:33:32 2004 Arnold D. Robbins Fix problem reported by Stephen Marchant on Thu, 30 Oct 2003 13:11:42 -0500. * regex_internal.h (re_realloc): Allow for SunOS pre-Standard C `realloc' which doesn't accept NULL pointers. * awk.h (erealloc): Same. Unrelated. Change suggested by Peter Sobisch , have PROCINFO["version"] be the version of gawk: * main.c (load_procinfo): Add in version. Unrelated: Avoid warning: * main.c (main): Cast calls to `bindtextdomain' and `textdomain' to void. Avoids diagnostic with `configure --disable-nls'. * dfa.c (check_matching_with_multibyte_ops): Remove unneeded nested #ifdef, per Scott Deifik (scottd@amgen.com). Also fix some spelling errors in comments. Tue Jun 1 18:26:45 2004 Paul Eggert Fix a bug reported by Mike Romaniw to bug-gnu-utils on 2003-09-27: compl(compl(0xf0f)) returned 0xfff on hosts with 64-bit uintmax_t and 64-bit IEEE-764 double, due to rounding errors. * doc/gawk.texi (Bitwise Functions): Leading nonzero bits are removed in order to fit the result into a C 'double' without rounding error. * builtin.c: Include if available. (FLT_RADIX, FLT_MANT_DIG, DBL_MANT_DIG): Define if not already defined. (AWKSMALL_MANT_DIG, AWKNUM_MANT_DIG, AWKNUM_FRACTION_BITS): New macros. (tmp_integer): New function. (do_lshift, do_rshift, do_and, do_or, do_xor, do_compl): Use them. Tue Jun 1 17:40:47 2004 Stepan Kasal * eval.c (push_args): Set var_value to Nnull_string for local variables. Mon May 31 11:49:20 2004 Stepan Kasal * replace.c: #undef DEBUG before including mktime.c, it has different meaning there. Mon May 31 08:25:30 2004 Arnold D. Robbins * array.c (get_actual): Add extra error message for `delete f' inside body of function `f'. Mon May 3 09:53:34 2004 Arnold D. Robbins * Makefile.in, */Makefile.in: Updated to automake 1.8.4. * config.guess, config.sub: Same. * aclocal.m4, depcomp, install-sh: Same. Mon May 3 09:24:45 2004 Arnold D. Robbins * configure.ac: Look for missing `strtoul'. * replace.c: Include missing_d/stroul.c if not HAVE_STRTOUL. * io.c (devopen): Use `strtoul' instead of `strtod' for extracting fd number from "/dev/fd/N". (Thanks to Jim Meyering.) * field.c (set_FIELDWIDTHS): Use `strtoul' instead of `strtod' when parsing FIELDWIDTHS values. (Thanks to Jim Meyering.) Mon Apr 19 20:12:57 2004 Arnold D. Robbins * Makefile.in, */Makefile.in: Updated to automake 1.8.3. * config.guess, config.sub: Same. 2004-03-18 Stepan Kasal * eval.c (make_scalar): Comment clarification. * array.c (get_actual): Remove the condition ``canfatal'' before ``cant_happen()''; if the data are consistent, we simply cannot get there with a non-func Node_param_list, no matter whether we are called via get_array or not. * awkgram.y (variable): Make one longer message, to help translators. Tue Mar 9 17:34:10 2004 Arnold D. Robbins Adapted `gofast' patch from Redhat Enterprise version of grep to current dfa.c. * dfa.c (buf_offset): New variable. (SKIP_REMAINS_MB_IF_INITIAL_STATE): Modified to use it, don't free `mblen_buf', `inputwcs'. (match_anychar, match_mb_charset, transit_state_consume_1char, transit_state): Use buf_offset in mblen_buf. (dfaexec): Realloc things instead of free and malloc. Thu Mar 4 16:46:55 2004 Arnold D. Robbins * configure.ac (AC_FUNC_MBRTOWC): Added. (AC_CHECK_FUNCS): Removed `mbrtwoc'. (REGEX_MALLOC): Removed. Not needed for new regex* routines. * re.c (research): Removed comment and check for return of -2 since that was for old regex using alloca or REGEX_MALLOC. Wed Mar 3 17:10:16 2004 Arnold D. Robbins * io.c (close_one): Don't close stdout or stderr; can happen if /dev/stdout or /dev/stderr are used in redirection and all the open files get used. Sun Feb 29 12:17:37 2004 Arnold D. Robbins * regcomp.c (build_charclass, build_charclass_op): Change type of `class_name' parameter to `const char *' from `const unsigned char *' and adjust callers. Thu Feb 26 15:20:22 2004 Arnold D. Robbins * awkgram.y (get_src_buf): Rewritten to better manage input and supplying newlines on command line programs. Fixes problems reading source files on Cygwin. Unrelated fixes from mary1john8@earthlink.net: * node.c (format_val): For no malloc case, free s->stptr if necessary. * io.c (nextfile): Add missing call to `unref(FILENAME_node->var_value)' for no files case. (close_redir): Remove file from redirection list even if fp is stdout or stderr. Tue Feb 24 12:11:34 2004 Arnold D. Robbins * regex_internal.c (build_wcs_upper_buffer): Enclose `offsets_needed' label in `#ifdef _LIBC' to silence `unused label' compiler warning. Tue Feb 24 11:57:18 2004 Nelson H.F. Beebe * regcomp.c (parse_expression): Add cast to (unsigned char *) in calls to `build_charclass_op'. * regex_internal.c (build_wcs_buffer): Add cast to char* in call to `wcrtomb'. * regex_internal.h (bitset_not, bitset_merge, bitset_not_merge, bitset_mask, re_string_char_size_a, re_string_wchar_at, re_string_elem_size_at): Change to use prototypes. (re_string_char_size_at, re_string_wchar_at, re_string_elem_size_at): Declare as `internal_function'. * Makefile.am: Add rule to make .i files. This assists in debugging. * awk.h (m_tree_eval): Add casts to NULL. (Some compilers are just dumb. ADR) Mon Feb 23 15:58:39 2004 Arnold D. Robbins Clean up occupied process slots of children that have died: * awk.h (child_catcher): New function, declare it. * main.c (main): Catch SIGCHLD/SIGCLD with `child_catcher'. * io.c (child_died): New static variable. (child_catcher): New function, sets `child_died', reinstalls self as signal handler. (close_rp): New function: isolates actual fp/iop closing logic. (close_redir): Call `close_rp'. (get_a_record): Check `child_died' and call `wait_any(0)' if so. Add descriptive comment. Unrelated clean up: * eval.c (fcalls): Renamed from `fcall_list'. All uses changed. (pop_fcall, push_args, dump_fcall_stack): Adjusted to use indexing on `fcalls' instead of a pointer into it. Avoids hassles if `fcalls' is realloc-ed during recursive tree_evals. Thanks to BWK. * config.guess, config.sub: Updated from Savannah. 2004-02-19 gettextize * configure.ac (AM_GNU_GETTEXT_VERSION): Bump to 0.14.1. Wed Feb 18 12:40:09 2004 Arnold D. Robbins * awkgram.y (rule): Use `msg' not `warning' for `must have an an action part' message. `warning' is wrong, since it's a real error. Mon Feb 16 12:17:39 2004 Arnold D. Robbins * main.c, eval.c, builtin.c: Change test for `#ifdef HAVE_LOCALE_H' to `#if ENABLE_NLS && defined(HAVE_LOCALE_H)' so that builds with `configure --disable-nls' will actually work on non-glibc systems. Thu Feb 12 02:05:34 2004 Stepan Kasal Lots of misc changes from Stepan Kasal integrated. * array.c: Various variables and parameters of static functions changed from int to long, in order to prevent overflow. * eval.c (make_scalar): New function; takes care of everything that has to be done when a node of type Node_var_new or Node_array_ref changes to a scalar variable. (tree_eval, get_lhs): Call it. (tree_eval): From now on, tree_eval(NULL) doesn't work; it reports an internal error. * awk.h (m_tree_eval): Likewise for the two macro versions. * awkgram.y (statement): Make sure the Node_K_return's lnode is always the return value, never NULL. * Makefile.am (install-exec-hook, uninstall-links): Make use of $(VERSION). (INCLUDES): Renamed to AM_CPPFLAGS. (AM_CPPFLAGS): The file libintl.h is generated in the build subdirectory intl, not in the directory $(srcdir)/intl. (diffout): New target is an alias for ``make -C test diffout.'' * awklib/Makefile.am (INCLUDES): Renamed to AM_CPPFLAGS. * README_d/README.hpux: Change the whitespace in the appended patch, so that it applies to the current source. * posix/gawkmisc.c: Change a tab to a space (needed for the above). Make version control more in the style of current autotools: * configure.ac: Remove obsolete versions of macros: AM_INIT_AUTOMAKE doesn't need any parameters. AC_OUTPUT shouldn't have any parameters either. Its parameters go to a new macro: AC_CONFIG_FILES. AC_CONFIG_HEADERS moved near the end of the file. * configure.ac: Add [version.c:version.in] to AC_CONFIG_FILES * version.in: Modify for autoconf substitutions. * version.c: Remove, it's generated at configure time now. * fixvers, patchlev.h, unsupported/tandem/ptchlvl.h: Nuke and ... * Makefile.am, main.c: ... forget them. Mon Feb 9 12:57:00 2004 Arnold D. Robbins * awk.h builtin.c eval.c field.c (HUGE): Changed to `UNLIMITED'. Avoids possible conflict with constant in svid-mode math.h. Thanks to Roman.Putanowicz@iecn.u-nancy.fr for pointing out the problem. Fri Feb 6 12:09:55 2004 Arnold D. Robbins * configure.ac: Check for `wctype', `wcscoll' and `iswctype'. * mbsupport.h: New file. Merges and centralizes testing for MBS support. * Makefile.am (base_sources): Add mbsupport.h to list. * dfa.c, dfa.h, awk.h (MBS_SUPPORT): Include "mbsupport.h" and use the test there. * regex_internal (RE_ENABLE_I18N): Same. * Makefile.am (CLEANFILES): Added. Thu Feb 5 18:05:12 2004 Arnold D. Robbins * configure.ac (HAVE_WCTYPE_T): New test code added. * dfa.h (wctype_t): Define if system doesn't. Needed here too for other files that include dfa.h. * dfa.c (wctype_t): Define if system doesn't. (lex): Manually fill in arrays used for char class range testing so will work on c89 and older compilers. (parse_bracket_exp_mb): Split up an assignment to avoid type complaints. * main.c (main): When checking for `close_io' failure, only set `exit_val' to 1 if not already exiting. * regcomp.c (regerror): Remove use of mempcpy. Generates too many compiler warnings. * configure.ac (AC_CHECK_FUNCS): Don't bother checking for it. Wed Feb 4 17:34:47 2004 Arnold D. Robbins * getopt.h (__THROW): Only define it if C++. The __GNU_PREREQ macro is a major headache. 2004-02-02 Paolo Bonzini * regexec.c (check_matching): Add P_MATCH_FIRST parameter. (re_search_internal): Pass new parameter to check_matching. (check_matching): Unless a parenthesized group is found at the beginning of the regexp, advance P_MATCH_FIRST until we entered a state different from the initial state. Mon Feb 2 15:52:37 2004 Arnold D. Robbins * re.c (research): Change last param to re_search to pass NULL if `need_start' is false. May give us a marginal speed gain. Thu Jan 29 17:04:51 2004 Arnold D. Robbins * builtin.c (sub_common): Fix logic for `&' in replacement for multibyte case. Simplify code a bit. Tue Jan 20 10:41:45 2004 Arnold D. Robbins * configure.ac: Add check for `memmove'. * replace.c: Include missing_d/memmove.c if don't have `memmove'. Sun Jan 18 12:01:29 2004 Arnold D. Robbins * builtin.c (sub_common): Add comment and support for 2001 POSIX behavior when --posix in effect. The masses have been clamoring for this one. 2004-01-16 gettextize * configure.ac (AM_GNU_GETTEXT_VERSION): Bump to 0.13.1. * intl/*: Updated to 0.13.1. Fri Jan 16 08:16:38 2004 Arnold D. Robbins * getopt.h, getopt.c, getopt1.c: Synced to GLIBC version: getopt.c: 1.51 getopt.h: 1.18 getopt1.c: 1.9 Thu Jan 15 15:28:48 2004 Arnold D. Robbins Here we go again: * regcomp.c, regex.h, regex.c, regex_internal.h, regex_internal.c, regexec.c: Sync to GLIBC version, but with bug fixes. GLIBC CVS versions: regcomp.c: 1.76 regexec.c: 1.55 regex.c: 1.125 regex.h: 1.30 regex_internal.c: 1.39 regex_internal.h: 1.45 regexec.c: 1.55 * acinclude.m4: Removed, not needed for automake 1.8.x. * configure.ac: Updated to autoconf 2.59. Everything else updated to automake 1.8x and autoconf 2.59. Wed Jan 14 14:26:36 2004 Arnold D. Robbins * dfa.c, dfa.h: Updated manually with most of the changes in grep 2.5.1. That version lost the ability to match newlines in the data, so the merge had to be done by hand. Sigh. 2004-01-12 Paolo Bonzini ALLOCA patch from http://sources.redhat.com/ml/libc-alpha/2004-01/msg00099.html added. * regcomp.c [_LIBC && !RE_ENABLE_I18N]: Drop code to support this, it is never true. (build_range_exp) [!_LIBC]: Do not create a range in MBCSET for a single-byte character set. (build_range_exp) [_LIBC]: Do not create a range in MBCSET for a single-byte character set without collation elements. (init_dfa): Do not conditionalize on _LIBC, it just makes the code less clear. (parse_bracket_exp): Use NON_MATCH variable in addition to "mbcset->non_match", not as an alternative. (build_charclass_op): rename NOT parameter to NON_MATCH, use it instead of declaring a variable. (parse_bracket_exp) [!_LIBC]: Pass NULL for MBCSET if the character set is single-byte. Wed Jan 7 15:23:04 2004 Arnold D. Robbins * awkgram.y (mk_rexp): Set n->re_cnt to 1. Makes reinstated dfa code actually take effect! Don't know how I missed this. Unrelated: sync regex code to glibc. * regcomp.c, regex.h, regex.c, regex_internal.h, regex_internal.c, regexec.c: Sync to GLIBC version, but with bug fixes. GLIBC CVS versions: regcomp.c: 1.74 regex.c: 1.124 regex.h: 1.30 regex_internal.c: 1.39 regex_internal.h: 1.43 regexec.c: 1.55 * regcomp.c (peek_token): Temporarily, we hope, disable \s and \S operators. Too much trouble to document right now. * dfa.c (lex): Add code for \s and \S but disable it until next release. Wed Dec 24 15:28:57 2003 Arnold D. Robbins * eval.c (r_tree_eval): For Node_assign_concat, add call `free_temp(r)'. Thanks to mary1john8@earthlink.net. Mon Dec 1 10:25:52 2003 Arnold D. Robbins CONSTVAL not needed if we test PERM instead. Thanks to mary1john8@earthlink.net. See test/concat3.awk. * awk.h [CONSTVAL]: Removed. * eval.c (flag2str): Removed CONSTVAL from table. (r_tree_eval): For Node_assign_concat, it's enough to check if l->flags has PERM clear. * awkgram.y (yylex): Removed use of CONSTVAL for YSTRING and YNUMBER. Mon Nov 3 16:33:26 2003 Arnold D. Robbins * awkgram.y (get_src_buf): Replace `memmove' with `memcpy' for marginal portability gain to older systems. * io.c (get_a_record): Ditto. Sun Nov 2 15:59:27 2003 Arnold D. Robbins * awk.h [CONSTVAL]: Renamed from `STRCONST'. * eval.c (flags2str): Fix in table. (r_tree_eval): For Node_assign_concat, check for the flag for both left and right hand sides. Also add a `force_string' call for the right hand side and the left hand side. * awkgram.y (yylex): Change flag value for YSTRING and add use of flag for YNUMBER. Wed Oct 29 14:23:29 2003 Arnold D. Robbins * awk.h [STRCONST]: New flag value. * eval.c (flags2str): Add it to table. (r_tree_eval): For Node_assign_concat, check for the flag so that we don't clobber string constants given: s = "" s = s something * awkgram.y (yylex): For YSTRING, set STRCONST flag. Tue Oct 28 18:00:00 2003 Arnold D. Robbins Force SUBSEP to always have a string value. Per bug report from mary1john8@earthlink.net. * awk.h (NODETYPE): New type, Node_SUBSEP. (set_SUBSEP): Add declaration. * awkgram.y (isnoeffect, isassignable): Add Node_SUBSPEP case. * array.c (set_SUBSEP): New function. * eval.c (nodetypes): Add Node_SUBSEP. (r_tree_eval, r_get_lhs): Add code for Node_SUBSEP. * main.c (varinit): Use Node_SUBSEP as type for SUBSEP. * profile.c (tree_eval, pp_lhs, is_scalar, prec_level): Handle Node_SUBSEP. Tue Oct 7 09:26:33 2003 Arnold D. Robbins * awk.h (NODETYPE): New member `Node_assign_concat'. * awkgram.y (exp): Look for case of `x = x y' and if so, create a Node_assign_concat. * eval.c (interpret): Add case for Node_assign_concat. * profile.c (prec_level): Ditto. (tree_eval): Ditto. For variables, call new function `vname' to print name; handles varname field for -v variables, which end up including the value. (vname): New function. Wed Sep 24 17:32:31 2003 Arnold D. Robbins Speed up `avoid_dfa' kludge, at least a little: * awk.h (struct Regexp): Add `has_anchor' member. Make it and `dfa' member shorts; keeps space the same. * re.c (make_regexp): Set `has_anchor' member correctly. (avoid_dfa): Test for `has_anchor' member instead of searching for it each time. Sun Sep 21 18:34:32 2003 Arnold D. Robbins * builtin.c (state): Only needs to be 256 bytes, initstate() can't use any more than that. Well whadayaknow. (do_rand, do_srand): Call `setstate' after calling `initstate'. Tue Sep 16 15:44:29 2003 Arnold D. Robbins * eval.c (interpret): For Node_K_switch, add kludge_need_start stuff as used in `match_op'. Sigh. * re.c (make_regexp): Add `no_dfa' variable, which is true if GAWK_NO_DFA exists in the environment. This enables run time testing of things with/without the dfa matcher. Mon Sep 15 18:36:38 2003 Arnold D. Robbins After much soul searching, reinstated old dfa code. The performance of the new regex is just disastrous. Sigh. * awk.h (re_cnt): Reinstated old definition. (struct Regexp): Added `dfareg' and `dfa' members. (make_regexp): New last parameter in function, changed decl. (avoid_dfa): Added declaration. * awkgram.y (regexp, mk_rexp): Added use of `re_cnt'. Fixed call to `make_regexp'. * Makefile.am: Add dfa.h and dfa.c. * eval.c (match_op): Complexified: added call to `avoid_dfa' and `kludge_need_start' variable where used to pass FALSE as last parameter of research(). * field.c (set_FS): Fixed call to `make_regexp'. * io.c (get_a_record, set_RS): Fixed calls to `make_regexp'. * re.c (make_regexp): Added last paramter (`dfa') to function. Complexified the code. (re_update): Fixed call to `make_regexp'. (research): Complexified the code, added calls to dfa stuff. (dfaerror): New function. (re_update): Fixed call to `make_regexp'. (avoid_dfa): New function. Tue Sep 9 15:57:38 2003 Arnold D. Robbins * awkgram.y (get_src_buf): Fix calculation of `offset' when shifting source lines around. In general, improve handling of things when moving the source code line around. What a mess this code is. Mon Sep 8 19:08:55 2003 Arnold D. Robbins * eval.c (fmt_ok): Make provision for %F format and printf %'f flag here too. 2003-07-23 Christophe Bisiere (tiny change) * posix/regex.h (RE_TRANSLATE_TYPE): Define it to "unsigned char," to avoid problems at hosts with signed char. * posix/regexec.c (re_search_internal): Don't say "unsigned RE_TRANSLATE_TYPE." Thu Aug 28 11:09:41 2003 Arnold D. Robbins * io.c (nextfile): Change use of variable `files' to make it clearer that it's a boolean flag. Tue Aug 26 22:58:15 2003 Arnold D. Robbins * io.c (useropen): Add `defined (HAVE_GETGROUPS)' as first test in `#ifdef'. Brings things in sync with same test in main.c and awk.h. Tue Aug 26 22:49:37 2003 Arnold D. Robbins * dbug.h: New file. * array.c, awkgram.y, builtin.c, eval.c, ext.c, field.c, io.c, main.c, msg.c, node.c, profile.c, re.c: Converted to use Fred Fish's `dbug' library. By default compiled out, thus not affecting speed. For the nonce, the `dbug' library itself is not shipped with gawk, since I expect no-one else but me to be using it. Thu Aug 21 23:15:36 2003 Arnold D. Robbins * awkgram.y (numfiles): Change extern decl to long, to match what's in main.c. Keeps things working on 64-bit systems. Thanks to bug report from Jan Oravec . Wed Aug 20 14:53:47 2003 Arnold D. Robbins * eval.c (genflags2str): Move test for out-of-space inside test for is the bit set. Mon Aug 11 11:26:51 2003 Arnold D. Robbins * regcomp.c (parse_bracket_exp): If `build_charclass' fails, just pass its value on as the return value. Sun Aug 10 16:59:14 2003 Arnold D. Robbins * regcomp.c (build_range_exp): Make sure we don't get WEOF on range characters. Tue Aug 5 21:49:32 2003 Arnold D. Robbins * array.c (get_actual): In `case Node_param_list' add test for `&& (symbol->flags & FUNC) == 0' to the if. Sun Jul 13 18:28:38 2003 Arnold D. Robbins Further bug fix: * awkgram.y (variable): Give the new variable an lnode of Nnull_string if it's not an array, even if it is a Node_var_new. Fri Jul 11 09:32:21 2003 Arnold D. Robbins Bug fix: * eval.c (r_tree_eval): For Node_array_ref, set tree->orig_array->var_value to Nnull_string too. Unrelated i18n and POSIX change: * configure.ac: Add check for local printf supporting %F format. * awk.h (loc): New variable declaration. * main.c (loc): Defined. (main): Call `localeconv' to set loc. * io.c (format_tree): Add support for printf quote flag, %'d for decimal formats (not %e, %E), adds thousand separator into value. 2003-07-10 Paul Eggert * io.c (two_way_open): If /bin/sh cannot be executed, exit with status 126 consistently. Mon Jul 7 11:01:43 2003 Arnold D. Robbins * Release 3.1.3: Release tar file made. Mon Jul 7 09:55:49 2003 Arnold D. Robbins * main.c (copyleft, usage): Make sure to fflush output fp. Per Jim Meyering, if error, exit non-zero. * ext.c (get_argument): Fix array paramater handling. 2003-07-06 Paul Eggert * builtin.c (do_substr): Issue better diagnostics when d_substr and d_length are NaN, or when 0 < d_length < 1. Be careful when comparing double to SIZE_MAX, as the comparison might return the "wrong" answer when `(double) SIZE_MAX' is a number that is not equal to SIZE_MAX. (do_gensub): Watch out for HOW values that are out of range or are NaN. (do_dcngettext): dcngettext wants an argument of type unsigned long, not long, so use a value of that type. Fri Jul 4 10:58:02 2003 Arnold D. Robbins * main.c (main): Make option letter 'D' fall through into default unknown case if not debugging. Let's us have just one version of `optlist'. Thu Jun 26 15:25:57 2003 Arnold D. Robbins * array.c (get_actual): Undo Stepan Kasal change of 2003-06-17. See test/match2.awk. Wed Jun 25 15:26:08 2003 Arnold D. Robbins * builtin.c (do_rand): Change calculation so that result obeys constraint: 0 <= N < 1. This is per history and POSIX. Thanks to Nelson Beebe (beebe@math.utah.edu) for reporting this issue. Mon Jun 23 15:13:39 2003 Arnold D. Robbins * io.c (rs1scan): Per advice from Bruno Haible, it's safe to skip the multibyte checking code if RS is '\n'. See the comment in the code. Big performance improvement for multibyte locales. 2003-06-20 Stepan Kasal * eval.c (comp_func): if memcmp returns 0, we have to compare the lengths. 2003-06-19 Stepan Kasal * eval.c (interpret) : use NULL, not 0, to initialize the variable list. (comp_func): array indices no longer are string values, you have to use ahname_str, ahname_len. Tue Jun 17 11:53:46 2003 Arnold D. Robbins * array.c (get_actual): Make check for isparam smarter: also check for FUNC flag. 2003-06-17 Stepan Kasal * array.c (get_actual): even if canfatal is FALSE, don't tolerate existence of things which can't happen. Mon Jun 16 16:21:44 2003 Arnold D. Robbins * configure.ac: Removed m4/Makefile. * m4/Makefile.am: Removed. 2003-06-16 gettextize * configure.ac (AC_OUTPUT): Add m4/Makefile. (AM_GNU_GETTEXT_VERSION): Bump to 0.12.1. Sun Jun 15 20:45:43 2003 Arnold D. Robbins * io.c (get_a_record): Enhance logic to fill buffers to include `|| no_data_left(iop)'. (rs1scan): Fix logic for setting recm fields for multibyte character case. 2003-06-10 Stepan Kasal * awkgram.y (release_all_vars): do not try to release a value of Node_var_new; after get_lhs, use the lhs directly, do not try to do (*lhs)->var_value; the Node_var case doesn't need special treatment. * builtin.c (do_match): `get_param' is successful iff it returns Node_var_array---if the variable was new, get_param has already changed the type. * field.c (do_split): likewise. Sun Jun 15 19:36:35 2003 Arnold D. Robbins * node.c (r_dupnode): Typo fix in hash tables: stptr -> ahname_str. Thanks to mary1john8@earthlink.net. * array.c (get_actual): Add `if (canfatal)' before call to `cant_happen'. Sun Jun 15 19:25:49 2003 Patrick T.J. McPhee * awk.h (memcpy_ulong): Add ! WIN32 to ifdefs. Mon Jun 9 18:38:20 2003 Arnold D. Robbins * custom.h (hpux): Added stuff to (hopefully) get `tzset' declared on HP/UX 10 and up. Mon Jun 9 17:12:24 2003 Patrick T.J. McPhee * awk.h (ATTRIBUTE_EXPORTED): New macro for dynamic libs on Windows32. * CONVMFTidx, stack_ptr, do_lint, lintfunc: Now have this attribute. Mon Jun 9 13:11:33 2003 Arnold D. Robbins Clean up of Stepan's patches. * array.c (get_actual): Renamed from r_get_array. Added second param canfatal if routine should print fatal message when not an array. (get_actual): Renamed 'prm' to `isparam'. (array_vname): Add static msglen var; only realloc string if it grows. Don't use `s += sprintf(...)'. No good on old systems where sprintf returns char *. Minor formatting cleanups. (do_adump): Restored separate `a' and `r' variables; helps for debugging. * awk.h (SCALAR, UNINITIALIZED): Removed entirely, renumbered other flags. (get_array, get_param): New macros, calls get_actual. (get_actual): Declaration changed from that of r_get_array. * awkgram.y (release_all_vars): Restored previous version of code; new version isn't right for Node_xx variables. (variable): Minor code cleanup for readability. * builtin.c (do_match): Use get_param and print our own message when third parameter is not an array. * eval.c: Added a few comments here and there, removed some no longer needed comments. * field.c (do_split): Use get_param and print our own message when second parameter is not an array. Mon Jun 9 11:46:21 2003 Arnold D. Robbins * array.c (do_delete): Initialize hash1 and last to keep gcc -Wall happy. * io.c (rsnullscan): Comment out label skip_leading for same reason. Wed May 28 08:31:23 CEST 2003 Stepan Kasal * eval.c (forloops_active, in_function): Nuked. (pop_all_forloops, pop_fcall_stack): are now inline. Wed May 28 07:58:35 2003 Stepan Kasal * field.c, awk.h (Null_field): no longer static. * field.c (init_fields): initial value of $0 can be Nnull_string, no need to copy it. * eval.c (r_get_lhs) : test for uninitialized field, which is Nnull_string for $0 and Null_field for $(>0). * builtin.c (do_print_rec): test for uninitialized $0. Tue May 27 17:03:02 2003 Stepan Kasal * awk.h (Node_var_new): New node type for variables which can be either scalar or array. From now on, Node_var is always scalar. (Node_gvar_ref): Nuked, its role can be taken by Node_array_ref. (orig_var): removed, orig_array is enough. (SCALAR, UNINITIALIZED): Flags nuked. (var_uninitialized): new macro to distinguish uninitialized vars; used in several other macros. * array.c (r_get_array, array_vname, do_adump): adapt to the above changes. * awkgram.y, eval.c, field.c, main.c, node.c, profile.c: ditto. Tue May 27 14:27:50 2003 Stepan Kasal * array.c (r_get_array): New function, which goes all the way through Node_param_list to actual Node_var_array; if it encounters non scalar Node_var, it changes it to Node_var_array. (in_array, do_delete, do_delete_loop, do_adump, assoc_sort_inplace): Use get_array. (assoc_lookup): The parameter must be a Node_var_array. * awk.h (get_array, r_get_array): Declare the new function and define a macro to speed it up. * builtin.c (do_match): Use get_array. * eval.c (interpret) : ditto. (r_get_lhs) : ditto. * field.c (do_split): ditto. Tue May 27 08:23:51 2003 Stepan Kasal Changed node->vname meaning for type Node_array_ref and Node_gvar_ref. It contains only the reference name; one has to (recursively) follow node->prev_array to find out the call history for the array. * array.c (array_vname): New function to print the array name. (assoc_lookup, do_delete): Use array_vname. * eval.c (interpret, r_tree_eval, r_get_lhs): Use array_vname. (push_args, pop_fcall): Things have simplified. * awk.h (array_vname): Declare. (prev_array): Define. Sun Jun 8 11:25:36 2003 Stepan Kasal * awkgram.y (append_right): when using savetail, remember that it is not necessarily the tail of the list---it's just a pointer to the last chunk appended. Thu Jun 5 12:01:41 2003 Arnold D. Robbins * builtin.c (do_strtonum): Make `strtonum(13)' work. Wed Jun 4 17:07:06 2003 Corinna Vinschen * io.c (binmode): Include function for __CYGWIN__ too. Tue Jun 3 12:40:50 2003 Arnold D. Robbins * eval.c (interpret): Node_K_switch. For regex case, don't call `free_temp' on the result of `force_string' if it's equal to switch_value. Thanks to John DuBois for finding the problem. Sun Jun 1 13:08:22 2003 Arnold D. Robbins * builtin.c (format_tree): For %c, force precision to 1. Wed May 28 11:55:48 2003 Arnold D. Robbins * awkgram.y (isnoeffect, isassignable): Add Node_TEXTDOMAIN to switches in both functions. Wed May 28 11:38:59 2003 Stepan Kasal * awkgram.y (switch_body): remove rule ``switch_body:/*empty/'' as ``switch_body:case_statements'' covers it---this disambiguation fixes a reduce/reduce conflict. Sun May 25 16:23:43 2003 Corinna Vinschen * configure.ac: Remove linking against /usr/lib/automode.o. * configure: Regenerate. Sun May 25 15:19:19 2003 Stepan Kasal * awk.h (get_lhs): For an initialized Node_var, you may return the address of var_value pointer, no matter whether reference bit was set or not. We were silly slowing down most of the assignements. * (get_a_record): after grow_iop_buffer, move recm.rt_start even if recm.len == 0. Mon May 19 16:55:59 2003 Arnold D. Robbins Code for C-style switch statements. Initial version contributed by Michael Benzinger . Disabled by default, use `configure --enable-switch' to turn it on. * configure.ac: New AC_ARG_ENABLE for switch statements. * awk.h (NODETYPE): New types for switch, case, default keywords and respective lists. * awkgram.y: New productions for switch statement. Does checking to avoid duplicate cases. * eval.c (nodetypes): New entries for new NODETYPEs. (interpret): New code to do switch execution. * profile.c (pprint): New code to print switch statements. Mon May 19 15:05:43 2003 Arnold D. Robbins * configure.ac: Renamed from configure.in. * fixvers: Now looks in configure.ac. * Makefile.am: Now cites configure.ac. * Misc other: Updated to Automake 1.7.5. Sun May 18 12:03:56 2003 Arnold D. Robbins * array.c (do_delete_loop): Fix bracing of logic for tests. Wed May 14 09:01:16 2003 Stepan Kasal Misc patches: * builtin.c (do_match): if third parameter to `match' is supplied, store all subexpressions which are applicable, even though there are some unused between them. * awkgram.y (yylex): when returning from unterminated REGEXP (which is /* kludge */), take care to fake a yylval, to prevent ``internal error'' later. Sun May 11 15:51:00 2003 Stepan Kasal * io.c (rsnullscan, get_a_record): Boundary condition bug fixes. Sun May 11 15:15:20 IDT 2003 Scott Deifik * awk.h: Add decls for `memcpy_ulong', `memset_ulong', and MSC defines. * regex.c: Include if MSC for size_t. Mon May 5 15:11:06 2003 Arnold D. Robbins * io.c (get_a_record): Only tweak RT's value in place if the current RS scanner is the same as the last one. Bug report submitted by John DuBois (). Fri May 2 14:39:48 2003 Arnold D. Robbins * io.c (get_a_record): Add logic at end to be smart about setting RT. Saves considerable time, esp for default case where RS = "\n". Wed Apr 30 11:44:38 2003 Arnold D. Robbins * field.c (do_split): Add check and code for Node_gvar_ref. * array.c (in_array, do_delete, asort_actual): Same. * builtin.c (do_split): Same for 3rd arg array parameter. * eval.c (interpret): Same for Node_K_array_for. (push_args): Same for evaluating extra args. Tue Apr 29 15:54:28 2003 Arnold D. Robbins Record reading code redone/simplified considerably. * awk.h (IOBUF): Removed total field, no longer used. * io.c (at_eof, has_data, no_data_left): New macros. (RECVALUE, SCANSTATE): New enumerated types. (rs1get_a_record, rsnull_get_a_record, rsre_get_a_record): Removed. (get_a_record): Rewritten, again. Now contains just buffer and record code; searching code moved into these functions: (rs1scan, rsnullscan, rsrescan): New functions to scan a buffer for record contents and terminator. Fill in values in: (struct recmatch): Holds found record and terminator. (spec_setup): Set iop->dataend to indicate data is already in buffer. (nextfile, inrec): Use new macros instead of flag and pointer tests. (set_RS): Set scanning function instead of record function. FWIW, it all passes `make test'. Sun Apr 27 21:02:39 2003 Arnold D. Robbins * io.c (do_close): At end, if do_posix return 0. See comment in code. Tue Apr 15 09:56:03 2003 Isamu Hasegawa * configure.in: Check existence of wcrtomb, and wcscoll. * configh.in: Likewise. * configure: Re-generate. Thanks to Kimura Koichi for reporting. Sun Apr 13 16:02:10 2003 Arnold D. Robbins * main.c (main): Add call to `setlocale' for LC_NUMERIC after program is parsed. (arg_assign): Switch back to "C" locale for LC_NUMERIC for command line assignments; this is per POSIX that period is decimal point for program and command line assignments and the locale's separator applies for input, output, and string to number conversion. 2003-03-26 Paul Eggert * builtin.c [HAVE_INTTYPES_H]: Include . [!HAVE_INTTYPES_H && HAVE_STDINT_H]: Include . (CHAR_BIT, INTMAX_MIN, UINTMAX_MAX): Define if the system does not. (TYPE_SIGNED, TYPE_MINIMUM, TYPE_MAXIMUM): New macros, taken from coreutils and many other GNU utilities. (format_tree): When formatting, use widest possible integers rather than settling with 'long'. (do_lshift, do_rshift, do_and, do_or, do_xor, do_compl): Likewise, when doing bitwise operations. * configure.in (jm_AC_TYPE_LONG_LONG, jm_AC_TYPE_UNSIGNED_LONG_LONG, jm_AC_TYPE_INTMAX_T, jm_AC_TYPE_UINTMAX_T): Add, since the mainline code now needs this. * doc/gawk.texi (Control Letters, Bitwise Functions): Document this. * m4/intmax_t.m4: New file, taken from coreutils (but renamed to avoid collision with our m4/inttypes.m4). * m4/longlong.m4: New file, taken from coreutils. * m4/uintmax_t.m4, m4/ulonglong.m4: Remove; superseded by the above new m4 files. * builtin.c (BITS_PER_BYTE): Remove; use CHAR_BIT instead, since it's the standard name. (do_lshift, do_rshift): Complain if the shift width is exactly equal to the word size, too. Thu Mar 27 10:44:11 2003 Arnold D. Robbins * io.c (rs1_get_a_record, rsnull_get_a_record, rsre_get_a_record): Enhance check for no data left in file to be only if file has non-zero size. Linux files such as /proc/filesystems stat as a regular file of size 0, but actually have contents. Ugh. Thanks to Martin Schlemmer for the bug report. Wed Mar 26 12:19:32 2003 Arnold D. Robbins * builtin.c (format_tree): Add a lint warning at label `out_of_range'. Tue Mar 25 12:24:38 2003 Arnold D. Robbins * awkgram.y (variable): For array subscript, if NAME is in the symbol table, but not a variable, array, or parameter, generate a syntax error. (isarray): New function, tests if a symbol can be an array. * custom.h: Add check for HP/UX, needed for GCC. Mon Mar 17 09:21:09 2003 Arnold D. Robbins Allow simultaneous manipulation of a global array directly and when passed as a parameter. * awk.h (Node_gvar_ref): New nodetype. [orig_var]: New macro. * array.c (do_delete_loop, do_delete): Add logic to handle seeing Node_gvar_ref. * eval.c (nodetypes): Add Node_gvar_ref. (r_tree_eval, r_get_lhs): Add Node_gvar_ref case. Wed Mar 19 14:10:31 2003 Arnold D. Robbins This time for sure. -- Bullwinkle * Release 3.1.2: Release tar file made. Wed Mar 19 14:08:11 2003 Arnold D. Robbins * awkgram.y: Production `program --> program error'. Add a return so that we don't produce an infinite stream of error messages. Thanks to Michael Mauch for pointing this out. Wed Mar 19 13:45:50 2003 Corinna Vinschen * regex.c [RE_ENBABLE_I18N]: remove definition; the one in regex_internal.h is better and makes things work with Cygwin. Tue Mar 11 11:54:20 2003 Arnold D. Robbins * regex_internal.h: Don't include after was included in regex.c, since it could redefine RE_DUP_MAX to a lower value. (bitset_set, bitset_clear, bitset_contain): Use 1UL instead of 1 in left shift operations. * regex.c: Include before * regcomp.c (re_compile_fastmap_iter, init_word_char, parse_expression): Use 1UL instead of 1 in left shift operations. Mon Mar 10 15:45:37 2003 Corinna Vinschen * configure.in: Update CYGWIN case to add /usr/lib/automode.o. Thu Mar 6 11:07:36 2003 Arnold D. Robbins Updated to automake 1.7.3. * config.guess, config.sub: Updated from prep. * Makefile.am (AUTOMAKE_OPTIONS): Add dist-bzip2 to get .bz2 files. Tue Mar 4 10:40:46 2003 Arnold D. Robbins * version.in: Added goop for K&R compilers; forgot that I have to fix this file which then is used to create version.c. Mon Mar 3 17:00:44 2003 Arnold D. Robbins * configure.in: New option --disable-lint. * awk.h (do_lint, do_lint_old): Conditionally declare based on NO_LINT. * eval.c (set_LINT): Ifdef out body if NO_LINT. * main.c (do_lint, do_lint_old): Conditionally compile properly. (main): Handle --lint argument code. Fri Feb 28 10:43:07 2003 Arnold D. Robbins * main.c (main): Add LC_TIME to the things that get set with setlocale(). * builtin.c (format_tree): Change test of `n0-- <= 0' to ==, avoids VMS diagnostic. Thu Feb 27 17:48:29 2003 Pat Rankin * regexec.c (proceed_next_node): Cast re_string_get_buffer to char *. (get_subexp): Likewise. Tue Feb 25 12:33:41 2003 Arnold D. Robbins * regex_internal.h, regex_internal.c, regcomp.c, regexec.c: Make MB_CUR_MAX into thread local variable re_mb_cur_max. Unrelated, from Scott Deifik: * io.c (grow_iop_buffer): Add checks for overflow of new buffer size. Mon Feb 24 13:30:59 2003 Arnold D. Robbins * awk.h (gawk_mb_cur_max): Declared: * main.c (gawk_mb_cur_max): Defined, init to 1. (main): Initialize gawk_mb_cur_max. * awkgram.y, builtin.c, eval.c, field.c, io.c, re.c (mb_cur_max): Replaces all instances of MB_CUR_MAX, which is a function call (!) in glibc. Big speed up, especially for -Fx case, where x is a single character. Unrelated: * awkgram.y (rule): For non-existent action, use a Node_K_print_rec node. Sun Feb 23 15:45:20 2003 Arnold D. Robbins Speed up plain `print' and `print $0': * awk.h (Node_K_print_rec): New node type. (do_print_rec): Declare function. * awkgram.y (simple_stmt): Create humongous test for plain `print' or `print $0', and if so, use a Node_K_print_rec for it. Modify test for lint message. * builtin.c (redirect_to_fp): New function for common code to get fp and rp for do_print{,f,_rec} functions. (do_print): Use redirect_to_fp(). (do_printf): Use redirect_to_fp(). (do_print_rec): New function to just print $0 from field_arr[0] directly; will rebuild the record first if necessary. * eval.c (nodetypes): Add Node_K_print_rec. (interpret): Add Node_K_print_rec case. * profile.c (pprint): Add Node_K_print_rec case. (pp_print_stmt): If null lnode, print "$0" else print the lnode. Unrelated: * regex_internal.h: Add ENABLE_NLS to the condition for using gettext so that --disable-nls really disables it. Sat Feb 23 22:46:00 2003 Arnold D. Robbins * io.c (rs1_get_a_record, rsnull_get_a_record, rsre_get_a_record): Modify buffer-filling algorithm to always read one or more multiples of the blocksize (iop->readsize). (grow_iop_buffer): Make sure there's room for the current partially read record and one disk block buffer. Thu Feb 20 22:02:00 2003 Arnold D. Robbins * re.c (research): Fix typo in cast of precision value to int. * regex.h, regex.c, re_internal.h, re_internal.c, regcomp.c, regexec.c: synced to GLIBC source, maintaining K&R portability changes, and bug fixes, although losing ability to compile each file separately. * Makefile.am (SOURCES): Moved placement of regex source files from here ... (EXTRA_DIST): ... to here. Tue Feb 18 14:17:33 2003 Arnold D. Robbins * re.c (research): Cast precision value to int. * builtin.c (format_tree): For toofew, cast field width value to int. * io.c (rsre_get_a_record): Initialize restart and reend. Add a variable to make sure they're set before used at end of function. (iopflags2str): Removed decl at top and made not static so that GCC stops complaining that it's defined but not used. Bleah. Mon Feb 17 11:02:34 2003 Arnold D. Robbins * config.guess, config.sub: Updated from prep. Sun Feb 16 15:47:15 2003 Scott Deifik * awk.h (format_tree, make_str_node): Changed decls to match how they are called. * builtin.c (format_tree, sub_common): Same. * node.c (make_str_node): Same. Wed Feb 5 14:18:01 2003 Arnold D. Robbins * awk.h: Removed duplicate decl of set_prof_file(). Removed undef of const for non-ANSI C; config.h should handle it. * msg.c (set_loc): Use srcfile and srcline in regular code to shut up stupid SGI compiler. Tue Feb 4 14:28:06 2003 Arnold D. Robbins All relevant files: Copyright year updated to 2003. Tue Feb 4 13:40:41 2003 Martin C. Brown * intl/libgnuintl.h: Preprocessor fixes for MacOS X. * regex.h: Ditto. Tue Feb 4 13:39:37 2003 Arnold D. Robbins * awkgram.y (builtin_func): New string for use in rationalizing function parsing and installation code. Sun Feb 2 16:00:55 2003 Arnold D. Robbins Cache function body code pointer so that only have to find it the first time a function is called. This potential for optimization brought to my attention by Stepan Kasal. * awk.h [funcbody]: New macro. * awkgram.y (FUNC_CALL): Set $$->funcbody to NULL. * eval.c (func_call): Changed to take top-level Node_func_call as the single parameter. Do the lookup and caching. (r_tree_eval): Change how func_call() is called in switch. * profile.c (pp_func_call): Similar changes. (tree_eval): Ditto. Sun Feb 2 15:32:42 2003 Stepan Kasal ADR: More grammar rationalization/repair from Stepan. * awkgram.y (common_exp, simp_exp): the rule from getline (without pipe) has been moved from common_exp to simp_exp. The redirection of print statements reworked. The idea comes from mawk-1.3.3; much thanks to Michael Brennan! * awkgram.y (IO_OUT, IO_IN): new tokens. (APPEND_OP, TWOWAYIO): swallowed by the above ones. (in_print, in_parens): new static variables, to trace whether IO_OUT is expected. (yylex): emit the new tokens, update in_parens on '(' and ')'. (exp): the print command(s) reworked. (oputput_redir): reworked. (print_expression_list): new non-terminal. (rexp, rexpression_list opt_rexpression_list): nuked. (exp, simp_exp): ``cmd|getline'' rule changed to ``cmd IO_IN getline'' and moved from exp to simp_exp. Unrelated: * awkgram.y (variable): Don't return Node_func, issue a fatal error instead. * eval.c (r_tree_eval, r_get_lhs): Omit special checks for Node_func, nodes of this type cannot get into the program tree. * profile.c (tree_eval, pp_lhs): Likewise. Thu Jan 30 17:42:05 2003 Stepan Kasal ADR: Applied lots of patches from Stepan. * array.c (do_delete_loop): Call after_assign for the loop index. * field.c (do_split): The third argument to split(), sep, has to be evaluated and the result dupnoded before assoc_clear is called, similarily as src. And we needn't to evaluate the third argument if it's CONSTant regex and the first parameter is null string. * awk.h (dupnode): Changed to macro, function renamed to r_dupnode. * node.c (dupnode, r_dupnode): Rename. * awkgram.y (parms_shadow): Return bool value, ... (shadow_funcs): ... which will enable us to end the program if lintfunc is fatal. (program): Cleanup of the rules defining the ``program'' non-terminal. (start, program, rule): no value associated, expression_value is now treated similarily as begin_block and end_block. (pattern, rule): bison actions for non-terminal `pattern' now add a new rule to the appropriate Node_rule_list, action for non-terminal `rule' now only adds the associated code block to the rnode of Node_rule_node. (io_allowed): renamed to !begin_or_end_rule. (append_pattern): New function, adds new Node_rule_node to a rule_list. (mkrangenode): Deleted, this tiny function was called only once. (function_body): non-terminal replaced by `action'. (statements, action, statement): `statements' can now be empty; both callers had to accomodate to this. (statements): Don't call isnoeffect($2->type) if $2 happens to be NULL. Mon Jan 27 14:12:19 2003 Arnold D. Robbins * io.c (iop_close): Based on report by Stepan Kasal and because of his changes, don't call reset_record() when saving a copy of contents of $0. * awkgram.y: Improved function parsing error messages for case where user uses a builtin name as a function name. Based on error report by Stepan Kasal. * ext.c (make_builtin): Set FUNC flag for new function. Based on error report by Stepan Kasal. Mon Jan 27 14:06:20 2003 Stepan Kasal * field.c (reset_record): No longer call set_record(), the code is moved to the function body. Do not set MAYBE_NUM. (set_record): Call reset_record() to perform the common tasks. The prototype has changed, change awk.h and all callers. Mon Jan 27 10:50:03 2003 Arnold D. Robbins * awk.h (NODE): `proc' renamed to `builtin,' to fix a conflict on some systems. Replaced on all spots where it was used. Sun Jan 26 11:52:01 2003 Arnold D. Robbins * awk.h [NUMSUBPATS]: New macro. * builtin.c (do_match): Use it in loop that fills in subpattern info. * eval.c (r_tree_eval): for Node_assign, don't call free_temp(), as assign_val() contains dupnode(), which would clear the TEMP flag. From Stepan Kasal . * config.sub: Updated from prep. Sun Jan 19 22:34:01 2003 Arnold D. Robbins * awk.h (do_asorti): Add declaration. * awkgram.y (tokentab): Add asorti() function to table. * array.c (ASORT_TYPE): New enumerated type for VALUE or INDEX array sorting. (assoc_sort_inplace): New second arg of type ASORT_TYPE. Additional code to rearrange array so rest of merge-sorting works; basically values are tossed and index moved into value spot. (asort_actual): Renamed from do_asort(). Takes new ASORT_TYPE argument. (do_asort): Calls asort_actual(tree, VALUE). (do_asorti): Calls asort_actual(tree, INDEX). * main.c (load_procinfo): Free groupset array when done with it. Thu Jan 16 18:30:50 2003 Arnold D. Robbins * builtin.c (do_match): Revised to provide start and length indices in array 3rd parameter. * config.guess, config.sub: Updated from prep. Thu Jan 2 11:09:12 2003 Arnold D. Robbins Updated to bison 1.875. Tue Dec 31 17:14:45 2002 Arnold D. Robbins Updated things to automake 1.7.2 and autoconf 2.57. Tue Dec 31 16:54:44 2002 Arnold D. Robbins * awk.h [IOP_CLOSED]: New flag. * io.c (iop_close): Set IOP_CLOSED flag. (inrec): Check for IOP_CLOSED; if set return EOF. (rs1_get_a_record, rsnull_get_a_record): Check for EOF before refilling buffers. (rsre_get_a_record): Ditto. Also, set RT before updating pointers in IOP. * Makefile.am (efence): New target to compile with Electric Fence. 2002-12-23 Kaveh R. Ghazi * awk.h (catchsig): Delete prototype. * main.c (catchsig): Make static and remove excess argument. (main): Delete unnecessary casts. * io.c (rs1_get_a_record, rsnull_get_a_record): Mark parameter with ATTRIBUTE_UNUSED. Mon Dec 23 11:54:07 2002 Arnold D. Robbins * regex_internal.h, regex_internal.c, regcomp.c, regexec.c, version.c: Fixed to compile, once again, under K&R compilers. * io.c (grow_iop_buffer): Fix calculation of new size to first subtract 2, double, then add 2 back in. Fri Dec 20 11:48:42 2002 Arnold D. Robbins get_a_record split into three routines. * awk.h (IOBUF): Structure reworked for new code. * io.c (get_a_record): Now a pointer to different functions. (rs1_get_a_record, rsnull_get_a_record, rsre_get_a_record): New functions. (iop_alloc, iop_close): Reworked for new structure. (do_getline, inrec): Modifiend for new EOF condition. (iopflags2str): New routine. Fri Dec 20 11:05:50 2002 Isamu Hasegawa * regex.c, regex_internal.c, regex_internal.h: Changes to allow separate compilation of the reg*c files. * regcomp.c: Fix bug in using translation tables with [[:upper:]] etc. * Makefile.am: Move regex files into sources from EXTRA_DIST. (ADR) Mon Dec 9 14:20:42 2002 Stepan Kasal * main.c (main): When processing option '-f' don't ignore spaces if optarg points at the beginning of the current argument (like ``gawk -f " " file''). 2002-11-30 Kaveh R. Ghazi * awkgram.y (stopme): Mark parameter with ATTRIBUTE_UNUSED. (yyerror): Add ATTRIBUTE_PRINTF_1. * builtin.c (do_systime, do_rand): Likewise. * field.c (set_field, re_parse_field, def_parse_field, posix_def_parse_field, null_parse_field, sc_parse_field, fw_parse_field): Likewise. * io.c (pidopen, useropen): Likewise. * main.c (catchsig): Likewise. * profile.c (init_profiling): Likewise. * awk.h (err): Add ATTRIBUTE_PRINTF. * msg.c (err): Delete redundant prototype. Fix format specifier. Wed Nov 27 06:04:20 2002 Pat Rankin * ext.c [#if !DYNAMIC] (do_ext): Cast string value for error node. Sun Nov 24 18:23:29 2002 Arnold D. Robbins From Paul Eggert, with some edits by me. * builtin.c (do_substr): Consistently use floating point values for lint messages, so they should be printed pretty much as the user saw them. Check for overflow before converting floating point to integer. Do the right thing with NaNs. Check for index out-of-range before checking for length out-of-range, to avoid some nasty effects if address arithmetic overflows (e.g., indx + length < index). Allow zero-length substrings when checking for lint if do_lint == LINT_INVALID. Sun Nov 24 18:21:06 2002 Arnold D. Robbins * awk.h (LINT_ALL, LINT_INVALID): New constants. * main.c (main): Allow --lint=invalid which restricts warning to things that aren't valid. * eval.c (set_LINT): Update setting logic. Wed Nov 20 13:14:58 2002 Arnold D. Robbins * awk.h (lintfunc): Improve ifdef for attribute to only work for GCC 3.2 and later. * io.c (PIPES_SIMULATED): Don't define if on AIX, which does define TANDEM in one of its header files. Ugh. Tue Nov 19 15:33:55 2002 Arnold D. Robbins * builtin.c (do_substr): Use %lu in warnings instead of %d. Mon Nov 18 14:42:53 2002 Arnold D. Robbins * config.guess: Synced from ftp.gnu.org. * config.sub: Ditto. Sun Nov 17 21:32:49 2002 Arnold D. Robbins Updated things to automake 1.7.1. Sun Nov 3 14:33:30 2002 Arnold D. Robbins * eval.c (r_get_lhs): For variables, always clear UNINITIALIZED, since the variable is about to be assigned to. From Stepan Kasal. Fri Nov 1 11:19:01 2002 Arnold D. Robbins * awk.h (lintfunc): Can only supply attributes for a function pointer if GCC >= 3. Added ifdefs. Bah, humbug. 2002-10-30 Kaveh R. Ghazi * array.c (array_init, concat_exp, assoc_find, do_delete): Const-ify. * awk.h (redirect, set_record, pp_func, pp_string_fp, format_val, parse_escape, make_regexp, research, reisstring, remaybelong): Likewise. * awkgram.y (dumpintlstr, dumpintlstr2, func_use, dup_parms, var_comp, finfo, fcompare, func_use, dumpintlstr, dumpintlstr2): Likewise. * builtin.c (stdfile, do_fflush, do_index, category_table): Likewise. * eval.c (push_forloop, push_args, PUSH_BINDING, RESTORE_BINDING, cmp_nodes, op_assign, loop_info, fcall, fmt_ok, set_LINT, comp_func): Likewise. * ext.c (do_ext): Likewise. * field.c (set_record): Likewise. * io.c (gawk_popen, two_way_open, binmode, redirect, getredirect, fatal): Likewise. * node.c (values, format_val, make_str_node, parse_escape): Likewise. * profile.c (pp_string, pp_match_op, pp_func, pp_string, pp_string_fp): Likewise. * re.c (make_regexp, research, reisstring, remaybelong): Likewise. 2002-10-30 Kaveh R. Ghazi * awk.h (__attribute__, ATTRIBUTE_UNUSED, ATTRIBUTE_NORETURN, ATTRIBUTE_PRINTF, ATTRIBUTE_PRINTF_1, __extension__): Define. (emalloc, erealloc): Fix format specifier warnings. (do_nextfile):Mark with ATTRIBUTE_NORETURN. (getredirect): Const-ify. (msg, error, warning, r_fatal, lintfunc): Mark with ATTRIBUTE_PRINTF_1. (r_fatal): Mark with ATTRIBUTE_NORETURN. * builtin.c (format_tree): Fix format specifier warning. * eval.c (interpret): Likewise. * main.c (usage, copyleft, catchsig, nostalgia, version): Mark with ATTRIBUTE_NORETURN. * profile.c (dump_and_exit): Likewise. 2002-10-29 Kaveh R. Ghazi * array.c (array_init): Use ISDIGIT, not isdigit. * awk.h (m_tree_eval, force_number, force_string): Use __extension__ in statement expressions. * main.c (lintfunc): Fix !__SDTC__ case. * regex_internal.c (calc_state_hash): Fix inline declaration. * regexec.c (proceed_next_node): Cast assignment to correct type. 2002-10-29 Kaveh R. Ghazi * awk.h (exp_node, Func_ptr): Add prototype arguments. * awkgram.y (yystype, token, getfname, nextc, pushback, allow_newline, yylex): Likewise. * io.c (wait_any): Likewise. * profile.c (indent_in, indent_out): Likewise. * random.h (random): Likewise. 2002-10-29 Kaveh R. Ghazi * array.c (grow_table): Const-ify. * awk.h (RE_TRANSLATE_TYPE): Define. (flagtab, casetable): Const-ify. (getfname, shadow_funcs, redflags2str): Prototype. (flags2str, genflags2str, nodetype2str, redflags2str, set_loc, msg, error, warning, r_fatal): Const-ify. * awkgram.y (tokentab, snode): Likewise. * builtin.c (format_tree, do_strftime, localecategory_from_argument): Likewise. * eval.c (casetable, nodetypes, nodetype2str, flags2str, genflags2str): Likewise. * io.c (redflags2str, socketopen): Likewise. * main.c (varfile, version_string, lintfunc, optab, copyleft, varinit, init_vars): Likewise. * msg.c (srcfile, msg, warning, error, set_loc, r_fatal): Likewise. * profile.c (pp_op_assign, pp_match_op, pp_redir): Likewise. * random.c (sccsid): Likewise. * version.c, version.in (version_string): Likewise. Tue Oct 29 10:50:52 2002 Arnold D. Robbins * configure.in: Update version in AC_INIT and AM_INIT_AUTOMAKE * fixvers: Make grep for pattern a little smarter. Mon Oct 28 16:35:39 2002 Arnold D. Robbins * awk.h (hash): Now a function pointer. * array.c (gst_hash_string, scramble): New functions. (awk_hash): Renamed from hash. (hash): Now a function pointer. (array_init): Change hash function based on environment for experimentation. Mon Oct 28 13:21:20 2002 Arnold D. Robbins Applied lots of patches from Stepan Kasal, tweaked as needed for current code base. * node.c (dupnode): When n->stref overfows, flag the node as PERM. Same for n->ahname_ref. (unref): Remove the check for n->stref == LONG_MAX and n->ahname_ref == LONG_MAX. * awk.h (make_string): The third argument to make_str_node changed from FALSE to 0, it's not Boolean. (free_temp): Evaluate the argument only once, so that we can call free_temp(tree_eval(n)) for achieving side effects. (load_environ, load_procinfo): Changed return type to NODE *. * main.c (load_environ): The ENVIRON_node should be created with type Node_var_array and lnode set to NULL. Return pointer to the created node and create an empty hash even on TANDEM. (load_procinfo): Same mods for PROCINFO_node. (init_args): ARGV_node should also have lnode set to NULL. * eval.c (r_tree_eval): case Node_assign moved just above the other assignment cases. (op_assign): ++ and -- cases merged with += and -=, respectively. (push_args): Evaluate all args, even in cases where more args are supplied then required. (interpret): In case Node_K_forarray, flag the variable num_elems also as volatile, so that it survives longjmp() and can be trusted when linting code. (r_get_lhs): Case Node_param_list was unreachable (unless something breaks really badly), remove it; (r_tree_eval): case Node_var_array removed from the last switch, it was caught in the first switch above. * profile.c (tree_eval): Again, case Node_var_array was caught above. * awkgram.y (variable): Code simplified, making use of the above changes. * field.c (sc_parse_field): IGNORECASE only applies to regex based field-splitting, so remove code that pays attention to it. (do_split): Don't use parse_field if RS_is_null. (set_FS): Beware of FS == "\\" even if RS_is_null. Code changes to make things work better: * field.c (set_FS): Don't use cmp_nodes() to compare old and new value of FS, that uses IGNORECASE, which is a bad idea. Improve logic for choosing sc_parse_field. Ensure that when RS_is_null but using a single character, that we do pay attention to case when doing regex splitting. * io.c (set_RS): Don't use cmp_nodes() to compare old and new value of RS, that uses IGNORECASE, which is a bad idea. Mon Oct 28 09:43:14 2002 Arnold D. Robbins * recomp.c (parse_expression): Change return statement into two so it'll compile for SGI cc. * awk.h (STR, CUR): Changed to STRCUR and NUMCUR respectively, to avoid conflict with STR on some System V systems. Changed in all source files. Thu Oct 24 16:14:34 2002 Arnold D. Robbins * array.c (AVG_CHAIN_MAX): Now a variable, to allow easy experimentation. (array_init): Pulls a new value from env var AVG_CHAIN_MAX if it exists and sets the variable. * awk.h: Add declaration for array_init(). * main.c (main): Call array_init(). Tue Oct 22 11:23:56 2002 Arnold D. Robbins * bisonfix.sed: Updated for current bison. Death to alloca! 2002-10-21 Isamu Hasegawa * builtin.c (tolower, toupper): Add casts to char* to fix some compiler warnings. * eval.c (cmp_nodes): Ditto. * regcomp.c (peek_token_bracket): Skip the byte already read. Wed Oct 16 15:02:09 2002 Arnold D. Robbins * io.c (set_RS): Make sure to always call set_FS(). 2002-10-11 Isamu Hasegawa * regcomp.c (re_compile_fastmap_iter): Remove the handling OP_CONTEXT_NODE. (regfree): Likewise. (create_initial_state): Likewise. (analyze): Remove the substitutions which became useless. (calc_first): Likewise. (calc_epsdest): Use edests of OP_BACK_REF in case that it has epsilon destination. (duplicate_node_closure): New function. (duplicate_node): Remove the handling OP_CONTEXT_NODE. (calc_inveclosure): Likewise. (calc_eclosure): Likewise. (calc_eclosure_iter): Invoke duplicate_node_closure instead of direct invocation of duplicate_node. (parse): Don't use comma operator in the return to avoid compiler warning. (parse_reg_exp): Likewise. (parse_branch): Likewise. (parse_expression): Likewise. (parse_sub_exp): Likewise. (parse_dup_op): Likewise. * regex_internal.c (re_dfa_add_node): Remove the substitutions which became useless. (create_ci_newstate): Remove the handling OP_CONTEXT_NODE. (create_cd_newstate): Likewise. * posix/regex_internal.h (re_token_type_t): Remove the obsolete type. (re_token_t): Likewise. (re_dfa_t): Likewise. (re_node_set_remove): New macro. * regexec.c (check_matching): Remove the handling OP_CONTEXT_NODE. (check_halt_node_context): Likewise. (proceed_next_node): Likewise. (pop_fail_stack): Fix the memory leak. (set_regs): Likewise. (free_fail_stack_return): New function. (sift_states_backward): Fix the memory leak. Remove the handling OP_CONTEXT_NODE. (update_cur_sifted_state): Append some if clause to avoid redundant call. (sub_epsilon_src_nodes): Use IS_EPSILON_NODE since it might be a back reference. (check_dst_limits): Remove the handling OP_CONTEXT_NODE. (check_subexp_limits): Likewise. (search_subexp): Likewise. (sift_states_bkref): Likewise. (transit_state_mb): Likewise. (transit_state_bkref_loop): Likewise. (transit_state_bkref_loop): Likewise. (group_nodes_into_DFAstates): Likewise. (check_node_accept): Likewise. (sift_ctx_init): Add initializing. Tue Oct 15 14:18:53 2002 Arnold D. Robbins * eval.c (set_IGNORECASE): Call set_RS() instead of set_FS_if_not_FIELDWIDTHS(). The former calls the latter for us, and also makes IGNORECASE affect RS like it's supposed to. * field.c (FS_re_yes_case, FS_re_no_case): New variables. (set_FS): Smarten up routine to not recompile FS_regexp if all that's changed is IGNORECASE or if switching back to FS from FIELDWIDTHS. Significant speed-up for cases where IGNORECASE is assigned to for every record. * io.c (RS_re_yes_case, RS_re_no_case): New variables. (set_RS): Similar changes as to set_FS(). In particular, IGNORECASE changing now affects record splitting too. * re.c (refree): Set rp->pat.tranaslate to NULL. It comes from casetable and shouldn't be freed. (Strictly necessary only for old regex, but a good idea anyway). Also, call regfree(& rp->pat) instead of manually free()ing things, since there's dynamically allocated stuff hiding in the buffer. Avoids a memory leak. Mon Oct 14 12:02:39 2002 Arnold D. Robbins Major space reduction in array management. Overhead reduced to two NODE's per element from three. * awk.h (ahash): Union is gone. (hash.ref): new union member. (ahnext): new definition into hash union. (ahvalue): new definition into hash union. (ahname_str): new member, points into hash union. (ahname_len): new member, points into hash union. (ahname_ref): new member, points into hash union. * array.c: Replaces uses of ahname member with string and length. Set the reference count correctly to 1 on new nodes. * eval.c (interpret): Case for Node_K_arrayfor. dupnode() the array indices, and set loop variable to new value made via make_string(). * node.c (unref, dupnode): Node_ahash nodes are now also reference counted, a la strings. Similar code is used to increment/decrement the counts, and/or copy nodes as needed. Unrelated: * awk.h (forsub): Removed. Not used. Sun Oct 13 16:58:27 2002 Stepan Kasal * profile.c (pprint): #undef the temporary defines at the end of the case. * eval.c (interpret): Likewise. (assign_val): we can unref() before doing dupnode(). Also, move the check for NF < 0 from here ... * field.c (set_NF): ... to here. * main.c (varinit): no need to call set_NF(). * awkgram.y (statements): don't be so generous when concatenating `statements' with a `statement'. 2002-10-13 Isamu Hasegawa * regcomp.c: Synced with development sources. * regex_internal.c: Synced with development sources. * regex_internal.h: Synced with development sources. * regexec.c: Synced with development sources. Sun Oct 13 21:35:35 2002 Arnold D. Robbins * awk.h (NODE): reflags is now unsigned long for: (exec_count): defined to be sub.nodep.reflags. Using `number' broke pgawk. * profile.c (Node_K_delete_loop): print out as a for loop with a comment that it's internally the same as `delete array'. * eval.c (Node_K_delete_loop): Increment the exec_count. Ooops. * configure.in (AM_GNU_GETTEXT_VERSION): New macro call. * custom.h: Updated description of the file at the top. Thu Oct 10 16:39:51 2002 Arnold D. Robbins * awk.h (vname, exec_count): Now macros into different parts of the NODE structure that can be safely used for them. Saves 16 bytes per NODE. * eval.c: Changed use of `vname' to `varname' to avoid new macro. * main.c (lintfunc): Made ifdefed decls match awk.h. * eval.c (comp_func): Use memcmp instead of strcmp. * configure.in (AC_CONFIG_HEADER): Physically append custom.h to config.h to avoid subdir compiliation problems. Sun Oct 6 17:36:15 2002 Arnold D. Robbins Updated to automake 1.7 and bison 1.50. * INSTALL: Replaced with current version from automake 1.7. * config.guess: Replaced with current version from automake 1.7. * config.sub: Replaced with current version from automake 1.7. * depcomp: Replaced with current version from automake 1.7. * doc/texinfo.tex: Replaced with current version from automake 1.7. * install-sh: Replaced with current version from automake 1.7. * missing: Replaced with current version from automake 1.7. * mkinstalldirs: Replaced with current version from automake 1.7. * ylwrap: Replaced with current version from automake 1.7. * configure.in (DYNAMIC): Updated AC_DEFINE(DYNAMIC) to three-argument form for autoheader. * acinclude.m4: Removed includes of jm-mktime.m4 and largefile.m4, which are now standard parts of Autoconf. * Makefile.in: Regenerated. * aclocal.m4: Regenerated. * awkgram.c: Regenerated. * awklib/Makefile.in: Regenerated. * configure: Regenerated. * doc/Makefile.in: Regenerated. * test/Makefile.in: Regenerated. Sun Sep 29 16:47:49 2002 Arnold D. Robbins * custom.h (__WIN32__): Added from gnuwin32 project, via Stepan Kasal. * awkgram.y: For tawk compatibility, added `delete(array)'. To remain undocumented, since it's WAY non-standard. Sun Sep 22 22:23:50 2002 Arnold D. Robbins * awk.h (re_cnt): Removed, not needed since no dfa code. * awkgram.y (regexp, a_regexp): Removed use of re_cnt. * re.c (re_update): Ditto. Thu Sep 19 10:55:37 2002 Arnold D. Robbins * io.c (binmode): Create function if defined(WIN32) also. Updated to gettext 0.11.5, autoconf 2.54 and automake 1.6.3. * aclocal.m4: Regenerated. * m4/codeset.m4: Updated. * m4/gettext.m4: Updated. * m4/glibc21.m4: Updated. * m4/iconv.m4: Updated. * m4/lcmessage.m4: Updated. * m4/lib-ld.m4: Updated. * m4/lib-link.m4: Updated. * m4/lib-prefix.m4: Updated. * m4/progtest.m4: Updated. * po/Makefile.in.in: Updated. * po/Rules-quot: Updated. * po/boldquot.sed: Updated. * po/en@boldquot.header: Updated. * po/en@quot.header: Updated. * po/insert-header.sin: Updated. * po/quot.sed: Updated. * po/remove-potcdate.sin: Updated. Tue Sep 17 23:46:01 2002 Arnold D. Robbins * configure.in: Moved override of INSTALL to just after AC_INIT so that it takes effect. Necessary for Autoconf 2.5x. Mon Sep 16 16:40:57 2002 Stepan Kasal * awkgram.y (want_assign): Removed. (SLASH_BEFORE_EQUAL, ASSIGN): New terminals; ``/='' is now formed from these two. (a_slash): New non-terminal, representing either '/' or SLASH_BEFORE_EQUAL. (assign_operator): New non-terminal, replaces ASSIGNOP. (REGEXP): yylex now eats the terminating '/' before returning REGEXP token. (exp): The check for C-like comments moved from here (regexp): ... to here. (common_exp): New non-terminal; contains common parts of exp and rexp. (a_relop, relop_or_less): New non-terminals. (rexp): some rules updated to be analogous to exp. (output_redir): Can contain only common_exp, not exp in general. Mon Sep 16 22:51:51 2002 Arnold D. Robbins * io.c (two_way_open): Move label use_pipes outsidef of ifdef, just in case. Thu Sep 12 15:11:28 2002 Arnold D. Robbins * awkgram.y (getfname): Return NULL if not found, remove fatal error. Could be an extension function. (dump_funcs): Walk symbol table counting functions before mallocing table, since there could be extension functions, func_count could be too small. * profile.c (pp_builtin): Handle NULL return from getfname(). Print it as "extension_function()" if so. Tue Sep 10 17:33:48 2002 Arnold D. Robbins Minor code simplification. * awk.h (in_array): Change return type to NODE*. (assoc_exists): Remove declaration. * array.c (in_array): Change return type to NODE *. Return value is pointer to element value or NULL. (assoc_exists): Removed function. * eval.c (r_tree_eval): Case Node_in_array, change value to test return of in_array() against NULL. * io.c (pty_vs_pipes): Change test to make a tmp_string() of the index and call in_array(). Add free_temp() of subscript and free() of full_index (oops). 2002-09-10 Isamu Hasegawa * posix/regcomp.c: Wrap #include wchar.h and wctype.h in #if. (build_range_exp): Add castings to strlen invocations. (build_collating_symbol): Restore the type of characters from "char" to "unsigned char", and supplement castings. (build_collating_symbol): Likewise. (build_equiv_class): Likewise. (build_charclass): Likewise. (seek_collating_symbol_entry): Likewise. (parse_bracket_exp): Likewise. (build_word_op): Supplement a casting. * posix/regex_internal.c: Wrap #include wchar.h and wctype.h in #if. (re_string_allocate): Fix castings. (re_string_construct): Likewise. (re_string_construct_common): Likewise. (re_string_realloc_buffers): Likewise. (build_wcs_buffer): Likewise. (build_wcs_upper_buffer): Likewise. (re_string_skip_chars): Likewise. (re_string_reconstruct): Likewise. * posix/regex_internal.h: Restore the type of characters in re_string_t and bracket_elem_t from "char" to "unsigned char". (re_string_elem_size_at): Fix castings. * posix/regexec.c: Wrap #include wchar.h and wctype.h in #if. (transit_state_bkref_loop): Restore the type of characters from "char" to "unsigned char", and append a cast to "char*" pointer in array subscript. (check_node_accept_bytes): Likewise. (find_collation_sequence_value): Likewise. Thu Sep 5 13:15:09 2002 Arnold D. Robbins * re.c (remaybelong): New routine. (reisstring): Simplified the code a bit. * awk.h (remaybelong): Declaration added. * io.c (get_a_record): Change fourth grungy special case to use remaybelong() instead of strchr() on last character. Wed Sep 4 13:20:26 2002 Arnold D. Robbins * io.c (do_input): Recode guts of main loop to be easier to trace with a debugger. (get_a_record): Fourth grungy special case for RE-based record splitting added. See explanatory comments there and test/rebuf.awk. 2002-09-03 Isamu Hasegawa * posix/regcomp.c (regcomp): Append "__restrict" modifier to avoid warnings of some compilers. (build_collating_symbol): Change the type of characters from "unsigned char" to "char", and append a cast to "char*" pointer in array subscript. (build_collating_symbol): Likewise. (build_equiv_class): Likewise. (build_charclass): Likewise. (re_compile_pattern): Remove incorrect cast. (re_compile_fastmap_iter): Change the type of characters from "unsigned char" to "char", and append a cast to "char*" pointer in array subscript. (parse_bracket_exp): Likewise. * posix/regex_internal.c (re_string_construct_common): Likewise. (re_string_allocate): Likewise. (re_string_construct): Likewise. (re_string_realloc_buffers): Likewise. (build_wcs_buffer): Likewise. (re_string_reconstruct): Likewise. * posix/regex_internal.h: Change the type of characters in re_string_t and bracket_elem_t from "unsigned char" to "char". * posix/regexec.c (regexec): Append "__restrict" modifier to avoid warnings of some compilers. (transit_state_bkref_loop): Change the type of characters from "unsigned char" to "char", and append a cast to "char*" pointer in array subscript. (check_node_accept_bytes): Likewise. (find_collation_sequence_value): Likewise. Wed Aug 21 15:40:36 2002 Corinna Vinschen * configure.in: Define --without-libintl-prefix and --without-libiconv-prefix for Cygwin by default. * Makefile.am: Call fixvers from $(srcdir). * awk.h: Don't define O_BINARY on Cygwin. Wed Aug 21 15:31:57 2002 Andreas Buening * configure.in (AC_OBJEXT, AC_EXEEXT): Added. Removed OS/2 goo. * Makefile.am (check-local): Add $(EXEEXT) suffixes, remove OS/2 goo. * regcomp.c, regex_internal.c, regexec.c: Conditionalize include of and on RE_ENABLE_I18N. Wed Aug 21 14:43:57 2002 Arnold D. Robbins * gettext.h (ENABLE_NLS): Add include of locale.h so that things compile even without optimization. Sheesh. * io.c (two_way_open, pty_vs_pipes): Conditionalize pty code on HAVE_TERMIOS_H. Thu Aug 8 22:16:10 2002 Arnold D. Robbins * main.c (main): Force LC_NUMERIC locale to "C", esp. for M$ systems. Ugh. Wed Aug 7 13:42:01 2002 Arnold D. Robbins * io.c (get_a_record): Improve test for newlines at beginning of record but with nothing following it. See test/nulrsend. Mon Aug 5 10:12:39 2002 Arnold D. Robbins Add option to use ptys instead of pipes for |&. Basic plumbing originally from Paolo Bonzini . * awk.h (RED_PTY): New flag. (assoc_exists): Add declaration. * array.c (in_array): Use FALSE not zero for return value. (assoc_exists): New routine to find and return value for an index in an array. * configure.in: test for termios.h and stropts.h, and grantpt function. * io.c: include termios.h and stropts.h if available. (redflags2str): Add RED_PTY to table. (redirect): Add RED_PTY to flags turned off when searching. (close_redir): close write channel for two-way pipes that use ptys by sending an EOF. (two_way_open): If pty_vs_pipe(), use pty's to open two-way pipes as they are line-buffered by default --> alleviates deadlock problems. If fails, fall back to using pipes. (pty_vs_pipe): New function. * main.c (arg_assign): Clean up English in some of the error messages. Sun Aug 4 00:37:38 2002 Stepan Kasal * re.c (make_regexp): don't pass the error message returned by re_compile_pattern() to gettext(); it's already gettextized. (make_regexp): minor reformat of code. Wed Jul 31 23:50:31 2002 Arnold D. Robbins Removed dfa code from gawk since not really needed with new regex. * Makefile.am: Removed dfa.h and dfa.c. * awk.h (struct Regexp): Removed `dfareg' and `dfa' members. (make_regexp): Last parameter in function went away, changed decl. (avoid_dfa): Removed declaration. * awkgram.y: Fixed call to make_regexp(). * eval.c (match_op): Simplified: removed call to avoid_dfa() and `kludge_need_start' variable. Instead, pass FALSE as last parameter of research(). * field.c (set_FS): Fixed call to make_regexp(). * io.c (get_a_record, set_RS): Fixed calls to make_regexp(). * re.c (make_regexp): Removed last paramter (`dfa') from function. Simplified the code. (research): Simplified the code, removed calls to dfa stuff. (dfaerror): Removed function. (re_update): Fixed call to make_regexp(). (avoid_dfa): Removed function. Thu Jul 25 21:55:45 2002 Arnold D. Robbins * regcomp.c, regex_internal.c, regex_internal.h, regexec.c: Bug fixes from Isamu Hasegawa and Stepan Kasal applied. Sat Jul 6 23:28:37 2002 Arnold D. Robbins * awkgram.y (yyerror): Change text of unexpected newline message to include end of string. Mon Jun 17 17:58:55 2002 Arnold D. Robbins * field.c (do_split): Per Michal Jaegermann, move free_temp(fs) above label `out'. Tue Jun 11 23:26:09 2002 Paul Eggert Update to autoconf 2.53 and automake 1.6.1. * acconfig.h: Removed. * m4/isc-posix.m4: Removed. * m4/jm-mktime.m4: Removed. * m4/largefile.m4: Removed. * m4/ssize_t.m4: Removed. * ansi2knr.c: updated. * depcomp: updated. * install-sh: updated. * missing: updated. * mkinstalldirs: updated. * ylwrap: updated. * configure.in: Improved quoting. * acinclude.m4: Use `m4_sinclude', not antiquated `sinclude'. Tue Jun 11 23:08:40 2002 Arnold D. Robbins * configure.in: Add `getgrent' to list of functions checked so that awklib/grcat is compiled correctly. Tue Jun 11 22:18:42 2002 Stepan Kasal Improve argument parsing and -v assignment. * awk.h (struct src): Add additional enum values. (arg_assign): Return type and arg list changes. * io.c (nextfile): Add extra arg in call to `arg_assign'. * main.c (pre_assign): Nuked. (allocfiles): New variable. (srcfiles_add, preassigns_add): New macros. (main): Logic cleaned up. (add_src): New function. Use `size_t' for optimal_bufsize function. * awkgram.y (yylex): `len' is now size_t. * pc/gawkmisc.pc (optimal_bufsize): Change return type to size_t. * posix/gawkmisc.c (optimal_bufsize): Change return type to size_t. * unsupported/atari/gawkmisc.atr (optimal_bufsize): Change return type to size_t. * unsupported/tandem/tmisc.c (optimal_bufsize): Change return type to size_t. * vms/gawkmisc.vms (optimal_bufsize): Change return type to size_t. * README_d/README.hpux: New file. Fri May 24 12:23:01 2002 Arnold D. Robbins * profile.c (init_profiling): Remove default initialization of `prof_fp' to stderr. Per Stepan Kasal . Wed May 15 15:39:17 2002 Arnold D. Robbins Work through builtin operations to make sure that anything that might have side effects gets dealt with. * array.c (do_delete): Evaluate subscript first before checking if something is or isn't an array. * builtin.c (sub_common): Evaluate replacement text, and free it if no match of regex in source text. Wed May 15 15:30:34 2002 Arnold D. Robbins Switch to new version of regex from IBM Japan. * regcomp.c: New file. * regex.c: Replaced with new version. * regex.h: Replaced with new version. * regex_internal.c: New file. * regex_internal.h: New file. * regexec.c: New file. * Makefile.am (EXTRA_SOURCES): New files added. Tue May 14 17:04:05 2002 Arnold D. Robbins * awk.h (): Move check and include into gettext.h. * gettext.h (): Add check and include per patch from Bruno Haible. * field.c (do_split): When checking for split of null string, evaluate seperator if it's not FS, since could have side effects. At end, free_temp(fs), not free_temp(sep). Both of these thanks to Stepan Kasal . Mon May 13 00:41:31 2002 Arnold D. Robbins * custom.h (ultrix): Add define GETGROUPS_NOT_STANDARD. * main.c (init_groupset): For GETGROUPS_NOT_STANDARD, use old way to set `ngroups'. 2002-05-10 Andreas Schwab * dfa.c (parse_bracket_exp_mb): Fix warning. Thu May 9 22:28:32 2002 Arnold D. Robbins * builtin.c (sub_common): Fix logic for match of null strings to get correct semantics. See test/gsubtst2.*. * field.c (do_split): Minor code cleanup; the third arg to split() is set to be FS by the grammar, so don't need to check it for NULL. Thanks to Stepan Kasal . * awk.h (locale.h): Move include before that of "gettext.h" for systems that define functions that gettext.h would use when NLS is disabled. Per bug report from Ayamura Kikuchi . Tue May 7 17:31:01 2002 Arnold D. Robbins Miscellanious patches courtesy of Stepan Kasal . * field.c, main.c: Tidy up some comments. * field.c (set_FIELDWIDTHS): Init fw_alloc to 4 so it isn't immediately realloced. * main.c (load_procinfo): Check value of FS/FIELDWIDTHS for value of PROCINFO["FS"]. * awk.h (set_FS_if_not_FIELDWIDTHS): Removed decl. * field.c (set_FS_if_not_FIELDWIDTHS): Removed function. * eval.c (set_IGNORECASE): Use inline code checking `using_fieldwidths()'. * io.c (set_IGNORECASE): Ditto. Sun May 5 14:28:34 2002 Arnold D. Robbins Fix a memory leak in array for loops if the body contains a `next' or `nextfile' statement. The changes maintain a stack of active for loops that is pushed and popped for each loop, and popped entirely for `next', `nextfile', etc. * eval.c (forloops_active, pop_forloop, pop_all_forloops, push_forloop): new functions. (interpret): Case Node_K_arrayfor, call push and pop functions. Case Node_rule_list: pop loops and pop fcalls after longjmp. Cases Node_K_next, Node_K_nextfile, Node_K_break and Node_K_continue, removed check before longjmp. Case Node_K_exit: add loop check. (loop_stack, nloops, nloops_active): New variables that implement the stack. Wed May 1 16:41:32 2002 Arnold D. Robbins * Release 3.1.1: Release tar file made. Wed May 1 16:07:49 2002 Arnold D. Robbins * getopt.c: Installed latest version from glibc. Sun Apr 28 17:19:07 2002 Arnold D. Robbins * fixvers: Changed patterns to allow test versions of the form `gawk-3.1.1a'. * patchlev.h: Patchlevel is now a string constant. * main.c (version): Print patchlevel using %s, not %d. * Makefile.am: Rework DEFPATH stuff and datadir stuff yet again. * config.sub: Updated with current version from ftp.gnu.org. * config.guess: Ditto. Upgrade to gettext-0.11.2: * ABOUT-NLS: Replaced with version from gettext 0.11.2. * config.rpath: Replaced with version from gettext 0.11.2. * intl/*: Replaced with version from gettext 0.11.2. * po/Makefile.in.in: Replaced with version from gettext 0.11.2. * po/Makevars.template: Replaced with version from gettext 0.11.2. * po/Rules-quot: Replaced with version from gettext 0.11.2. * po/boldquot.sed: Replaced with version from gettext 0.11.2. * po/en@boldquot.header: Replaced with version from gettext 0.11.2. * po/en@quot.header: Replaced with version from gettext 0.11.2. * po/insert-header.sin: Replaced with version from gettext 0.11.2. * po/quot.sed: Replaced with version from gettext 0.11.2. * po/remove-potcdate.sin: Replaced with version from gettext 0.11.2. * m4/codeset.m4: Replaced with version from gettext 0.11.2. * m4/gettext.m4: Replaced with version from gettext 0.11.2. * m4/glibc21.m4: Replaced with version from gettext 0.11.2. * m4/iconv.m4: Replaced with version from gettext 0.11.2. * m4/isc-posix.m4: Replaced with version from gettext 0.11.2. * m4/lcmessage.m4: Replaced with version from gettext 0.11.2. * m4/lib-ld.m4: Replaced with version from gettext 0.11.2. * m4/lib-link.m4: Replaced with version from gettext 0.11.2. * m4/lib-prefix.m4: Replaced with version from gettext 0.11.2. * m4/progtest.m4: Replaced with version from gettext 0.11.2. Wed Apr 17 15:09:45 2002 Arnold D. Robbins * regex.c (PREFIX): Change test for token concatenation ability to `#ifdef HAVE_STRINGIZE'. If a cpp has one, it ought to have the other. Tue Apr 16 12:26:06 2002 Arnold D. Robbins * profile.c (tree_eval): Make unary minus case smarter, use is_scalar test and if false parenthesize expression. Add Node_TEXTDOMAIN case. (pp_lhs, is_scalar, prec_level): Add Node_TEXTDOMAIN cases. Thu Apr 11 21:28:33 2002 Arnold D. Robbins * array.c (do_adump): Spelling fix in output message. * builtin.c: Ditto, in multiple routines. (do_toupper, do_tolower): Add cast to size_t in assigment to mbclen for some compilers. * re.c (research): Fix way returning is done to silence some compiler diagnostics. Wed Apr 10 19:30:51 2002 Arnold D. Robbins * Makefile.am (datadir): Set directly to have `/awk'. (DEFPATH): Go back to using $(datadir) for path. Tue Apr 9 17:34:09 2002 Arnold D. Robbins Upgraded to gettext 0.11.1. * Makefile.am (LDADD): Use @LIBINTL@ instead of @INTLLIBS@. * ABOUT-NLS: Version from 0.11.1. * config.rpath: Version from 0.11.1. * aclocal.m4: Regenerated based on new files. * intl/*: Replaced with version from 0.11.1. * m4/ChangeLog: New file. * m4/codeset.m4: New file. * m4/gettext.m4: Version from 0.11.1. * po/ChangeLog: New file. * po/Makefile.in.in: Version from 0.11.1. * po/remove-potcdate.sin: New file. Mon Apr 8 22:22:58 2002 Arnold D. Robbins * Makefile.am (libexecdir): Set directly to have `/awk'. (DEFPATH): Use $(pkgdatadir) for path. (install-exec-hook): Add version link for pgawk. (uninstall-links): Remove pgawk version link. Wed Mar 20 13:44:21 2002 Isamu Hasegawa * regex.c (__alignof__): Definition for non-GCC compilers. Sun Mar 17 17:41:55 2002 Arnold D. Robbins * io.c (do_pathopen): Malloc buffers to hold constructed filenames: No Arbitrary Limits! Thanks to keoki@techie.com for the bug report. Sun Mar 10 16:59:06 2002 Scott Deifik * awk.h (LOCALEDIR): Provide a definition in case not using i18n stuff. Wed Mar 6 18:14:44 2002 Arnold D. Robbins * main.c (usage): Add some explanatory text and examples at end. Sun Mar 3 16:42:50 2002 Arnold D. Robbins * getopt.h, getopt.c, getopt1.c: Update to current version from glibc CVS. Fri Feb 22 15:53:38 2002 Isamu Hasegawa * dfa.c (fetch_wc): Fix type from wchar_t to wint_t. (parse_bracket_exp_mb): Likewise. * regex.c (extract_number): Retrieve the sign information from byte-code in case of AIX. Thu Feb 21 16:44:24 2002 Arnold D. Robbins * re.c (resetup): Moved setting re_max_failures into regex.c. * regex.c (re_max_failures): Set to really big if REGEX_MALLOC defined. Do this in both places that define re_max_failures. Thu Feb 21 19:02:22 2002 Isamu Hasegawa * builtin.c (sub_common): Avoid index_multibyte_buffer invocation in single byte character environments. Thu Feb 21 10:08:56 2002 Isamu Hasegawa * dfa.c (parse_bracket_exp_mb): For ':', use wctype_t in MALLOC, not wchar_t. Thu Feb 21 09:52:16 2002 Arnold D. Robbins Upgraded to automake 1.5 and gettext-0.11. Also bug fix to multibyte code. * ABOUT-NLS: upgraded. * config.guess, config.sub, config.rpath, gettext.h, ylwrap: new files. * Makefile.am: added above to appropriate places. * awk.h: Replace libintl.h and macros with include of gettext.h. (emalloc, erealloc): Add num bytes to error message, put string inside _(). (index_multibyte_buffer): Removed decl. * awklib/Makefile.am: Use $(EXEEXT) for grcat and pwcat targets. * builtin.c (index_multibyte_buffer): Made static to this file. (sub_common): Add checks that replacement string is length > 0 so that we don't try to malloc(0): this fails on some systems. * configure.in (AM_GNU_GETTEXT): Update macro for gettext 0.11. (ALL_LINGUAS): Removed. * m4/codeset.m4: new file. * m4/gettext.m4: updated. * m4/glibc21.m4: new file. * m4/iconv.m4: new file. * m4/isc-posix.m4: new file. * m4/lcmessage.m4: updated. * m4/lib-ld.m4: new file. * m4/lib-link.m4: new file. * m4/lib-prefix.m4: new file. * m4/progtest.m4: updated. * intl/*: Replaced with version from gettext 0.11. * po/*: Revised for gettext 0.11. Mon Feb 18 14:42:39 2002 Arnold D. Robbins * builtin.c (nondec2awknum): Change assert to runtime check in case user passed in bad data. 2002-02-17 Paul Eggert * re.c (resetup): Try to avoid silly limitation of regex.c by setting re_max_failures to the largest reasonable value. Sun Feb 17 14:57:43 2002 Arnold D. Robbins * builtin.c (research): If re_search() returns -2, the match failed since regex couldn't allocate enough memory for what it needed. Fail with a fatal message instead. This is a workaround, not a fix, but I don't mess with regex.[ch]. Fri Feb 8 16:01:11 2002 Arnold D. Robbins * awkgram.y (LEX_FOR): Fix case of array loop with body of single delete statement to actually check the right things to make the optimization. * profile.c (tree_eval): Add case for Node_K_delete_loop. (prec_level): Ditto. Mon Feb 4 10:38:00 2002 Bruno Haible * awk.h (dcngettext): New macro. (do_dcngettext): New declaration. * awkgram.y (tokentab): Add dcngettext. (snode): Add a warning for incorrect use of dcngettext. (dumpintlstr): fflush at the end, not in the middle. (dumpintlstr2): New function. * builtin.c (localecategory_from_argument): New function, extracted from do_dcgettext. (do_dcgettext): Call it. (do_dcngettext): New function. Sun Feb 3 17:56:20 2002 Bruno Haible * builtin.c (do_bindtextdomain): Don't free the same variable twice. * main.c (main): Call setlocale for LC_MESSAGE, to make dcgettext function work on glibc systems. Wed Jan 23 15:03:36 2002 Andreas Buening * configure.in (PATH_SEPARATOR): Code added for OS/2. Makefile.am (PATH_SEPARATOR): Added. (DEFPATH): Make use of PATH_SEPARATOR. Wed Jan 23 14:46:04 2002 Arnold D. Robbins * awkgram.y (yylex): Add test for lasttok != '$' when looking at _"...". See comments in code. Wed Aug 15 07:43:10 2001 Isamu Hasegawa * regex.c : Implements the codes for exactn_bin to work correctly in multibyte environments, in case of invalid multibyte sequence. Wed Aug 15 07:36:56 2001 Isamu Hasegawa * regex.c : Implements the codes for charset/charset_not to work in multibyte environments. Wed Aug 15 05:04:34 2001 Isamu Hasegawa * regex.c : Add some comments. Wed Aug 15 05:04:15 2001 Isamu Hasegawa * regex.c (count_mbs_length): New function, check the mutibyte strings and count how many wchar_t the substring occupy. (CHAR_T): New macro, character type depending on environments(singlebyte/multibyte). (UCHAR_T): New macro, unsigned character type. (COMPILED_BUFFER_VAR): New macro, the buffer containing the compiled buffer. Adapt singlebyte/multibyte environments with CHAR_T, UCHAR_T, and COMPILED_BUFFER_VAR. Mon Jun 25 09:00:41 2001 Isamu Hasegawa * regex.c : Reorganize code to build code twice. byte_* are for single byte, wcs_* are for multibyte character sets. Chose functions according to current locale dynamically. * regex.c (convert_mbs_to_wcs): New function, convert multibyte strings to wide character strings for multibyte environments. Fri Jun 22 05:43:50 2001 Isamu Hasegawa * regex.c (MBS_SUPPORT): New macro, defined if the environment can handle multibyte characters. (OFFSET_ADDRESS_SIZE): Offset address size in the compiled buffer. Rewrite offset addresses with OFFSET_ADDRESS_SIZE. Thu Apr 26 08:03:17 2001 Isamu Hasegawa * builtin.c (index_multibyte_buffer) : inspect the buffer and write the index. (sub_common) : in multibyte environment, skip multibyte characters when we check special characters. * awk.h (index_multibyte_buffer) : add prototype. * eval.c (cmp_nodes) : in multibyte environment, compare per character. * field.c (re_parse_field) : in multibyte environment, avoid to call research() on invalid boundary. (sc_parse_field) : in multibyte environment, avoid to compare on invalid boundary. (null_parse_field) : in multibyte environment, split per character, not per byte. * io.c (get_a_record) : in multibyte environment, avoid to compare on invalid boundary. Wed Apr 25 08:29:47 2001 Isamu Hasegawa * awk.h (strncasecmpmbs) : add prototype. * builtin.c (strncasecmpmbs) : new function like strncasecmp but for multibyte strings. (do_index) : in multibyte environment, compare per character. * builtin.c (do_tolower) : in multibyte environment, user towlower instead of TOLOWER. (do_toupper) : in multibyte environment, user towupper instead of TOUPPER. Tue Apr 24 10:38:06 2001 Isamu Hasegawa In multibyte environments, handle multibyte characters as single characters in bracket expressions. * dfa.h (mb_char_classes) : new structure. (mbcsets) : new variable. (nmbcsets) : new variable. (mbcsets_alloc) : new variable. * dfa.c (prtok) : handle MBCSET. (fetch_wc) : new function to fetch a wide character. (parse_bracket_exp_mb) : new function to handle multibyte character in lex(). (lex) : invoke parse_bracket_exp_mb() for multibyte bracket expression. (atom) : handle MBCSET. (epsclosure) : likewise. (dfaanalyze) : likewise. (dfastate) : likewise. (match_mb_charset) : new function to judge whether a bracket match with a multibyte character. (check_matching_with_multibyte_ops) : handle MBCSET. (dfainit) : initialize new variables. (dfafree) : free new variables. Mon Apr 23 01:40:09 2001 Isamu Hasegawa Implement the mechanism to match with multibyte characters, and use it for `period' in multibyte environments. * dfa.h (mbps) : new variable. * dfa.c (prtok) : handle ANYCHAR. (lex) : use ANYCHAR for `period' in multibyte environments. (atom) : handle ANYCHAR. (state_index) : initialize mbps in multibyte environments. (epsclosure) : handle ANYCHAR. (dfaanalyze) : handle ANYCHAR. (dfastate) : handle ANYCHAR. (realloc_trans_if_necessary) : new function. (transit_state_singlebyte) : new function. (match_anychar) : new function. (check_matching_with_multibyte_ops) : new function. (transit_state_consume_1char) : new function. (transit_state) : new function. (dfaexec) : invoke transit_state if expression can match with a multibyte character in multibyte environments. (dfamust) : handle ANYCHAR. Fri Apr 20 11:31:24 2001 Isamu Hasegawa Avoid incorrect state transition in multibyte environments. * dfa.h (nmultibyte_prop) : new variable. (multibyte_prop) : new variable. * dfa.c (addtok) : set inputwcs. (dfastate) : avoid incorrect state transition in multibyte environments. (dfaexec) : likewise. (dfainit) : init multibyte_prop. (dfafree) : free multibyte_prop. (inputwcs): new variable. (mblen_buf) : new variable contains the amount of remain byte of corresponding multibyte character in the input string. Fri Apr 20 06:28:59 2001 Isamu Hasegawa Handle a multibyte character followed by '*', '+', and '{n,m}' correctly. * dfa.c (update_mb_len_index): new function. Support for multibyte string. (FETCH) : call update_mb_len_index. (lex) : check cur_mb_index not to misunderstand multibyte characters. (atom) : make a tree from a multibyte character. (dfaparse) : initialize new variables. (mbs) : new variable. (cur_mb_len) : new variable. (cur_mb_index) : new variable. Thu Apr 19 09:32:47 2001 Isamu Hasegawa * awkgram.y (cur_mbstate) : new varialble containing means current shift state. (cur_char_ring) : new varialbe reffering the buffer which contains last some character from the buffer. (cur_ring_idx) : new variable containing the current index on cur_char_ring. (nextc_is_1stbyte) : new macro, means that last nextc() return a singlebyte character or 1st byte of a multibyte character. (nextc) : check the buffer and update cur_ring_char in multibyte environments. (pushback) : adjust cur_ring_idx in multibyte environments. (yylex) : add check whether nextc() returned 1st-byte in multibyte environments. * re.c (make_regexp) : in multibyte environment, skip multibyte characters when we check special characters. Wed Apr 18 07:58:20 2001 Isamu Hasegawa * awk.h (MBS_SUPPORT) : New flag, means supporting multibyte strings. * configure.in : add check for wchar.h, wctype.h, mbrtowc, and mbrlen. Wed Jan 16 16:32:40 2002 Arnold D. Robbins * builtin.c (do_strtonum): Simplified. Check first if the value matches a non-decimal number, and if so convert it. Otherwise do a regular force_number. Mon Jan 7 22:12:15 2002 Arnold D. Robbins * awkgram.y (statement): moved delete, print, and expressions into new non-terminal `simple_stmt'. Allow opt_simple_stmt in the first and third part of a for loop, per latest POSIX, which documents an otherwise undocumented historical oddity in Unix awk. This has the pleasant side effect of making line numbers more accurate for messages involving delete statements. (opt_simple_stmt, simple_stmt): new non-terminals. Based on bug report from drj@pobox.com. Mon Dec 24 14:04:02 2001 Arnold D. Robbins * configure.in: Changes for VMS with new strftime: (AC_HEADER_TIME): added (AC_CHECK_HEADERS): Check for sys/time.h. (TIME_T_IN_SYS_TYPES_H): Add header check. * acconfig.h (TIME_T_IN_SYS_TYPES_H): Added. Wed Dec 19 16:01:58 2001 Peter J. Farley III * configure.in: Add MS-DOS to getpgrp special case. * dfa.c, getopt.c, regex.c: Fix code to work with --disable-nls. Wed Dec 19 15:59:25 2001 Eli Zaretskii * profile.c (init_profiling_signals) [__DJGPP__]: Use SIGINT instead of SIGHUP and SIGQUIT instead of SIGUSR1. Tue Dec 18 20:56:07 2001 Andreas Buening More OS/2 stuff. * awk.h (O_BINARY): Don't redefine for EMX. * io.c (gawk_popen): Add __EMX__ in case compiling DOS executable. * configure.in: Add OS/2 to case for manual GETPGRP_VOID. Tue Dec 4 17:54:30 2001 Arnold D. Robbins New configure time option, --with-whiny-user-strftime. * configure.in (AC_ARG_WITH): Add appropriate code for autoconf. * accondig.h (USE_INCLUDED_STRFTIME): add #undef for it. * custom.h (USE_INCLUDED_STRFTIME): set things up write. Tue Dec 4 16:44:07 2001 Andreas Buening Mongo patch for updated OS/2 support. * awk.h (TOUPPER, TOLOWER): Define only if not already defined. * awkgram.y (extproc feature): Add ifdef for __EMX__. * gawkmisc.c (__EMX__): include pc/gawkmisc.c directly. * io.c (__EMX__): Added for a number of places in addition to OS2 def. (two_way_open): Added OS/2 specific code added that uses spawn. (gawk_popen): ditto. Mon Dec 3 14:07:56 2001 Arnold D. Robbins Fix use of getgroups to use dynamic memory, solves problem of systems where NGROUPS_MAX lies. * awk.h (groupset, ngroups): New extern variables. * configure.in (AC_CHECK_FUNCS): Add getgroups to list. * io.c (user_open): Use global ngroups and groupset variables, don't call getgroups here. * main.c (init_groupset): New function to init global vars using malloc. Declare it at top. (main): Call init_groupset(). (load_procinfo): Use global ngroups and groupset variables. Sun Nov 18 11:56:01 2001 Arnold D. Robbins * random.c (srandomdev): ifdef-out. Lots of compile time problems on multiple platforms, and gawk doesn't even use the routine. The heck with fine-grained solutions. Wed Nov 14 16:12:40 2001 Pat Rankin * builtin.c (bchunk_one): Use `ofre < 1' instead of `ofre <= 0' to avoid compiler complaint about suspicious comparison for unsigned variable. (`ofre == 0' ought to suffice...) Tue Nov 13 17:27:52 2001 Arnold D. Robbins * awkgram.y (yyerror): Fix the code to behave like it used to. Keep "no arbitrary limits" by mallocing the buffer and freeing it. Wed Nov 7 16:46:20 2001 Arnold D. Robbins * awkgram.y (yyerror): Remove dependency upon buf[] to hold prepended space and `^' pointer. Avoids core dumps for long source lines. Sat Nov 3 22:27:21 2001 Arnold D. Robbins * m4/strtod.m4: Add missing `#endif'. Oops. Mon Oct 29 14:53:57 2001 Arnold D. Robbins * awkgram.y: Add semicolons in calls to count_args(). Apparently bisoon adds a semicolon to each body automatically and byacc doesn't. Sun Oct 28 16:53:18 2001 Arnold D. Robbins * builtin.c (format_tree): Fix off-by-one error in "ran out for this one" diagnostic. Also fix lint check for too many arguments vs. count in format string. Wed Oct 10 11:01:47 2001 Arnold D. Robbins * fixvers: Check that files exist before doing `cmp', in case they're in a source code system and aren't there. Fix from Grant Erickson (gerickson@brocade.com). Thu Oct 4 18:20:36 2001 Arnold D. Robbins * eval.c (r_tree_eval): For comparison, dupnode() results of evaluation so that we can hang on to them and avoid memory corruption. Change calls to free_temp() to unref(). Tue Sep 25 15:19:53 2001 Arnold D. Robbins * io.c (iop_open): Only call os_close_on_exec() for fd > fileno(stderr). 2001-09-07 Paul Eggert * io.c (redirect): When deciding to use the fdopen bug hack, use "__sun" rather than "solaris". No compilers predefine "solaris", but both GCC and Sun C predefine "__sun". Thu Aug 30 15:17:12 2001 Arnold D. Robbins * main.c (copyleft): Use a printf %d for last year of update to avoid translation strings changing when the file is updated from now on. Suggestion from Ulrich Drepper. Thu Aug 23 14:01:14 2001 Arnold D. Robbins * awkgram.y (unary minus production): Add check that value isn't a string. Based on bug report from drj@pobox.com. * profile.c (tree_eval): For node_val, only test NUMBER to see if value is numeric, not NUM|NUMBER. Thu Aug 16 12:21:28 2001 Arnold D. Robbins * configure.in (ALL_LINGUAS): Added `fr' and `tr'. * po/fr.po, po/tr.fo: New files. 2001-08-13 Paul Eggert This patch fixes a bug that causes gawk to rewind standard input incorrectly. It also removes all instances of fseek, from the gawk source proper, which should make gawk a bit more portable. (The original patch removed off_t & lseek too, but I need that for something else. ADR.) * posix/gawkmisc.c (optimal_bufsize): Don't use lseek on the input, because that might change its state. Instead, just check whether it is a regular file. This obviates the need to invoke isatty. (Also, fix a spelling error in the first line of the source.) * pc/gawkmisc.pc, unsupported/atari/gawkmisc.atr: Likewise. * awk.h (S_ISREG): Move this macro here ... * io.c (S_ISREG): from here. * protos.h (fseek): Remove prototype; no longer used. Fri Aug 3 13:38:54 2001 Arnold D. Robbins * array.c (assoc_lookup): Change assert test on type to real test to protect against FS[1] = "x" kinds of things. It'd be better to do this in the grammar, but this is easier and just as effective. Undid BECAMEARRAY changes of 25 June 2001 in favor of correct code: * eval.c (pop_fcall): Change test and comment for freeing n->vname. (flags2str): removed BECAMEARRAY entry. * awk.h (BECAMEARRAY): Removed define. * array.c (assoc_lookup): Removed setting of BECAMEARRAY flag. Mon Jul 23 17:33:13 2001 Arnold D. Robbins * io.c (get_a_record): Handle case where RS = "" and input file is only newlines. See test/onlynl. Bug report by Michel Jouvin . Wed Jul 4 18:34:19 2001 Arnold D. Robbins * eval.c (assign_val): Don't allow negative values for NF. * field.c (set_NF): Robustify field-freeing code to make sure values are always positive. Sun Jul 1 19:15:01 2001 Arnold D. Robbins * builtin.c (do_index): If second string is "", return 1. Mon Jun 25 19:34:24 2001 Arnold D. Robbins Further rationalization of treatment of dynamic regexes, so that profiling code works correctly. * awk.h (NODETYPE): New type, Node_dynregex. * awkgram.y (mk_rexp): Use Node_dynregex. * eval.c (nodetypes): Add Node_dynregex. (r_tree_eval): Add Node_dynregex to case for match_op(). * profile.c (tree_eval): Add Node_dynregex to case for pp_match_op(). (pp_match_op): Handle Node_dynregex, simplify cases for ~ and !~. * re.c (re_update): Add assertion that type is Node_regex when flags indicate CONST. New lint warning. * awkgram.y (yylex): Added lint warning that constant with leading zero is treated as octal or hex. Generalized code for those who are Strong In The Ways of the Source. * awk.h: New boolean variable. * main.c (main): set it. * eval.c (interpret): For arrays, check it. Remove variable 'first', not needed anymore. * profile.c (pp_string_fp): Enable printing of non-ascii characters verbatim if variable set. Fix memory corruption on SCO for array vars as params changed globally. * awk.h (BECAMEARRAY): new flag. * array.c (assoc_lookup): set the flag as appropriate. * eval.c (flags2str): Add the flag. (pop_fcall): Check the flag, don't free memory if set. Wed Jun 13 18:07:06 2001 Arnold D. Robbins * eval.c (fmt_index): Actually call erealloc() to grow fmt_list if that's really necessary. Bug report from David Jones, djones@zoonami.com. Sun Jun 10 14:24:48 2001 Arnold D. Robbins * profile.c (pp_match_op): Rationalized the code. Thu Jun 7 11:54:36 2001 Arnold D. Robbins * awk.h (O_BINARY): Don't define if already defined (as is true for cygwin/gcc --- oops). Sun Jun 3 13:04:44 2001 Arnold D. Robbins * Release 3.1.0: Release tar file made. And there was rejoicing. Wed Apr 25 11:44:07 2001 Arnold Robbins * Makefile.am (AM_MAKEFLAGS): Add definition per advice from Nelson Beebe. Tue Apr 24 14:28:00 2001 Arnold Robbins * io.c (devopen): Patch from Jeurgen to robustify pulling out hostname, port numbers, etc, to avoid any buffer overrun problems. Mon Apr 23 10:26:38 2001 Arnold Robbins * awkgram.y: Fix grammar so that `print ... |& ".." |& getline' dies with a parse-time error message. Sun Apr 22 16:46:48 2001 Arnold Robbins * io.c (socketopen): Fix from Juergen in recursive call. Thu Apr 19 18:39:20 2001 Pat Rankin * awk.h: Really fix logic around include of . * awk.h (callresult): New name for `result' macro. * eval.c (r_get_lhs, case Node_builtin): Use it. Thu Apr 19 16:31:09 2001 Pat Rankin * io.c: Move code around to allow compilation with DEC C. Thu Apr 19 16:21:56 2001 Arnold D. Robbins * random.h: Move decl of random() here. * random.c: Remove decl of random(). Mon Apr 9 11:41:58 2001 Arnold D. Robbins * dfa.c (dfainit): Initialize more members in the structure, based on bug report in bug.gnu.utils by aaronl@vitelus.com (Aaron Lehmann). * awk.h: Fix logic around include of . Thu Apr 5 20:12:05 2001 Pat Rankin * dfa.c: for VMS, #include instead of . * missing_d/mktime.c: likewise. * random.c: reorder include directives to get gawk config info from random.h sooner. [fcntl.h]: guard #include with HAVE_FCNTL_H test. [unistd.h]: guard #include with HAVE_UNISTD_H test. * random.c (srandomdev): skip /dev/urandom usage if O_RDONLY is not defined. Tue Mar 20 11:07:11 2001 Arnold D. Robbins * awkgram.y (function_body): Add opt_nls to end of production. Tue Mar 20 09:30:32 2001 Pat Rankin * awk.h (BROKEN_STRNCASECMP): Add decl of strcasecmp. * io.c (two_way_open): Add `return FALSE;' for fussy compilers. Sun Mar 18 15:10:56 2001 Arnold D. Robbins * io.c (gawk_pclose): Set the exit value for close correctly if the pipe died with a signal. Wed Mar 7 11:28:52 2001 Arnold D. Robbins * io.c (get_a_record): Correctly handle the case of a leading single newline at the front of the file when RS = "". 2001-02-26 Paul Eggert * COPYING: Incorporate latest version from FSF, which fixes a Y2k bug. * builtin.c (do_mktime): Allow the user to specify the tm_isdst member as an optional trailing integer, and to specify "out-of-range" members. Check for overflow when subtracting 1 from month or 1900 from year. Allow years just past INT_MAX, as they work on some hosts when INT_MAX - 1900 is representable as an int. * doc/gawk.1, doc/gawk.texi: Document the above changes. Also, document that the origin-zero Gregorian calendar is used. Fix confusing wording about "midnight" by replacing it with 00:00 ("midnight" is also 24:00, the end of the day). Mention the typical range for time stamps. Do not assume that years are nonnegative and are less than 10,000. Suggest TZ=UTC0 instead of TZ=GMT0, as that's how recent versions of GNU date behave. GMT is not always the time of day in Greenwich these days. Fix typos: "Emporer/Era", "1980's", "1970's". * m4/largefile.m4: Synchronized with latest version. Tue Feb 27 12:10:11 2001 Arnold D. Robbins * profile.c (pp_in_array): Change test to tree->type == Node_expression_list. Wed Feb 7 14:46:50 2001 Arnold D. Robbins * awkgram.y (LEX_FOR): Allow newline after `;' in for loops. Per bug report from Brian Kernighan, bwk@research.bell-labs.com. Tue Feb 6 18:35:27 2001 Martin C. Brown * io.c (socket_open): Conditionalize various options based on ifdef. Needed for BeOS port. Tue Feb 6 18:17:13 2001 Michal Jaegermann * regex.c (re_match_2_internal): Case maybe_pop_jump, for charset and not_charset: change cast from (unsigned char) to (unsigned). Catches last 8 chars with high bit set if backtracking. See test/rebt8b1.awk, test/rebt8b2.awk. Tue Feb 6 11:20:21 2001 Arnold D. Robbins Have `for (iggy in foo)' save the elements and loop over them. Make sorted for loops a dynamic test instead of a compile time test. Still requires being Strong In The Ways Of The Source. * awk.h: (struct search): removed. (assoc_scan, assoc_next): removed declarations. * array.c (assoc_scan, assoc_next): removed functions. * eval.c (interpret): remove Node_K_array_sorted_for. Change code at Node_K_arrayfor. (nodetypes): remove Node_K_array_sorted_for. * configure.in: removed array sorting test. * awkgram.y: removed sorted_in keyword and associated code. Sun Feb 4 14:57:49 2001 Arnold D. Robbins * eval.c (interpret): use tree->rnode->exec_count to hold count of times if was true. profile.c (interpret): ditto. * main.c (pre_assign): gross hack. malloc fresh copy of assign so can clear the '=', otherwise screws up profiling print out. Sun Jan 28 16:16:02 2001 Arnold D. Robbins Per request from Nelson Beebe, SIGHUP to pgawk dumps profile and function call stack and exits, SIGUSR1 dumps and continues running. * eval.c (dump_fcall_stack): New function, dumps awk function call stack. * awk.h (dump_fcall_stack): Add declaration. (init_profiling_signals): Ditto. * main.c (main): Call init_profiling_signals. * profile.c (init_profiling_signals, dump_and_exit, just_dump): new functions. Sun Jan 28 15:50:02 2001 Eli Zaretskii * io.c (gawk_popen): Restore the mode of stdin before running the child process and switch it back if BINMODE is in effect after the child returns. (redirect): Restore the mode of stdin before running the child process. (close_redir): Switch mode of stdin back to binary if BINMODE is in effect, after the child returns. * builtin.c (do_system): Restore the mode of stdin before running the child process and switch it back if BINMODE is in effect after the child returns. * awk.h (os_restore_mode): Add prototype. Thu Jan 18 14:03:06 2001 Arnold D. Robbins * custom.h, README_d/README.ultrix: Fixes for Ultrix from Juergen Kahrs. Wed Jan 17 11:03:40 2001 Eli Zaretskii * io.c (redirect) [F_GETFL && O_APPEND]: Use binmode in the call to fdopen. Mon Jan 15 16:29:52 2001 Arnold D. Robbins * profile.c (prec_level): Made Node_K_getline higher than < but lower than others. Allows use of getline with redirection inside an if. Wed Jan 10 15:35:06 2001 Arnold D. Robbins * eval.c (set_BINMODE): Rationalized string assignment. Sun Jan 7 15:26:16 2001 Arnold D. Robbins * getopt.h: Removed names in prototypes for getopt_long and getopt_long_only, fixes problems on MINGW32. Thu Jan 4 10:13:46 2001 Arnold D. Robbins * configure.in: Add check for mcheck.h * main.c: Include mcheck.h if have it. (main): If TIDYMEM turned on in environment, also call mtrace(). Wed Jan 3 16:41:33 2001 Arnold D. Robbins Fixed minor memory leaks. * re.c (re_update): When IGNORECASE changed, unref(t->re_text). * eval.c (pop_fcall): Fix the logic to correctly free the vname when copying array args back to their underlying source. Fixed massive memory leaks. * node.c (dupnode): If PERM is set, do nothing. (unref): Fix logic. Always turn off TEMP. Check just for MALLOC when incrementing the stref. * array.c (assoc_lookup): Turn off PERM also when saving subscript. * builtin.c (sub_common): Turn off PERM also when making private copy of string. Add a minor memory cleanup facility (undocumented): * awk.h (do_tidy_mem, release_all_vars): Add declarations. * main.c (do_tidy_mem): Add declaration. (main): if $TIDYMEM exists, do_tidy_mem is true, and call mtrace(). * awkgram.y (release_all_vars): New function. Sun Dec 31 10:47:37 2000 Arnold D. Robbins * awkgram.y (in_end_rule): Renamed `parsing_end_rule' to avoid conflict with global var of same name. Sun Dec 24 10:36:54 2000 Eli Zaretskii * awkgram.y (snode): Reword the error message about the number of arguments for a builtin, so as not to use the English `s' as a plural suffix. Tue Dec 12 08:38:03 2000 Arnold D. Robbins * ext.c (do_ext): ifdef out use of `dummy'. Duh. * regex.c (re_error_msgid): Revert to array of `char *' so that can compile on K&R compilers. Fix all uses appropriately. (re_error_msgid_idx): Removed. Fri Dec 8 11:47:26 2000 Arnold D. Robbins * ext.c (dummy): Make gcc specific via ifdef. * builtin.c (do_dcgettext): make conditional compilation smarter. * msg.c (warning, error, r_fatal): Finish switching back to multi-version function header. Wed Dec 6 13:28:58 2000 Arnold D. Robbins * random.h: include to get ssize_t definition. * awkgram.y (yyerror): Restore multi-version function header, it seems that what ansi2knr produces doesn't quite do the job on old compilers. msg.c (msg): Ditto. Tue Dec 5 15:05:35 2000 Arnold D. Robbins * configure.in (AC_C_INLINE): Added macro call. * Makefile.am (LN): Define it for install hooks. Sun Dec 3 17:28:53 2000 Arnold D. Robbins * awk.h (os_setbinmode): Declare new function. (setmode): Remove definition: conflicts with MacOS X. * main.c (main): Change call of setmode to os_setbindmode. * builtin.c (do_dcgettext): Improve ifdef for code, fixes MacOS X. * custom.h (__APPLE__): Force definition of HAVE_MKTIME, won't link otherwise. Harumph. Sun Nov 26 11:58:52 2000 Arnold D. Robbins * builtin.c (do_dcgettext, do_bindtextdomain): Add calls to free_temp the various arguments. Sigh. * io.c (yylex): nuked bstart variable, put all uses of mend variable into TANDEM ifdef. * main.c (load_environ): removed cp variable, value never used. * random.c: Remvoed uses of `inline' keyword. * Makefile.am (install-exec-hook, uninstall-local): new targets. Adds creation of gawk-X.Y.Z and awk links, as in 3.0.x. * configure.in (GAWK_AC_TYPE_SSIZE_T): Added. m4/ssize_t.m4: new file. Wed Nov 22 14:47:18 2000 Arnold D. Robbins After consultation with Brian Kernighan and Michael Brennan, nuked the abort keyword. * awk.h (Node_K_abort): removed. * eval.c (aborting): removed decl. (interpret): Removed Node_K_abort case. * io.c (do_input): Removed checks for aborting. * main.c (aborting): removed. (main): Removed checks for aborting. * profile.c (pprint): Removed Node_K_abort case. * awk.y (LEX_ABORT): All stuff removed. Wed Nov 22 10:45:57 2000 Arnold D. Robbins * ext.c (dummy): Move inside #ifdef DYNAMIC. Helps on PCs and other platforms that don't do dynamic loading. * awk.h (RED_TCP): New flag, means use shutdown. io.c (redflags2str): Add RED_TCP. (SHUT_RD, SHUT_WR, SHUT_RDWR): Add conditional defines. (redirect): Add RED_TCP to tflag if appropriate. Add more #ifdef HAVE_SOCKETS as needed. (close_redir): If RED_TCP set, shutdown(2) on each end of the socket. Tue Nov 21 16:25:41 2000 Arnold D. Robbins * awk.y: for (iggy in foo) loops: add test that index in delete statement is a simple variable. Tue Nov 14 16:11:39 2000 Arnold D. Robbins * awk.h: Add appropriate conditional versions of the gettext functions if we don't have or if ENABLE_NLS is not defined or zero. * configure.in: Add check for libintl.h header. From Scott Deifik for PCs. * awk.h (lintwarn): Call set_loc unconditionally, makes compilation work on PCs. * builtin.c (do_dcgettext): Compile out cat_tab and code if not ENABLE_NLS. * ext.c: for MSC, no long long variable. * random.c: use clock() instead of gettimeofday(). * builtin.c: Fixed prototypes for new random functions (ADR). Sun Nov 12 17:45:44 2000 Arnold D. Robbins * builtin.c (parse_next_arg): Fix call to >= num_args so running out of args check is correct, instead of core dumping. (format_tree): Save and restore `the_args' and `args_size' if a nested call is in progress, see explanatory comment. See also tests/addcomma. * Makefile.am: Fix things so that gawk/pgawk built first, even if `make check' called before make. Add some commentary. Wed Nov 8 14:39:20 2000 Arnold D. Robbins * configure.in: Only add -rdynamic for linux. * dfa.h, dfa.c: upgraded to versions in grep 2.4.2. Tue Nov 7 18:17:17 2000 Arnold D. Robbins * All: Switched to ANSI function headers and added `ansi2knr' automake option. Really cool. Tue Nov 7 16:57:49 2000 Arnold D. Robbins * io.c (redirect): Check for O_APPEND in flags when doing fdopen() of /dev/fd/N. Thanks to bug report from "John H. DuBois III" . Tue Nov 7 14:09:14 2000 Arnold D. Robbins * awk.h (os_is_setuid): declare function. * main.c (main): call it if do_lint and warn if true. * awkgram.y (tokentab): - Made sure all extensions are actually marked as such. Ouch. - Changed "sort" to "asort". Potential to break too much old code. * getopt.h, getopt.c, getopt1.c: replaced with current versions from glibc CVS archive. Mon Nov 6 18:14:33 2000 Arnold D. Robbins * random.c: Replaced with recent version from FreeBSD. Mon Nov 6 15:37:12 2000 Arnold D. Robbins Major simplification of automake machinery. * configure.in: - INSTALL is forced only if not provided in environment - lots of Makefile.in files removed since move to automake 1.4a * Makefile.am, */Makefile.am: Moved directories that don't need the automake machinery into EXTRA_DIST as appropriate and removed the Makefile{,.am,.in} files as needed. * eval_p.c, profile_p.c: New files to make it easier with automake to compile pgawk. Tue Oct 24 12:20:18 2000 Arnold D. Robbins * awkgram.y (valinfo, var_comp, dump_vars): New functions to dump the list of global variables. * awk.h: Declare dump_vars. * main.c (optab): new option "dump-variables". (main): Code to handle it, set the output file and then call dump_vars() at the end. (usage): New option added to usage message. Sat Oct 21 22:59:59 2000 Arnold D. Robbins * awkgram.y (parms_shadow): For a function, check if any parameters shadow global variables and print a warning. (shadow_funcs): Go through all functions and call parms_shadow(). (isnoeffect, isassignable): Add Node_LINT and NODE_BINMODE. * main.c (main): If do_lint, call shadow_funcs(). * awk.h: add declaration of shadow_funcs(). * configure.in: added m4/Makefile and awklib/eg/network/Makefile to list of generated makefiles. Tue Oct 17 10:47:35 2000 Arnold D. Robbins * array.c (assoc_lookup): Reverted change that did dupnode of array indices. Creates significant problems if index is numeric value and CONVFMT changes. Added fix to set bucket->ahname->stfmt to -1 so that force_string never recalculates the string value, and also turned off NUM and turned on STR. See test/arynasty.awk. Mon Oct 16 12:21:26 2000 Arnold D. Robbins * All: Cleaned up various lint warnings for consistent phrasing. * awk.y (in_end_rule): New variable for warning about unredirected getline. It's ok in a BEGIN, but not in an END. Sun Oct 15 14:14:05 2000 Arnold D. Robbins * field.c (set_FS): Add lint warning for FS = "". (do_split): Ditto for 3rd arg = "". Fri Oct 13 09:17:04 2000 Arnold D. Robbins * io.c (close_redir): Clear rp->fp on all closes. Remove rp from list if either closing both ends or both ends have been closed separately. Add exitwarn message for co-process. (flush_io): Add warning message if fflush of co-process fails. Rationalize return value to either 0 or -1. * builtin.c (do_gensub): 3rd arg of zero generates a warning. (do_fflush): rationalize return value: -1 for unopen or read-only redirection, status of fflush otherwise. Wed Oct 11 22:11:19 2000 Arnold D. Robbins * awk.y (for loop): Check that there is a body as part of the `is it a delete statement' check. Thu Oct 5 11:56:42 2000 Arnold D. Robbins * awk.h, awkgram.y, configure.in, eval.c: enabled `for (i in_sorted array)' loops for those who are Strong In The Way Of The Source. So there. Mon Oct 2 10:09:32 2000 Arnold D. Robbins * io.c (do_close): make close(x) for non-open x return -1 and update ERRNO. close(FILENAME) no longer does anything magic; this is all for better consistency with other awks and is more logical, anyway. Thu Sep 28 17:27:16 2000 Arnold D. Robbins * io.c (close_one): Added a lint warning if it becomes necessary to start multiplexing fd's, per ancient suggestion from Scott Deifik, . Tue Sep 26 14:41:41 2000 Arnold D. Robbins * profile.c: Move enum for redirection placement to top of file, and make the value a parameter to pp_redir. Fix all the calls. This gets `|&' right everywhere. Sun Sep 24 16:38:04 2000 Arnold D. Robbins * awk.h (freenode): set the flags straight to UNINITIALIZED. * node.c (unref): Fix test for MALLOC|TEMP to test the actual flags, not zero. * builtin.c (format_tree): ala print and concat, dupnode the temp nodes from tree_evaling the arguments. See test/nasty2.awk. Mon Sep 18 10:16:58 2000 Arnold D. Robbins * awkgram.y (snode): Make match 3rd arg and close 2nd arg fatal errors if --tradtional. Thu Sep 14 12:22:42 2000 Arnold D. Robbins * eval.c (update_ERRNO): Call gettext on result of strerror. i18n rules. Wed Sep 13 14:56:11 2000 Arnold D. Robbins * eval.c (r_tree_eval): Case for Node_concat. Dupnode the strings ala do_print to get more consistent results. Compare gawk 3.0.6 to nawk/mawk on test/nasty.awk. Thanks to Andrew Sumner (andrewsumner@yahoo.com) for pointing this one out. Wed Sep 13 10:06:47 2000 Arnold D. Robbins * io.c (two_way_close_type): New enumerated type. (close_redir): New third param of type two_way_close_type. Add smarts to two-way case for different close types. Only remove it from the redir list if closing is for both ends. (gawk_pclose): Check that rp->iop != NULL before closing, all three versions. * awkgram.y (tokentab): Allow 2nd argument to close. (snode): Add lint warning. Sun Sep 10 14:16:10 2000 Arnold D. Robbins * field.c (set_FIELDWIDTHS): Generate a fatal error upon encountering a negative width. Sun Sep 10 10:37:35 2000 Arnold D. Robbins * awkgram.y (snode): If first argument to dcgettext is a string constant and --gen-po, dump the string constant to the .po file too. * main.c (nostalgia): Add call to fflush(stderr). * eval.c (r_tree_eval): Add entries for Node_LINT and for NODE_TEXTDOMAIN. Thu Sep 7 10:46:20 2000 Arnold D. Robbins * builtin.c (do_dcgettext): Per suggestion from Ulrich Drepper, make the awk interface: str = dcgettext(string [, domain [, category]]) Wed Sep 6 16:28:12 2000 Arnold D. Robbins Bring gettext features out to the awk level! * awk.h: Add declarations of new functions `do_dcgettext' `do_bindtextdomain', `set_TEXTDOMAIN' and variables `TEXTDOMAIN', `TEXTDOMAIN_node'. New NODETYPE enum `Node_TEXTDOMAIN'. * eval.c (nodetypes): add Node_TEXTDOMAIN at end. (set_TEXTDOMAIN): new function. (r_get_lhs): add case for Node_TEXTDOMAIN. * main.c (varinit): add entry for TEXTDOMAIN. * node.c (format_val): If INTLSTR use dcgettext of string and TEXTDOMAIN. * awkgram.y (tokentab): Add entries for "dcgettext" and "bindtextdomain". * builtin.c (do_dcgettext, do_bindtextdomain): new functions. Tue Sep 5 17:01:34 2000 Arnold D. Robbins * profile.c (pp_string_fp): Use lower case versions of isascii and isprint to avoid printing high-bit-set characters. Make it smarter to break strings at 70 chars or after embedded newline, for --gen-po. Fix the calls to it everywhere for new boolean option to yes/no break lines. * m4/strtod.m4: new file, defines GAWK_AC_FUNC_STRTOD_C89. * configure.in: GAWK_AC_FUNC_STRTOD_C89 call added * acinclude.m4: include strtod.m4. * acconfig.h: add entry for STRTOD_NOT_C89. Remove entries for BITOPS and NON_DEC_DATA. * missing/missing.c: add check for STRTOD_NOT_C89, use ours if set. * missing/strtod.c: make smarter for input like 0x345. * awk.h: [STRTOD_NOT_C89]: define strtod gawk_strtod to get our version. Avoids linker weirdness. Mon Sep 4 09:16:43 2000 Arnold D. Robbins * field.c (set_record): fix from Utz-Uwe Haus to make sure there's always enough room in the record. * builtin.c (nondec2awknum): Fix octal conversions to exit when hitting a non-digit, and not go to decimal. Make check for non-octal better. Based on bug report from Morris_Lee@tvratings.com. Sun Sep 3 13:52:11 2000 Arnold D. Robbins * builtin.c (format_tree): Allow positional parameters for %*.* kinds of things. Made octal/hex constants and strtonum on by default. Made --enable-non-decimal-data a runtime switch `--non-decimal-data'. * configure.in: Removed AC_ARG_ENABLE for --enable-bitops and --enable-non-decimal-data. In .developing check, remove the AC_DEFINEs. * awk.h: Decls for bitwise functions now there by default. Add decl of `do_non_decimal_data'. * main.c (do_non_decimal_data): new variable (optlist): add new entry for `--non-decimal-data'. (main): turn off `do_non_decimal_data' if `do_traditional'. (usage): add the new option. * node.c (r_force_number): make check for non-decimal data a runtime check based on do_non_decimal_data. * awkgram.y (yylex): make non-decimal constants a runtime check. * builtin.c: remove the ifdefs around the bit functions and nondec2awknum. Tue Aug 29 18:45:56 2000 Arnold D. Robbins * configure.in: go back to ARRAYDEBUG if .developing set. * awkgram.y: use ARRAYDEBUG for adump(), use multiple tests for stopme(). Mon Aug 28 17:09:06 2000 Arnold D. Robbins * field.c (do_split): Add check for first arg is null string, if so, skip the work and return zero. Mon Aug 14 23:01:55 2000 Arnold D. Robbins * Add %COUNT$... handling to printf. awk.h (printf_count): new define in NODE structure. (format_tree): added decl. awkgram.y (count_args): new function to set printf_count in a node. [print productions]: call the function. (snode): for do_sprintf, call count_args, set the count in the lnode. builtin.c (format_tree): new fourth arg is argument count. Add smarts to handle the `$' in a format. (do_sprintf): use new argument to format_tree. node.c (format_val): ditto. Sun Aug 13 11:10:41 2000 Arnold D. Robbins Changes from Alan J. Broder (ajb@woti.com): - Array third arg to match puts subtexts into the array: * awk.y (tokentab): "match" gets third arg, and lint warning * builtin.c (do_match): if third arg there, fill it with subtexts - New builtin sort function: * awk.h (do_sort): declared. * array.c (do_sort, dup_table, merge, merge_sort, assoc_from_list, assoc_sort_inplace): new functions. * eval.c (tree_eval): in debug code, make uninitialized var a warning, not a fatal error. Breaks too many things. Wed Aug 9 10:51:41 2000 Arnold D. Robbins * eval.c (func_call): Increment the exec_count on the function's node; this allows printing a call count for functions. profile.c (pp_func): print the count for functions. * ALL: Changed DEBUG to GAWKDEBUG in all gawk files, so that I don't get regex/dfa debugging. In some cases, changed memory-related stuff to MEMDEBUG. Still have work to do. * awk.h, node.c, profile.c: removed exec_count_init variable; code has been cleaned up to not need different values for profiling/not profiling. Thu Jul 5 21:10:59 2000 Arnold D. Robbins * eval.c (casetable): Removed the USE_PURE_ASCII stuff; it was never documented. Latin 1 forever. * main.c (main): only call `init_profiling' after arg parsing if `do_profiling' is still false. Avoids resetting `prof_fp' back to stderr. 2000-02-17 Akim Demaille * m4: New directory. * acinclude.m4: Removed, replaced by m4/*.m4. * Makefile.am: Adjusted. Added ACLOCAL_AMFLAGS. * configure.in Adjusted. Use AC_SYS_LARGEFILE not GAWK_AC_SYS_LARGEFILE, jm_FUNC_MKTIME, not GAWK_FUNC_MKTIME. * acconfig.h: Removed _FILE_OFFSET_BITS, _LARGEFILE_SOURCE and _LARGE_FILES now templated by m4/largefile.m4. 2000-02-15 Arnold Robbins * MOVED TO AUTOMAKE AND GETTEXT. Just about every file touched. Work done by Arno Peters. Sun Jan 2 14:48:23 2000 Arnold D. Robbins First edit of the new millenium! * awk.y (yylex): if lint checking, be obnoxious about gotos. Mon Oct 25 19:12:02 1999 Arnold D. Robbins * awk.h: remove C_ALLOCA ifdef. * main.c (main): remove C_ALLOCA code. * io.c (do_input): ditto. Mon Aug 9 17:36:24 1999 Arnold D. Robbins * bisonfix.sed: unconditionally #undef YYSTACK_USE_ALLOCA. * configure.in: remove all alloca and ALLOCA related stuff. * Makefile.in: ditto Thu Jul 29 18:32:05 1999 Arnold D. Robbins * awk.h (NODE): exec_count now in #ifndef NO_PROFILING. * Makefile.in: changes to only recompile eval.c and profile.c to a special version for profiling. * custom.h [MSC_VER]: turn on NO_PROFILING to omit the exec_count and save space. * node.c (more_nodes): move setting of exec_count to #ifndef NO_PROFILING. Thu Jul 1 12:12:05 1999 Arnold D. Robbins * configure.in (AC_PREREQ): update to 2.13. GAWK_AC_C_STRINGIZE: convert to AC_C_STRINGIZE. * aclocal.m4 (GAWK_AC_C_STRINGIZE): remove definition, now part of autoconf. * acconfig.h (HAVE_STRINGIZE): ditto. Wed Apr 28 11:08:05 1999 Arnold D. Robbins * array.c (assoc_lookup): fix call to free_temp(subs) to after last use of subs. Sun Apr 25 16:48:06 1999 Arnold D. Robbins * io.c (redirect): add lint warning when same file is used for > and >>. Thu Apr 22 15:05:30 1999 Arnold D. Robbins * array.c (assoc_lookup): Fix call to fatal to lintwarn instead. * node.c (r_force_number): Use `0 &&' to disable warnings about conversions: they're overzealous, methinks. Thu Apr 8 14:27:58 1999 Arnold D. Robbins New features for profiling: * awk.h (NODE): add `exec_count' member. (freenode): clear `exec_count' upon free. * awk.y (func_count): new variable, counts total number of functions. (func_install): increment func_count. (struct finfo): information for use in sorting functions when pretty printing. (fcompare): compare two finfo structures. (dump_funcs): print the functions in sorted order for profiling. (getfname): return the name of a builtin function. * eval.c (INCREMENT): new macro for counting execution of nodes. (interpret): call INCREMENT() appropriately. * main.c (do_profiling): new flag if doing profiling. `--profiling': new option added to getopt_long machinery. (main): For profiled version, set do_profile and output file. Call `dump_prog' and `dump_funcs' if do_profiling at end. (usage): add new argument. * node.c (more_nodes, freenode): set exec_count to zero. * profile.c: new file, does pretty printing and prints counts. * Makefile.in: update to create two versions of gawk, regular and `pgawk' which does profiling. Wed Mar 10 21:38:14 1999 Arnold D. Robbins * io.c (close_redir): use update_ERRNO() instead of manually doing it. Mon Dec 21 15:58:21 1998 Arnold D. Robbins * configure.in: add BeOS to list of cases where we hardwire GETPGRP_VOID. custom.h: remove the #define from __be_os case. Cleaner to do it all in configure. Based on email from Martin C. Brown, mc@whoever.com. Mon Nov 30 20:52:52 1998 Arnold D. Robbins * eval.c (update_ERRNO): new function, mainly for use by extension functions. * awk.h: add decl. Tue Nov 24 18:13:29 1998 Arnold D. Robbins * Changes based on submission from Christos Zoulas at D.E. Shaw that adds the following features: - checking for use of uninitialized variables - checking if a string that's not a number converts to 0 - ability to load a dynamic library to add built-ins - VERSION variable (may or may not stay) Additional change: - --lint=fatal makes lint errors become fatal - LINT="fatal" has the same effect, any other positive value makes lint errors be just warnings * Makefile.in (includedir): new variable for gawk header files (ext.c, ext.o): new source and object files (OTHERS, extension): new directory for macro with example extension (install): install header files * acconfig.h (DYNAMIC): new macro, true if can do dynamic loading * array.c (assoc_lookup): new parameter `reference' is true if we want to do reference checking. Add appropriate reference checking code. * awk.h (UNITITIALIZED): new flag (lintfunc): function pointer for correct function to use (lintwarn): new macro to produce warnings (result): new macro for func call result, used in commented out code in eval.c. (getnode, freenode): revised to set UNINITIALIZED. (get_lhs): third arg for reference checking, change all calls -- Add appropriate decls of new/changed functions * awk.y (tokentab): new builtin "extension" for adding extensions (node_common): set flags to UNINITIALIZED for Node_var. * configure.in (dynamic linking): new check. Probably should be a separate macro. * eval.c (flag2str): add UNINITIALIZED to the table. (r_tree_eval): add checks for UNINITIALIZED. (push_args): appropriate changes for UNINITIALIZED to work. (r_get_lhs): new third argument for reference checking. (set_LINT): add code to handle setting `lintfunc' appropriately. * ext.c: new file, for doing dynamic library extensions. * extension/*: new directory with simple example code. * main.c (VERSION_node, EXTENSION_node): new nodes for new vars. (optab): change for "lint" to allow optional argument. (lintfunc): definition. (main): add case in option processing for --lint. (varinit): add entries for VERSION and EXTENSION. * node.c (r_force_number): checks that string really is a number. (morenodes): set UNITIALIZED in the flags. * re.c (all): change `result' to `res' globally to avoid conflict with new macro. * GLOBAL: change lint calls to warning() to lintwarn(). * GLOBAL: change all calls to get_lhs() to have 3rd arg. * GLOBAL: change all calls to assoc_lookup() to have 3rd arg. Sun Nov 22 17:07:39 1998 Arnold D. Robbins * patchlev.h: renamed from patchlevel.h to make life easier for the PC guys. (main.c): changed to include patchlev.h. (Makefile.in): changed to ref patchlev.h where needed. Sat Nov 7 21:29:52 1998 Arnold D. Robbins * eval.c (r_get_lhs): case Node_field_spec. Fix the lint warnings for field reference of null string or non-numeric value. When turned on, $0 generated a warning! Oops. Thu Nov 5 16:58:38 1998 Arnold D. Robbins * main.c (init_fds): new function to pre-open 0, 1, and 2 on /dev/null if they're not open. Robustness, more or less. (main): call init_fds. * io.c (str2mode): add smarts for two-letter strings such as "rw", "r+", "wr", "w+" and "a+". Mon Nov 2 16:55:46 1998 Arnold D. Robbins * builtin.c (do_*): added lint checks for non-numeric and/or non-string arguments, as appropriate. This should have been done long ago. Tue Oct 20 21:56:06 1998 Arnold D. Robbins * awk.h (LINT_node): new variable for LINT special var (Node_LINT): new node type. (set_LINT): declare function. * main.c (varinit): add LINT variable. (usage): print an emphatic pointer to the manual for bug reports. * eval.c (nodetypes): new entry for Node_LINT. (r_get_lhs): case added for Node_LINT. (set_LINT): set do_lint from LINT variable. Mon Oct 19 22:35:46 1998 Arnold D. Robbins * configure.in: for GCC, add -Wall to get warnings for development. * Makefile.in (awktab.c): move sed stuff to separate script. * bisonfix.sed: new script, with old fix and Solaris x86 fix. * awk.h (nodetype2str): add declaration. (load_procinfo): add declaration. Tue Oct 13 22:28:56 1998 Arnold D. Robbins * Changes to make PROCINFO["FS"] reflect the use of FIELDWIDTHS or FS. eval.c (assign_val): new function that does the mechanics of assignment main.c (load_procinfo): add setting of PROCINFO["FS"] to "FS". field.c (update_PROCINFO): new function to update the array. (set_FS): call update_PROCINFO. (set_FIELDWIDTHS): ditto. Sun Sep 27 10:18:05 1998 Arnold D. Robbins * awk.h (reisstring): new prototype. * re.c (reisstring): new function, returns true if the re did a simple string match. This is rather simplistic in its logic. * io.c (get_a_record): in the case that RS is a regexp, AND the re matched at the exact end of the buffer, add a call to `reisstring' in case it's a simple string match. If so, we don't need to read more into the buffer because we don't have a regex like `x.*y' that might extend longer. This should be very helpful for interactive /inet clients where something like `RS = "\r\n"' happens. Thu Aug 13 22:07:40 1998 Arnold D. Robbins * io.c (socketopen): fixes from Juergen Kahrs to socket opening code for "any host". Tue Jul 14 19:02:33 1998 Arnold D. Robbins * aclocal.m4 (GAWK_AC_LIB_SOCKETS): removed the caching; configure gave different results the second time it was run! Fri Jul 10 09:11:06 1998 Arnold D. Robbins * eval.c (interpret): minor cleanups: add variable name to fatal error Node_K_array_for and other minor changes. Mon Jun 22 16:53:34 1998 Arnold D. Robbins * Makefile.in (tags, TAGS): add $(LIBSRC). Tue Jun 2 15:23:05 1998 Arnold D. Robbins * io.c (devopen): relax previous change, don't require "any", just that a port be there. The user can put 0 if they don't care. Wed May 27 21:33:45 1998 Arnold D. Robbins * io.c (devopen): for /inet, require that local and remote ports and the remote hostname be there, and that `any' be used for a port if they don't care. Thu May 21 14:13:46 1998 Arnold D. Robbins * node.c (parse_escape): Add warning that is always on for \q for any unknown q inside string or regex constant. I got bit by this myself once too often. Or else I'm just getting old and senile. Mon May 4 12:42:49 1998 Arnold D. Robbins * awk.h (NODETYPE): Sorted the Node_xxx entries for the builtin variables. Gotta look nice, don't we? * eval.c (nodetypes): ditto. (genflags2str): added code to check that we don't overflow the static buffer. This is just a debugging routine, not worth the hassle of dynamic allocation. Mon Mar 2 16:06:16 1998 Arnold D. Robbins * Makefile.in (dist): remove any embedded copied RCS or CVS directories. Mon Feb 23 00:09:52 1998 Arnold D. Robbins * awk.h (genflags2str): add declaration. * eval.c (genflags2str): new function. (flags2str): use new general purpose function. * io.c (redflags2str): same. Sun Feb 22 23:57:29 1998 Arnold D. Robbins Significant changes to add two-way i/o and sockets!!! * Makefile.in: add @SOCKET_LIBS@ to LIBS variable. * acconfig.h: add HAVE_SOCKETS and HAVE_PORTALS defs. * aclocal.m4: new macro GAWK_AC_LIB_SOCKETS. * awk.h: new node type, Node_redirect_twoway, and new redirection flags: RED_TWOWAY, and RED_SOCKET. * awk.y (parser): add TWOWAYIO token and appropriate productions. (yylex): recognize `|&' token if not traditional. * builtin.c (do_print, do_printf): flush buffer if TWOWAYIO. * configure.in: add header checks for networking header files, add --enable-portals switch, call GAWK_AC_LIB_SOCKETS * eval.c (nodetypes): add string constant for Node_redirect_twoway. * io.c (redflags2str): new function. (redirect): better error message in default case, add code for Node_redirect_twoway. (socketopen): new function. (iop_open, devopen): add recognition of `/inet/...'. (two_way_open): new function. Sat Dec 13 21:15:07 1997 Arnold D. Robbins * awk.h (struct node): new member, `param_list' in union `x', becomes `node->parmlist' in the code. * awk.y (func_install): rearranged a bit, to build up a list of the function parameter names and to save it in the `parmlist' field. * eval.c (push_args): new parameter, `varnames', which is the list of variable names. Use this to set the vname field of each parameter's value as it's created. Special case arrays to include where they came from, mainly for array vs. scalar diagnostics. (r_tree_eval): don't set the `vname' field for parameters. (pop_fcall): free the `vname' field if it's an array. (func_call): pass in the `parmlist' field to call of push_args(). (r_get_lhs): for Node_subscript, change error message to use the `vname' field. (stopme): new do-nothing function for use with debugging code and setting breakpoints. Thu Dec 4 15:18:17 1997 Arnold D. Robbins * awk.y: fixed several lint checks and moved some into test for do_lint_old. * eval.c (fmt_index): add value of bad format spec to error message. Tue Nov 18 22:19:02 1997 Arnold D. Robbins * Makefile.in (install): strip the installed binary. From Anatoly A. Orehovsky (tolik@mpeks.tomsk.su). Sun Nov 16 22:12:39 1997 Arnold D. Robbins * array.c (in_array, assoc_lookup): add symbol->vname to fatal calls for scalar in array context. Wed Nov 12 22:18:33 1997 Arnold D. Robbins * awk.h [ISASCII]: on all IS* macros, add cast to unsigned char. [TOUPPER, TOLOWER]: new macros using unsigned char. * awk.y: change to use of IS* vs. is* macros. * builtin.c (nondec2awknum): change to use of IS* vs. is* macros, change casts for casetable[] from int to unsigned char. use new TOLOWER, TOUPPER macros * dfa.c [ISASCII]: on all IS* macros, add cast to unsigned char. (lex): change isdigit to ISDIGIT. [TOUPPER, TOLOWER]: new macros using unsigned char, now used. * eval.c (fmt_ok): change to use of IS* vs. is* macros. * field.c (sc_parse_field): change to use of IS* vs. is* macros, change casts for casetable[] from int to unsigned char. (set_FS): change to use of IS* vs. is* macros. * io.c (get_a_record): change to use of IS* vs. is* macros, change casts for casetable[] from int to unsigned char. * main.c (main): change to use of IS* vs. is* macros. * node.c (r_force_number, parse_escape): change to use of IS* vs. is* macros. * re.c (make_regexp): change to use of IS* vs. is* macros. * regex.c [ISASCII]: on all IS* macros, add cast to unsigned char. Sun Oct 19 12:36:47 1997 Arnold D. Robbins * ALL: Change email address to arnold@gnu.org in all relevant places. Wed Oct 15 03:38:12 1997 Arnold D. Robbins * awk.y (yylex): Don't allow newlines after ? or : if do_posix. Thu Oct 9 19:28:39 1997 Arnold D. Robbins * custom.h [SEQUENT]: removed; not needed any more since the mmap code was ripped out. Wed Oct 8 17:22:03 1997 Arnold D. Robbins * configure.in: remove check for madvise; don't need it any more after nuking use of mmap. Tue Oct 7 11:14:21 1997 Arnold D. Robbins * eval.c (flags2str): made the code table driven. Shortened a lot. Tue Sep 30 20:59:17 1997 Arnold D. Robbins * eval.c (r_get_lhs): case Node_field_spec. Add lint warnings for field reference of null string or non-numeric value. Based on patch submitted by Alan Broder, ajb@dtmr.com. Wed Sep 24 20:47:59 1997 Arnold D. Robbins * custom.h [TANDEM]: new changes. Finishes up Tandem integration. Mon Sep 22 00:42:34 1997 Arnold D. Robbins * custom.h [__be_os]: remove BROKEN_TOKEN definition. dfa.c, dfa.h: change `token' to `dfa_token' to avoid BeOS compile problems. Thu Aug 7 22:35:17 1997 Arnold D. Robbins * Changes for BeOS from mc@whoever.com awk.h (strncasecmp): bracket prototype. custom.h [__be_os]: new stuff. dfa.h, dfa.c [BROKEN_TOK]: new ifdefs to use dfa_token, not token. Fri Aug 1 13:32:49 1997 Arnold D. Robbins * Tandem changes: awk.h [TANDEM]: misc additions, as needed. io.c (get_a_record): changes for fixed length records; not used on other systems. main.c (MRL): new variable, TANDEM specific. (main): update handling -mr option for TANDEM. (load_environ): comment out whole routine if TANDEM. missing.c [TANDEM]: new includes. gawkmisc.c [TANDEM]: include `tmiscc'. Wed Jul 30 19:53:52 1997 Arnold D. Robbins * Close-on-exec changes: awk.h: (os_close_on_exec, os_isdir): new functions. gawkmisc.c: add include fcntl.h. configure.in [AC_CHECK_HEADERS]: add fcntl.h. io.c (devopen, iop_open): change to use os_isdir(), not S_IFDIR(). (redirect, devopen, iop_open, gawk_popen): change all calls to fcntl() to os_close_on_exec(). Tue Jul 29 11:09:45 1997 Arnold D. Robbins * eval.c (set_BINMODE): fixed check for digits to use isdigit() instead of looping over digits and using strchr(). Duh. Sat Jul 26 22:52:08 1997 Arnold D. Robbins * eval.c (set_BINMODE): fix so that `-v BINMODE=w' works. * node.c (r_force_number): add decl of strtod(); makes things work on MIPS. * Makefile.in (install-strip): new target. Fri Jul 18 13:28:05 1997 Arnold D. Robbins * io.c (redirect, devopen, iop_open, gawk_popen): set the close-on-exec flag on all files and pipes opened for I/O. Keeps children run via system() or other pipes from running out of file descriptors. (Reported by Kenny McCormack, gazelle@yin.interaccess.com.) Tue Jul 8 22:18:00 1997 Arnold D. Robbins * awk.y [LEX_NEXT]: Removed support for `next file' as two words. Tue Jul 8 06:46:32 1997 Arnold D. Robbins * dfa.c: changes from pjr@jet.UK (Paul J Rippin) from an old bug report against 2.14.0 that speed up initialization and rewrite the inner loop into readable code. Thu Jul 3 11:44:50 1997 Arnold D. Robbins * Atari support moved into new `unsupported' directory. awk.h, Makefile.in, gawkmisc.c, and missing.c modified. Sun Jun 29 14:17:37 1997 Arnold D. Robbins * awk.y (exp): fixed warning about `x = /foo/'. Wed Jun 25 09:07:57 1997 Arnold D. Robbins * PORTS: removed from distribution. * Makefile.in (MISC): removed PORTS. Sun Jun 22 11:52:57 1997 Arnold D. Robbins * BINMODE changes awk.h (Node_BINMODE): added. (struct redirect): added mode field to save for io.c:close_one(). (BINMODE, BINMODE_node, set_BINMODE): add declarations. awk.y (isnoeffect): add Node_BINMODE. eval.c (nodetypes): add Node_BINMODE string. (r_tree_eval, r_get_lhs): add cases for Node_BINMODE. (set_BINMODE): new function. io.c (binmode): new function. (nextfile, redirect, gawk_popen): add calls to binmode(). main.c (BINMODE, BINMODE_node): add decls. (main): add call to setmode() if BINMODE is set. (varinit): add entry for BINMODE. Wed Jun 4 21:52:25 1997 Arnold D. Robbins * configure.in [AC_FUNC_MMAP]: removed call. * awk.h [struct iobuf]: removed IOP_MMAPED flag and `getrec' member. * io.c: removed all mmap related code. Sun Apr 27 16:23:56 1997 Arnold D. Robbins * aclocal.m4 [GAWK_AC_FUNC_MKTIME]: new macro. * configure.in (GAWK_AC_FUNC_MKTIME): call it. Thu Apr 24 23:25:06 1997 Arnold D. Robbins * io.c (devopen): remove stat test for /dev/foo files. Finally. Fri Jul 26 09:23:15 1996 Arnold D. Robbins * Changes to add an abort statement, a la tawk awk.h (Node_K_abort): new enum value for NODETYPE. main.c (aborting): new flag variable. (main): add logic to handle aborting. eval.c (interpret): add case for Node_K_abort. io.c (do_input): if aborting, break loop. awk.y (tokentab): add entry for "abort" keyword (PRODUCTIONS): add production for LEX_ABORT. Wed Jul 24 12:49:52 1996 Arnold D. Robbins * First cut at changes for i18n. awk.h (do_intl): declare new flag variable. [INTLSTR]: new flag def. (m_tree_eval): fix definitions for INTLSTR. (force_string): fix definitions for INTLSTR. awk.y (yylex): add _"..." for international strings. (dumpintlstr): new function. main.c (do_intl): define new flag variable. (optab): add "gen-po" entry. (main): if do_intl, exit, don't run the program. (gawkoption): add "gen-po" entry. node.c (r_force_string): call gettext if flags indicate INTLSTR. Thu Mar 14 06:29:42 1996 Arnold D. Robbins * awk.h (do_mktime): added declaration of new function. * builtin.c (do_mktime): new function. * awk.y (tokentab): added "mktime" to list of gawk extensions. * missing.c [HAVE_MKTIME]: added include of mktime.c if needed. Mon Feb 26 22:32:19 1996 Arnold D. Robbins * io.c (pidopen, useropen): added warnings to use PROCINFO[], not special files. * main.c (load_procinfo): new function. * awk.y (variable): added call to load_procinfo() function. Mon Aug 7 15:23:00 2000 Arnold D. Robbins * Release 3.0.6: Release tar file made. Thu Aug 3 17:47:53 2000 Greg McGary * regex.c: patches for gcc bounded pointer handling. Thu Aug 3 13:09:09 2000 Arnold D. Robbins * array.c (in_array, do_delete): fix tests for index equality when searching through the array to work correctly when index is "". Fri Jul 14 21:40:17 2000 Pat Rankin * builtin.c (format_tree): Workaround a DEC C V5.7 bug by splitting `strcpy() + 3' into two expressions (the builtin inline strcpy evidently has erroneous return type of void * instead of char *; reputedly fixed in V6.1). * eval.c (C): New macro. [casetable]: Use it to add explicit casts for the character values outside the range of 0 to 127. * missing/strncasecmp.c [C, charmap]: Likewise. * io.c (redirect): Add EIO check on failed open for VMS. Fri Jul 14 11:57:23 2000 Arnold D. Robbins Efficiency hack: turn `for (iggy in foo) delete foo[iggy]' into moral equivalent of `delete foo'. * array.c (do_delete_loop): new routine. * awk.h [NODETYPE]: new Node_K_delete_loop value. Add declaration of do_delete_loop. * awk.y [LEX_FOR]: Fix code to recognize special case. * eval.c (nodetypes): new entry for Node_K_delete_loop. (interpret): add case for Node_K_delete_loop, add more diagnostic info in default (cant_happen) case. Tue Jul 11 22:15:10 2000 Pat Rankin * awk.y (nextc): Recast unsigned char values back to int to prevent VAX C from truncating EOF to 255. Tue Jul 11 14:08:23 2000 Arnold D. Robbins * array.c (do_delete): switch to string comparison, not cmp_nodes. (assoc_find): add call to force_string on subscript. * eval.c (interpret): Case Node_K_arrayfor: check for Node_array_ref and fetch original_array. Yowser. Fri Jun 30 21:57:00 2000 Arnold D. Robbins * array.c (assoc_lookup): Don't force the subscript to be a string. Not a good idea after the change to using dupnode. Sun Jun 25 15:08:19 2000 Arnold D. Robbins * Release 3.0.5: Release tar file made. Wed Jun 14 13:03:45 2000 Arnold D. Robbins * field.c (set_record): manage a private buffer for $0. Keeps things safe in case `getline var' rearranges the IOBUF's contents that $0 is still pointing into. Tue Jun 13 16:27:55 2000 Paul Eggert Upgrade to latest and greatest version of largefile code. * configure.in (AC_CANONICAL_HOST): Remove. (GAWK_AC_SYS_LARGEFILE): Defer until after AC_MINIX, to avoid autoconf warnings. Rewrite largefile configuration so that we don't need to run getconf and don't need AC_CANONICAL_HOST. * config.guess, config.sub: Remove these files. * Makefile.in (MISC): Remove config.guess, config.sub. * m4/largefile.m4 (GAWK_AC_SYS_LARGEFILE_FLAGS, GAWK_AC_SYS_LARGEFILE_SPACE_APPEND): Remove. (GAWK_AC_SYS_LARGEFILE_TEST_INCLUDES): New macro. (GAWK_AC_SYS_LARGEFILE_MACRO_VALUE): Change arguments from CODE-TO-SET-DEFAULT to VALUE, INCLUDES, FUNCTION-BODY. All uses changed. Instead of inspecting the output of getconf, try to compile the test program without and with the macro definition. (GAWK_AC_SYS_LARGEFILE): Do not require AC_CANONICAL_HOST or check for getconf. Instead, check for the needed flags by compiling test programs. (GAWK_AC_SYS_LARGEFILE): Define _XOPEN_SOURCE to be 500 to work around glibc 2.1.3 bug. (GAWK_AC_SYS_LARGEFILE_FLAGS): Don't use -n32 on IRIX if the installer said otherwise. (GAWK_AC_SYS_LARGEFILE_FLAGS): Work around a bug in the QNX shell, which doesn't propagate exit status of failed commands inside shell assignments. Wed Jun 7 13:23:09 2000 Arnold D. Robbins * Updated copyright dates in appropriate files. Mon May 22 17:29:43 2000 Arnold D. Robbins * Makefile.in (clean): get `*/core' too. Sun May 7 16:33:05 2000 Arnold D. Robbins * array.c (concat_exp): Change ref to `lnode->stlen' and `lnode->stptr' for SUBSEP to use `var_value->...'. Tue May 2 09:54:29 2000 Arnold D. Robbins Fix referencing freed memory as shown by test/arynocls.* tests. * awk.h [Node_array_ref]: new node type. [orig_array]: new macro element in NODE structure. * field.c (do_split): handle case for Node_array_ref, fetch the original array. * array.c (in_array, do_delete): ditto. * eval.c (nodetypes[]): add Node_array_ref string. (r_tree_eval): handle case for Node_array_ref. (push_args): push arrays as Node_array_ref, and pass them on. (pop_fcall): don't unref lnode if it's an array when releasing local arguments. Check for both Node_array and Node_array_ref. (r_get_lhs): choke on Node_array_ref as for Node_array. For Node_subscript, handle Node_array_ref. Tue May 2 09:52:12 2000 Bruno Haible * io.c (redirect): After reopening a `struct redirect', move it to the head of the list. Sun Apr 2 17:51:40 2000 Arnold D. Robbins * re.c (re_update): Check if IGNORECASE has changed, and if so recompute the re. See test/igncdym.awk. Mon Mar 20 16:18:34 2000 Arnold D. Robbins * io.c (set_RS): Added a lint warning about multicharacter RS, per suggestion from Akim DeMaille (akim@epita.fr). Sun Feb 13 14:40:32 2000 Arnold D. Robbins * eval.c (push_args): Fix from Nide Naoyuki , re-assign `f' in case tree_eval moved fcall_list around. Sun Feb 6 11:39:33 2000 Arnold D. Robbins * eval.c (op_assign): Fix it right. For ++ and --, get the lhs in the operations, do the op, and then return. For += etc, get the rhs FIRST, since the lhs can move around as a result, *then* get the lhs and do the operation. See test/opasnidx.awk. Tue Feb 1 18:41:40 2000 Arnold D. Robbins * eval.c (op_assign): reget the rval after regetting the left hand side. See test/opasnslf.awk for why. Thu Jan 27 18:06:31 2000 Arnold D. Robbins * awk.y (yylex): Made ']' not one of the characters that sets `want_assign' to false. `a[i] /= 2' was broken. Per bug report from Kristofer T. Karas . Wed Dec 22 15:06:37 1999 Arnold D. Robbins * awk.y: Removed declarations of functions before definition of `tokentab[]'. They're redundant with what's in awk.h. Thu Dec 9 17:01:07 1999 Arnold D. Robbins * node.c (parse_escape): Add lint warning for unrecognized escape sequences. Mon Dec 6 15:17:34 1999 Arnold D. Robbins * main.c (usage): Changed bug reporting email addresses to be a reference to `Bugs' node in the online and printed doc, instead. Thu Dec 2 13:08:18 1999 Arnold D. Robbins * builtin.c (do_compl): test `d' for negative inside the do_lint test, not uval. Ooops. Fri Nov 26 10:58:36 1999 Arnold D. Robbins * array.c (assoc_find): ALWAYS compare indexes as strings, don't use cmp_nodes in case they are numeric. Oh my. Talk about a Day 1 bug! Tue Nov 23 11:58:53 1999 Arnold D. Robbins * regex.c (SYNTAX): cast argument to `unsigned char' instead of &-ing with 0xFF. Hopefully somewhat more portable, ala 21 Nov 99 changes to awk.y. Sun Nov 21 22:25:27 1999 Paul Eggert * aclocal.m4 (AC_SYS_LARGEFILE_FLAGS): Work around a problem with the QNX 4.25 shell, which doesn't propagate exit status of failed commands inside shell assignments. Sun Nov 21 20:33:35 1999 Arnold D. Robbins * awk.h (nextc): remove declaration, don't need it here. awk.y (nextc): Cast values to unsigned char so that latin-1 characters in strings don't turn themselves into EOF. Most notably y-umlaut, which is decimal 255. Mon Nov 1 20:00:25 1999 Arnold D. Robbins * regex.c (init_syntax_once): move below definition of ISALNUM etc., then use ISALNUM to init the table, so that the word ops will work if i18n'ed. (SYNTAX): And subscript with 0xFF for Latin-1 characters. Mon Oct 25 18:37:13 1999 Arnold D. Robbins * awk.h, main.c, io.c: undo previous changes (22 Oct 1999). * main.c (main): move call to `init_fields()' to before arg parsing. This allows `-v NF=blah' to work ok. Fri Oct 22 17:43:40 1999 Arnold D. Robbins * main.c (arg_assign): Add new arg, `initing' for icky special casing of -v of special variables. Use it to check for NF. May need to add other cases later. (pre_assign): change call arg_assign, passing initing=TRUE; io.c (nextfile): change call arg_assign, passing initing=FALSE; awk.h: Change prototype for arg_assign. Tue Oct 19 16:06:48 1999 Paul Eggert * io.c (close_redir): Don't munge errno between setting it and using it. Wed Oct 6 17:47:47 1999 Arnold D. Robbins * main.c (arg_assign): return NULL on bad variable. Allows things like `./3x=stuff' to work as a filename. Thu Sep 23 21:35:46 1999 Paul Eggert * aclocal.m4 (GAWK_AC_SYS_LARGEFILE_FLAGS): Work around GCC 2.95.1 bug in HP-UX 10.20 or later. (Had to fix the fix. ADR. :-) Tue Sep 21 13:31:36 1999 Arnold D. Robbins * builtin.c (format_tree): For '0', only set zero_flag if we haven't seen the field width or precision yet. Mon Aug 9 13:06:01 1999 Arnold D. Robbins * array.c (assoc_lookup): Removed code that gave each array a private copy of each index. Balloons memory usage for no good reason that I can see. Just use dupnode in all cases. * configure.in: check for $srcdir/.developing adds extra defines for my testing/debugging use. Yes, hack alert. Sun Aug 1 11:02:02 1999 Arnold D. Robbins * node.c (dupnode): turn off FIELD when copying nodes. * array.c (do_adump, assoc_dump): new functions for array debugging. * awk.y (tokentab): conditionally add "adump" function for debugging. * awk.h: delcare new functions. Thu Jul 29 23:26:40 1999 Arnold D. Robbins From wsanchez@apple.com: * Makefile.in (install-strip): new target, coding stds. compatibility. * config.guess, config.sub: Add MacOS X recognition. Thu Jul 29 19:09:19 1999 Arnold D. Robbins * awk.y (func_install): make `function foo(foo)' a fatal error. eval.c (r_tree_eval): diagnose use of a function name as a variable inside the function. Sun Jul 4 16:53:14 1999 Arnold D. Robbins * eval.c (eval_condition): add extra braces to avoid gcc warning. I'm not going to bother for the library code like dfa and regex. Wed Jun 30 16:14:36 1999 Arnold D. Robbins * Release 3.0.4: Release tar file made. This time for sure. Wed Jun 30 16:10:11 1999 Arnold D. Robbins * awk.h: add include of , and comment about config.h having to be included before any system headers. Otherwise, with egcs-2.91.66 and later on Linux systems, and possibly others, things break badly, due to the LFS macros. * awk.y, builtin.c, eval.c, field.c, io.c: removed include of assert.h Wed Jun 9 11:39:19 1999 Paul Eggert Port the large-file code to AIX, HP-UX, and IRIX. Add cross-compilation support for large files. * config.guess, config.sub: New files. * configure.in (AC_CANONICAL_HOST): Add; GAWK_AC_SYS_LARGEFILE needs this. (GAWK_AC_SYS_LARGEFILE): Renamed from GAWK_AC_LARGE_FILES. * aclocal.m4 (GAWK_AC_SYS_LARGEFILE): Renamed from GAWK_AC_LARGE_FILES. Add support for AIX and HP-UX. (GAWK_AC_SYS_LARGEFILE_FLAGS, GAWK_AC_SYS_LARGEFILE_SPACE_APPEND, GAWK_AC_SYS_LARGEFILE_MACRO_VALUE): New macros. * acconfig.h (_FILE_OFFSET_BITS, _LARGEFILE_SOURCE, _LARGE_FILES): New macros. * Makefile.in (MISC): add config.guess and config.sub so they get included in the distribution. Wed Jun 9 11:29:29 1999 Paul Eggert * io.c (iop_alloc): Don't mmap files whose sizes don't fit in `int'. [ This isn't really needed, as HAVE_MMAP is #undef'ed at the top, but it's there in case people want to take their life in their hands. ] Sun Jun 6 11:28:07 1999 Arnold D. Robbins * BETA Release 3.0.46: Release tar file made. Wed Jun 2 14:36:24 1999 Arnold D. Robbins * PORTS: Updated with a more recent list of systems that gawk compiles and tests ok on. Tue Jun 1 14:24:59 1999 Arnold D. Robbins * BETA Release 3.0.45: Release tar file made. Tue May 25 16:32:37 1999 Arnold D. Robbins * builtin.c (format_tree): more smarts for weird cases, such as zero precisions and zero values used with the `#' flag. Thanks to Andreas Schwab (schwab@gnu.org) for pointing these out. Wed May 19 14:02:54 1999 Arnold D. Robbins * io.c (do_close): move test for `close(FILENAME)' to after loop through all open redirections. Fixes problems in obscure cases with redirections in END rules. Sun May 16 14:08:39 1999 Arnold D. Robbins * awk.y (yylex): fix group of characters including ',' to set want_assign = FALSE. Fixes bizarre parsing problems in function call lists, for example. * io.c (get_a_record): repair logic for single-leading-newline case. Tue May 11 16:48:11 1999 Arnold D. Robbins * aclocal.m4 (GAWK_AC_AIX_TWEAK): new macro. * configure.in: call it * Makefile.in: (awklib/all): pass CFLAGS on to sub-make so that password programs will get AIX magic defines. Avoids having to tweak program code for those in doc/gawk.texi. Mon May 3 16:56:23 1999 Arnold D. Robbins * array.c (do_delete): don't free_temp(subs) until after all references to it are finished. Mon May 3 13:41:16 1999 Arnold D. Robbins * BETA Release 3.0.44: Release tar file made. Sun May 2 18:25:43 1999 Arnold D. Robbins * io.c (get_a_record): Do a really good job of stripping newlines from the front of records when RS = "" and there's only one newline at the front of the file, which the regex didn't catch. Wed Apr 28 12:27:49 1999 Arnold D. Robbins * configure.in: more HP stuff: fix the manual alloca code so that gawk will compile and link on HP systems. See the comments. Sun Apr 25 13:39:16 1999 Arnold D. Robbins * Makefile.in (gawk): add $(CFLAGS) to linking step. * configure.in: correctly do AC_FUNC_GETPGRP on HP systems too. Tue Apr 13 20:21:00 1999 Arnold D. Robbins * BETA Release 3.0.43: Release tar file made. Tue Apr 13 19:02:20 1999 Arnold D. Robbins * io.c (useropen, pidopen): add casts to int on arguments to silence gcc warnings. * regex.c (regcomp,regexec,regfree): add ifdef for APPLE. Thu Feb 4 10:38:02 1999 Arnold D. Robbins * custom.h: hacks for BeOS. Not documented in the manual right now. * configure.in: hacks for BeOS. Check for HP-UX and define C_ALLOCA if not using gcc. I wish they'd just fix bison already. Sun Dec 20 16:57:38 1998 Arnold D. Robbins * BETA Release 3.0.42: Release tar file made. Sun Nov 15 21:05:39 1998 Arnold D. Robbins * io.c (gawk_popen): Add WIN32 to list of systems that use the non-real-pipe version. From the PC gawk guys. Wed Nov 4 11:32:24 1998 Arnold D. Robbins * BETA Release 3.0.41: Release tar file made. Tue Nov 3 16:24:35 1998 Arnold D. Robbins * eval.c (r_get_lhs): Fix the cases for the special variables, don't unref their current value if it's the same as the internal copy; perhaps the current one is used in a concatenation or some other expression somewhere higher up in the call chain. Ouch. See test/getnr2tm.awk. Sun Nov 1 15:24:52 1998 Arnold D. Robbins * builtin.c (format_tree): improve handling of zero-fill when a precision is present. See test/zeroflag.awk. Wed Oct 28 20:40:17 1998 Arnold D. Robbins * eval.c (r_tree_eval): Case for Node_concat. Get lengths separately, in case one expression has a side effect that that changes another. Ugly, but it keeps gawk from core dumping. See test/nasty.awk. Sun Oct 18 21:27:24 1998 Arnold D. Robbins * awk.y (append_right): bug fix, if `list' or `new' are NULL, return `list', so that things don't break too badly. * regex.c (re_compile_fastmap): remove unused variable `num_regs'. Thu Oct 8 19:36:57 1998 Arnold D. Robbins * BETA Release 3.0.40: Release tar file made. Mon Jul 27 10:14:33 1998 Arnold D. Robbins * node.c (parse_escape): Remove assignment with side effects from ISXDIGIT test. Thanks to "Mihai T. LAZARESCU" for pointing this out. Mon Apr 27 11:31:32 1998 Arnold D. Robbins * main.c (usage): fix the email address for the bug list. (copyleft): update the copyright year. Mon Mar 23 21:22:32 1998 Arnold D. Robbins * eval.c (r_get_lhs): make sure that values of type Node_param_list don't have the FUNC flag set. This means we don't allow the use of a function name as a variable or array from within the function. Sun Mar 22 19:12:32 1998 Paul Eggert * aclocal.m4 (GAWK_AC_LARGE_FILES): new macro that checks for large file support, and updates CPPFLAGS, LDFLAGS, LIBS as needed. * configure.in: call GAWK_AC_LARGE_FILES. * Makefile.in (CPPFLAGS, LDFLAGS): Let autoconf configure. (COMPFLAGS): Add $(CPPFLAGS). Mon Mar 16 14:06:41 1998 Arnold D. Robbins * field.c (using_FIELDWIDTHS): new macro. (using_fieldwidths): use new macro. (do_split): in case for FS_DFLT, also check that we're not using FIELDWIDTHS. Otherwise, split() would use FIELDWIDTHS, not current value of FS. Oops. Sun Nov 16 20:08:59 1997 Arnold D. Robbins * builtin.c (sub_common): fix for count of matches in gsub from Geert.Debyser@esat.kuleuven.ac.be. Wed Oct 15 03:38:12 1997 Arnold D. Robbins * field.c (set_FS): Use `sc_parsefield' if the value of FS is not alphabetic OR if not ignoring case. Bug fix if IGNORECASE is true and FS happens to be '^'. Sheesh, talk about obscure. (rebuild_record): Add more smarts to the code that sets up the fields. Thanks to Alan J. Broder (ajb@dtmr.com). Sun Oct 5 11:56:52 1997 Arnold D. Robbins * configure.in: if ISC add -D_SYSV3 to CFLAGS, per email from Mario Vanoni (vanonim@dial.eunet.ch). Fri Sep 26 00:57:49 1997 Arnold D. Robbins * awk.y (append_right): return if either list is NULL. Prevents syntax errors from causing core dumps. Wed Sep 17 15:34:15 1997 Arnold D. Robbins * field.c (rebuild_record): set things up so that all fields point into the new record and release any changed fields without causing memory leaks. Avoids problems when fields are extended with the value of $0 or other fields and then $0 is assigned to. Mon Sep 15 16:12:55 1997 Arnold D. Robbins * builtin.c (do_print): when testing for NUMBER, make sure it's not a string too. Thanks to Michael Brennan for clarifying the semantics. Sun Sep 14 19:55:12 1997 Arnold D. Robbins * node.c (format_val): always format values ourselves: avoids problems if OFMT is bizarre, like %s. Sun Sep 14 00:08:53 1997 Arnold D. Robbins * io.c (get_a_record): replace all occurrences of the test `grRS == FALSE' with `RS_is_null' which makes ` RS = "\0" ' actually work, is clearer code, and actually makes use of the `RS_is_null' variable! Sun Aug 17 07:15:12 1997 Arnold D. Robbins * field.c (set_FS): Change logic to always set parse_field, even if FS hasn't changed. Thanks to Igor Sheyn for catching this. Wed Aug 6 21:04:37 1997 Arnold D. Robbins * io.c (VMS et al gawk_popen): use pclose, not fclose, if iop_alloc fails. Wed Jul 30 19:53:52 1997 Arnold D. Robbins * awk.y [variable]: fix case for subscript if $3 == NULL. Sun Jul 27 22:47:30 1997 Arnold D. Robbins * awk.y (get_src_buf): don't close file if it's stdin. Sun Jul 27 22:47:15 1997 Pat Rankin * io.c (#if VMS: vmsrtl_fileno): new routine. (#if VMS: fileno): new macro substituted for stdio one. Thu Jul 17 20:05:59 1997 Arnold D. Robbins * builtin.c (do_print): When OFMT != CONVFMT, create a new temporary node with just the numeric value valid and format it, and use that for printing. Avoids memory corruption. Wed Jul 16 10:01:16 1997 Arnold D. Robbins * regex.c: When SYNTAX_TABLE is defined, but not emacs, then CHAR_SET_SIZE is not defined, though used in regcomp. It should be taken out of #ifdef SYNTAX_TABLE. Fix from bug group, from Akim Demaille, demaille@inf.enst.fr. * awk.h (isnondecimal): make test a little smarter. builtin.c (nondec2awknum): add bailout for decimal numbers, e.g. `00.1'. Fix from Larry Schwimmer . Thu Jun 19 19:00:40 1997 Arnold D. Robbins * eval.c (interpret): case Node_K_next, Node_K_nextfile: fatal error if called from BEGIN or END. (Fixed completely Mon May 3 13:31:42 1999.) Mon Jun 9 22:40:04 1997 Arnold D. Robbins * builtin.c (nondec2awknum): Allow `f' and `F' in hexadecimal numbers. Gotta get more sleep... * array.c (assoc_lookup): Fix from Tom Karzes (karzes@equator.com) for memory leak when forcing type to Node_var_array. Thu May 15 12:49:08 1997 Arnold D. Robbins * Release 3.0.3: Release tar file made. Wed May 14 08:06:08 1997 Arnold D. Robbins * io.c (do_close): add lint warning if closing something that isn't open. Tue May 13 12:14:12 1997 Arnold D. Robbins * random.c, builtin.c: remove __GLIBC__ tests, since it breaks `make test'. I prefer consistency across platforms. * Makefile.in (gawk): undid April 25 changes and added comment. Putting COMPLAGS in breaks with -g on VMS POSIX. Sun May 11 14:48:04 1997 Darrell Hankerson * io.c [MSC_VER]: add cases for WIN32. * regex.c [MSC_VER]: add cases for WIN32. Sun May 11 07:04:01 1997 Arnold D. Robbins * builtin.c (do_print): in the loop that evaluates each expression to be printed, do a dupnode to avoid bizarre output. Thanks to Michal for finding this problem. * awk.y (yylex): fix scanning of hexadecimal constants. Wed May 7 15:09:25 1997 Arnold D. Robbins * io.c (get_a_record): fix casetable indexing with cast to int. Keeps Michal happy. Tue May 6 16:40:19 1997 Arnold D. Robbins * eval.c (func_call): removed unneeded variables. Mon May 5 21:17:37 1997 Pat Rankin * missing/strftime.c [case 'v', VMS_EXT]: for VMS date format, two digit day of month should not be zero padded on the 1st through the 9th. Mon May 5 06:33:47 1997 Arnold D. Robbins * regex.h, regex.c: merge with current GLIBC version. Mon May 5 06:33:47 1997 Pat Rankin * io.c (nextfile): move the check for null return from iop_open in the normal case and add one for the "no args" case. Fri Apr 25 16:52:33 1997 Arnold D. Robbins * array.c (grow_table): add a bunch more large primes so arrays can get really big. Thanks to christos@deshaw.com. * all files: remove ifdef'ed out code and update copyrights. * Makefile.in (gawk): add $(COMPFLAGS) to command line. * eval.c (flags2str): added case for FIELD. Thu Apr 24 22:39:23 1997 Arnold D. Robbins * COPYING: changed to current official version from FSF. * regex.c: merge with GLIBC version. * awk.h [_GNU_SOURCE]: bracket definition inside ifdef. (NODE.source_line): move name member out of `x' union and into `nodep'; avoids problems doing diagnostics. (nondec2num): put decl into #if BITOPS || NONDECDATA * posix/gawkmisc.c, missing/system.c, missing/strtod.c, missing/strerror.c: move to generic GPL statement at top. * builtin.c (nondec2num): put into #if BITOPS || NONDECDATA Wed Apr 23 22:14:14 1997 Arnold D. Robbins * dfa.c: misc changes for really pedantic SGI compilers. * builtin.c: bracket defs of random() etc for GLIBC. * random.c: bracket whole file for GLIBC. * configure.in: extra goop for GETPGRP test for VMS POSIX. * custom.h [VMS]: remove hard definition of GETPGRP_VOID. Fri Apr 18 07:55:47 1997 Arnold D. Robbins * BETA Release 3.0.34: Release tar file made. Tue Apr 15 21:35:45 1997 Arnold D. Robbins NEW UNDOCUMENTED FEATURE. USE THE SOURCE LUKE! * acconfig.h [NONDECDATA]: new macro. * awk.h: add decl of do_strtonum. * awk.y (tokentab): add entry for strtonum function. * builtin.c (do_strtonum): new function. * configure.in (non-decimal-data): new --enable-* option. * node.c (r_force_number): change to allow non-decimal data inside ifdef NONDECDATA. Tue Apr 15 06:32:50 1997 Pat Rankin * missing/strftime.c (malloc, realloc, getenv, strchr): only declare these when STDC_HEADERS is not defined. : include these when STDC_HEADERS is defined. * awk.h (freenode, tree_eval, m_tree_eval): reorganize definitions. * alloca.c (malloc): if malloc is already defined as a macro, presumeably by config.h, don't define or declare it. Wed Apr 9 22:45:27 1997 Arnold D. Robbins * Makefile.in [COMPFLAGS]: per suggestion from Karl Berry, put $(CFLAGS) last. Tue Apr 8 23:54:46 1997 Arnold D. Robbins * eval.c (interpret): For Node_K_break and Node_K_continue, if treating them like `next', also check the function call stack and pop it if necessary. Mon Apr 7 18:22:37 1997 Arnold D. Robbins * awk.h: Add decls of new routines do_compl() and set_loc(). * awk.y (tokentab): add entry for "compl" function. * builtin.c (do_compl): new function to do ones complement. (do_substr): rationalized yet again, now notices negative start and length parameters. * eval.c (push_args): fix if call_list gets realloc'ed in the middle of things. Avoids crash for deeply nested function calls. * main.c (catch_sig): add call to set_loc(). * msg.c (set_loc, srcfile, srcline): new function and private variables to help out in tracing down source of error messages. Fri Mar 28 08:42:27 1997 Arnold D. Robbins * io.c (iop_alloc, iop_close): Undo changes of Feb 11, apparently other cleanups in io.c made mmap stuff start working again. BAH! It's a mess, the test suite still fails. I'm leaving the mmap stuff undefined for now. It'll probably get ripped out in 3.1. Thu Mar 27 08:48:57 1997 Arnold D. Robbins * custom.h [_SEQUENT_]: undef HAVE_MMAP. Wed Mar 26 09:08:16 1997 Arnold D. Robbins * io.c (iop_alloc): fix definition to make it static. Mon Mar 24 23:09:07 1997 Arnold D. Robbins * field.c (init_fields, etc..): more clean up use of Null_field and the various flags. * node.c (unref): if a field, free the node itself. Fixes memory leak problems. Sun Mar 23 22:51:09 1997 Arnold D. Robbins * awk.h [FIELD]: new flag for node->flags field. * builtin.c (sub_common): if FIELD is set, dup the string. * field.c (init_fields): set up a new Null_field global var. (init_fields, set_field, set_record) use the FIELD flag. (getfield): use Null_field instead of private variable. * io.c (wait_any): comment out calls to pclose and iop_close, caused weird race conditions. See test/pipeio1.awk. Thanks to Darrell Hankerson for tracing this one down. Tue Mar 18 20:57:18 1997 Arnold D. Robbins * dfa.c (inboth): free templist; plugs memory leak. * field.c (init_fields, grow_fields_arr, set_field, rebuild_record, set_record): remove PERM flag from entries in fields_arr[]. Fixes nasty memory leak. Tue Mar 18 06:33:00 1997 Arnold D. Robbins * awk.y (dup_parms): robustified against parameter errors. Sun Mar 16 21:31:40 1997 Arnold D. Robbins NEW UNDOCUMENTED FEATURE. USE THE SOURCE LUKE! * acconfig.h [BITOPS]: new macro. If set, do octal & hex and bit ops. * awk.h [isnondecimal]: new macro, and decl of new functions. * awk.y (yylex): add recognition of octal and hex constants. * builtin.c (do_and, do_or, do_xor, do_lshift, do_rshift): new functions that do bit operations. (nondec2awknum): new function to convert octal or hex to double. * configure.in: Add AC_ARG_ENABLE for bit operations. * node.c (r_force_number): add octal and hex conversion. Sun Mar 16 21:28:56 1997 Arnold D. Robbins * awk.h [IOP_NOFREE_OBJ]: new macro. * io.c (iop_open, iop_alloc): add new third parameter, which is either NULL, meaning allocate a new IOP, or the address of one already allocated. Have a static one in the `nextfile' routine, and use the IOP_NOFREE_OBJ flag for it. All of this keeps us from reading freed memory. The `swaplns' test fails otherwise. (iop_close): if IOP_NOFREE_OBJ is set, don't free the IOBUF. Wed Feb 26 06:21:02 1997 Arnold D. Robbins * eval.c (in_function, pop_fcall_stack, pop_fcall, push_args): new functions. These manage "frames" of awk function call arguments. The problem is that a `next' or a `nextfile' from a function leaks memory. These changes allow us to free up that memory. (interpret): for Node_K_next and Node_K_nextfile, check if in a function call and free all function call frames. Fri Feb 21 06:23:19 1997 Arnold D. Robbins * Misc changes from Katsuyuki Okabe : * builtin.c (do_substr): change a %d to %ld in warning message. * eval.c (op_assign): fix format string for warning about %=. Wed Feb 19 23:29:02 1997 Arnold D. Robbins * main.c (main): add do_intervals to condition that causes resetup() to be called again. Makes the --re-interval option actually work. What a concept. Fri Feb 14 09:47:31 1997 Arnold D. Robbins * io.c [#include "awk.h"]: undef HAVE_MMAP to just use the old code. Something is causing a file descriptor leak, and this is getting to be just too much hair. I reserve the right to rip out the mmap code entirely at a future date. Tue Feb 11 06:28:29 1997 Arnold D. Robbins * io.c (iop_alloc): for an mmap'ed file, close the file descriptor, and then touch each page to get a private copy. Fixes nasty case of truncating our input file. (iop_close): don't call close on mmap'ed file. Wed Feb 5 17:59:04 1997 Arnold D. Robbins * eval.c (interpret): For Node_K_delete, just call do_delete; let it handle the case of `delete array'. * array.c (do_delete): Changed to handle case of `delete array', and made smarter if the array is actually an uninitialized parameter. Sun Jan 26 22:58:29 1997 Arnold D. Robbins * getopt.h, getopt.c, getopt1.c: replaced with new versions from GLIBC 2. Sun Jan 19 23:37:03 1997 Arnold D. Robbins * eval.c (nodetype2str): not static, for debugging. (flags2str) new function: for debugging. * field.c (get_field): add new var that is like Nnull_string but does not have numeric attributes, so that new fields are strings. (set_record): turn off PERM flag before unrefing fields and field 0. * array.c (in_array): always evaluate subscript, could have side effects. * builtin.c (do_strftime): way increase size of buffer to make sure we don't have overflow problem. Keeps Paul Eggert happy. * custom.h [__amigaos__]: define fork to vfork. From Fred Fish. * dfa.c: move include of config.h to top, for RSXNT. From Kai Uwe Rommel. (ISALPHA, etc): change from Jacob Engelbrecht (jaen@novo.dk) to better handle non-ascii environments. * gawkmisc.c: remove amigados case, posix should now work fine. * amiga/*: nuked per previous entry. * Makefile.in: removed all references to amiga * io.c [HAVE_SYS_PARAM_H]: Add #undef RE_DUP_MAX to avoid spurious conflict with regex.h. (flush_io): remove amiga ifdefs, not needed anymore. (spec_setup): set getrec field for special files. Fix from Mark Gray (markgray@pdt.net). * node.c (more_nodes): fix to get the last entry in the array. Wed Jan 8 17:42:37 1997 Andreas Schwab * io.c (mmap_get_record): Fix return value if file ends without record separator. Fri Jan 3 19:57:16 1997 Pat Rankin * awk.y (get_src_buf): Test for an empty source file by detecting an initial read of 0 bytes rather than by relying on info from stat(). Wed Dec 25 11:25:22 1996 Arnold D. Robbins * Release 3.0.2: Release tar file made. Wed Dec 25 11:17:32 1996 Arnold D. Robbins * Makefile.in (install, uninstall): use $(srcdir)/patchlevel.h. Thanks to Richard Levitte, LeViMS@stacken.kth.se. (install): remove chmod command; let $(INSTALL_PROGRAM) use -m. Mon Dec 23 20:36:59 1996 Pat Rankin * custom.h (#if VMS_POSIX): Define GETPGRP_VOID. Fri Dec 20 08:59:55 1996 Arnold D. Robbins * getopt.c, getopt1.c: comment out the `#if defined (_LIBC) || !defined (__GNU_LIBRARY__)' and `#endif' to force use of this getopt, even on systems like linux. This will be handled better in 3.1 / glibc 2. Thu Dec 19 22:52:39 1996 Arnold D. Robbins * awk.y (yylex): In several places, after yyerror(), add call to exit(). Otherwise, infinite messages. This should probably be handled better. Wed Dec 18 22:42:10 1996 Darrel Hankerson * getopt.c (_getopt_internal): if 'W' and ';', if optind == argc, return c, don't fall through. Wed Dec 18 10:09:44 1996 Arnold D. Robbins * configure.in [AC_PREREQ]: Update to 2.12 in order to switch to autoconf 2.12. Lots of other files will be rebuilt automatically. [AM_SANITY_CHECK_CC]: Removed, autoconf does it now. * aclocal.m4 [AM_SANITY_CHECK_CC]: Removed, autoconf does it now. Tue Dec 17 22:23:16 1996 Arnold D. Robbins * builtin.c (do_strftime): fix case if format string is "". Also fix it if format is not "" but result of strftime is "". See comments in code. Tue Dec 10 23:09:26 1996 Arnold D. Robbins * Release 3.0.1: Release tar file made. Tue Dec 10 22:39:41 1996 Arnold D. Robbins * Makefile.in (dist): add dependency on `info'. Remove line that does makeinfo. (install): use $(LN) not $(LN_S) to link gawk gawk-version. Sun Dec 8 07:53:44 1996 Arnold D. Robbins * Makefile.in (gawk): took COMPFLAGS out of link line for help on VMS posix. Shouldn't (I hope) affect anything else. Thu Nov 28 11:52:24 1996 Arnold D. Robbins * configure.in (AC_PROG_INSTALL): Set INSTALL to install-sh. Tue Nov 26 22:42:00 1996 Arnold D. Robbins * PORTS: Updated list of systems. * Makefile.in (install): Fix some typos and add some improvements for Ultrix. Sun Nov 24 22:16:26 1996 Arnold D. Robbins * builtin.c (do_printf): if no args, fatal error. Return silently if --traditional. Thu Nov 7 20:54:43 1996 Arnold D. Robbins * io.c (inrec): make sure EOF hasn't already happened before trying to read; prevents accessing freed buffer. Thanks to Michal Jaegermann. * Makefile.in [AWKSRC]: add random.h. random.h: new file, redefines names of the `random' functions. random.c, builtin.c: add include of random.h. Thu Nov 7 09:06:21 1996 Arnold D. Robbins * awk.y (snode): undo 4 Oct change, put do_split code back. field.c (do_split): restore old code; add test for CONST, so that re_parse_field is used if third arg to split is a regexp constant. Mon Nov 4 12:57:11 1996 Arnold D. Robbins * main.c (main): Research -m[fr] options don't need literal '=' characters. Brian's documentation was confusing. Fixed, not that anyone actually uses these options with gawk. Sun Nov 3 11:23:21 1996 Arnold D. Robbins * field.c (def_parse_field): add \n to list of acceptable white space. (posix_def_parse_field): new routine, just like def_parse_field(), but only allows space and tab as separators. (do_split, set_FS): make appropriate choice between the two *def_parse_field() routines. Fri Oct 25 10:13:06 1996 Arnold D. Robbins * configure.in: remove test for random. * Makefile.in: add random.c to list of files always compiled. * missing.c: remove HAVE_RANDOM test. * builtin.c: remove ifdef's for HAVE_RANDOM. [GAWK_RAND_MAX]: use constant we know works with our random(). * random.c: new file - moved from missing/ directory. Wed Oct 23 19:46:01 1996 Pat Rankin * builtin.c (do_tolower, do_toupper): Add `unsigned char *' casts. Tue Oct 22 21:27:52 1996 Arnold D. Robbins * builtin.c [GAWK_RANDOM_MAX]: Try to make definition a bit smarter; don't use RAND_MAX if it's equal to SHRT_MAX, blows things up. Tue Oct 22 08:49:20 1996 Arnold D. Robbins * main.c (copyleft): update copyright date to 1996. too many files to list: update copyright date to 1996. Sun Oct 20 12:21:09 1996 Arnold D. Robbins * awk.y, dfa.c, eval.c, io.c, re.c: added various FIXME comments. Sat Oct 19 22:06:42 1996 Arnold D. Robbins * eval.c (nodetype2str): make static, add prototype. * field.c (sc_parse_field): cast array subscripts to int to shut up gcc warnings. * gawkmisc.c: add prototype for xmalloc. * awk.h: add prototype for getredirect. * builtin.c (do_fflush): remove extern decl of getredirect. * io.c (get_a_record, mmap_get_record): change decl of rs to int, to shut up gcc warnings. * awk.y (isassignable): add a default to switch to quiet gcc. * getopt.c (_getopt_internal): give default value to `indfound'. Fri Oct 18 09:00:49 1996 Arnold D. Robbins * regex.h [RE_SYNTAX_AWK]: add RE_CONTEXT_INDEP_ANCHORS. Thu Oct 17 22:32:55 1996 Arnold D. Robbins * aclocal.m4 [AM_SANITY_CHECK_CC]: added. * configure.in: use it. Thu Oct 17 21:43:25 1996 Arnold D. Robbins * configure.in: add checks for locale.h and setlocale(). awk.h: include locale.h and define out setlocale() if not available. main.c (main): call setlocale(). builtin.c (do_tolower, do_toupper): use unsigned char pointers, to get other charsets right in different locales. Wed Oct 16 21:32:53 1996 Arnold D. Robbins * builtin.c (format_tree): Change initial buffer size to 512 and use a constant. Allows large values of %f per bug report from sheyn@cs.bu.edu. Wed Oct 16 21:22:08 1996 Arnold D. Robbins * Makefile.in [MISC]: removed TAGS and tags (local-distclean): added TAGS and tags (maintainer-clean): removed TAGS and tags Wed Oct 16 12:28:43 1996 Arnold D. Robbins * main.c (version): Add call to copyleft(), per new standards. version.c: Fix text of version string to match new standards. Sun Oct 6 22:19:45 1996 Arnold D. Robbins * regex.c: updated to Emacs 19.34b base. Sun Oct 6 21:57:34 1996 Arnold D. Robbins * re.c (make_regexp): fixed to handle \8 and \9 in the middle of a regexp. Fri Oct 4 10:26:16 1996 Arnold D. Robbins * awk.y (snode): remove case for do_split; always making the third arg a Node_regex is wrong. field.c (do_split): rationalized to distinguish `/ /' from `" "'. Generally fixed up. * node.c (parse_escape): Allow single digit \x escapes. 1996-10-02 Paul Eggert * builtin.c (format_tree): Fix bug in %d and %i format: NaNs, and values in the range LONG_MAX+1 .. ULONG_MAX, were mishandled. Don't assume that double values <= -1 are converted to unsigned long in the expected way; the C Standard doesn't guarantee this. 1996-10-02 Paul Eggert * awk.h (INT_MAX): Remove unused symbol. Mon Sep 30 22:19:11 1996 Arnold D. Robbins * getopt.c (_getopt_internal): If 'W' is in the optstring followed by a ';' then search through the long opts table. This makes `-W foo=bar' same as `--foo=bar'. * main.c (main): 'W' now prints an error message. (gawk_option): deleted the routine. Sun Sep 29 23:04:54 1996 Arnold D. Robbins * builtin.c (sub_common): fix several bugs with gsub when matching null strings. See test/gsubtest.awk. Fri Sep 20 17:35:54 1996 Pat Rankin * alloca.c (NULL): don't define if has already done so. Fri Sep 20 11:54:31 1996 Arnold D. Robbins * builtin.c (do_print): evaluate all the expressions first and then print them. Avoids surprising behavior. See test/prtoeval.awk for an example. Tue Sep 10 06:21:40 1996 Arnold D. Robbins * awk.h [FUNC]: new flag, marks a Node_parameter_list as really being the function name; allows more checking in awk.y. * awk.y (isassignable): now takes a NODE * instead of a type, to check if a function parameter is marked FUNC, then it's the function name, which is not assignable. Fix call from snode(). (function_prologue): mark function name as FUNC. (yyerror): don't call exit() anymore; gawk will now report all syntax errors. Sun Sep 1 19:36:30 1996 Arnold D. Robbins * field.c (rebuild_record): after building new field 0, go through all old fields, and if they used to point into the old one, have them point into the new one. Then turn off PERM flag before unref-ing field 0. Wed Aug 28 19:13:34 1996 Arnold D. Robbins * eval.c (set_IGNORECASE): Correctly parenthesize bit operations in test and fix logic for string value. Wed Aug 28 22:06:33 1996 Arnold D. Robbins * main.c (usage): add email addresses for bug reporting, per change in GNU Coding Standards from RMS. Sun Aug 11 23:13:22 1996 Arnold D. Robbins * Makefile.in (install): correct use of $(INSTALL_PROGRAM). Thu Aug 8 23:29:43 1996 Arnold D. Robbins * parse.y (isassignable): new function, checks in type can be assigned to. (snode): changed checking for 3rd arg of gsub to be more general, supersedes earlier change. Thu Aug 8 13:58:26 1996 Arnold D. Robbins * parse.y (snode): If third arg to sub or gsub is builtin function, complain, since can't substitute into result. * eval.c (r_get_lhs): diagnose Node_builtin as an error, instead of falling through into default case and using cant_happen(). Thu Aug 1 07:13:14 1996 Arnold D. Robbins * regex.h [RE_DEBUG]: new macro. [RE_SYNTAX_GNU_AWK]: add RE_DEBUG. [RE_SYNTAX_POSIX_AWK]: add RE_INTERVALS. * regex.c (re_set_syntax): add #ifdef DEBUG code to turn on `debug' flag if RE_DEBUG set, and turn off debug if not set and debug was on. * main.c (main): remove `do_intervals = TRUE' from `if (do_posix)', it's now handled in the definition of RE_SYNTAX_POSIX_AWK. Mon Jul 29 17:49:07 1996 Pat Rankin * io.c (O_ACCMODE): define it if doesn't. Mon Jul 29 12:02:48 1996 Arnold D. Robbins * eval.c (set_IGNORECASE): made somewhat smarter. gawk -v IGNORECASE=0 was acting the same as -v IGNORECASE=1. Thanks to Darrell Hankerson for the bug report. Fri Jul 26 12:04:43 1996 Arnold D. Robbins * awk.h (format_val): add declaration of new routine. * node.c (format_val): new routine, abstracts old guts of r_forcestring; accepts format string and index as additional params. (r_force_string): changed to call format_val. * builtin.c (do_print): don't tree_eval the tree twice in case OFMTidx != CONVFMTidx; doing so could cause side effects (from bug report by Tobias Rettstadt, xassp@ipds.uni-kiel.de). Instead, call format_val. Mon Jul 22 21:59:15 1996 Arnold D. Robbins * io.c (iop_close): change check for "is $0 in the input buffer" to use `< (iop->buf + iop->secsiz + iop->size)' instead of `< iop->end'. The latter is bogus if EOF has been hit on the file. Fix from Darrel Hankerson based on bug report by Charles Howes (howes@grid.direct.ca). See test/eofsplit.awk. Thu Jul 18 19:43:20 1996 Arnold D. Robbins * builtin.c (sub_common): backed out change of Feb 14 in favor of: (do_gensub): Changed to use make_string and then to |= TEMP flag, based on bug report and patch from Katsuyuki Okabe, hgc02147@niftyserve.or.jp. Thu Jul 18 19:23:53 1996 Arnold D. Robbins * custom.h: added ifdef for QNX, based on bug report from Michael Hunter, mphunter@qnx.com. Mon Jul 15 09:31:01 1996 Arnold D. Robbins * io.c (redirect): When finding the rp pointer, if it's not NULL, set str = rp->value. This gets the '\0' terminated version. Motivated by bug report from John Hawkinson (jhawk@bbnplanet.com). Sun Jul 14 18:40:26 1996 Arnold D. Robbins * configure.in: added call to AC_CHECK_LIB(m, fmod), since apparently some systems have fmod in the math library. Portability: the Holy Grail. Sigh. Sun Jul 14 18:08:01 1996 Arnold D. Robbins * awk.h: add Jim Meyerings ISASCII etc hacks for ctype macros. * builtin.c (do_toupper, do_tolower, sub_common): changed to use upper-case versions of ctype macros. * main.c (main): ditto. * node.c (r_force_number, parse_escape): ditto. Sun Jul 14 06:34:18 1996 Arnold D. Robbins * field.c (set_record): made it always do the PERM flag. Fixes cases where $0 is assigned to, e.g. by gsub, keeps the fields valid. (get_field): removed the call to reset_record in case where ! field0_valid. We want to leave the fields alone if they've been changed. Thu Jul 11 23:04:20 1996 Arnold D. Robbins * io.c (devopen): change tests of (flag & O_fooONLY) to (flag & O_ACCMODE) == O_fooONLY. Per (long standing) bug report from Chapman Flack. (close_redir): change final conditional to just (status != 0) so that ERRNO always set; the warning had its own `if (do_lint)' anyway. * eval.c (do_split): force type of array to be Node_var_array instead of Node_var. Per (long standing) bug report from Chapman Flack. Thu Jul 11 22:17:14 1996 Arnold D. Robbins * Makefile.in (install): added symlink of gawk to awk if no awk in $(bindir). (LN_S): new variable for symlinking. (uninstall): remove awk if it's the same gawk. * Configure.in: Added call to AC_PROG_LN_S for Makefile.in. Sun Jul 7 15:47:13 1996 Arnold D. Robbins * main.c (main): made `--posix' turn on interval expressions. Gawk now matches its documentation. (What a concept!) Wed Jul 3 15:02:48 1996 Arnold D. Robbins * regex.h, regex.c: upgraded to changes from Emacs 19.31. Fri May 17 08:46:07 1996 Arnold D. Robbins * io.c (get_a_record): added `continued' flag. Fix from Darrell Hankerson for when RS = "\n|something". Wed May 15 02:34:55 1996 Arnold D. Robbins * Makefile.in (awklib/all): now depends on gawk, fixes problem with parallel make. Tue May 14 15:02:52 1996 Arnold D. Robbins * builtin.c (format_tree): fix handling of '*' to deal with negative value for fieldwidth -- make positive and turn on left justify. Per bug report from Michael Brennan. Sun May 12 20:42:06 1996 Arnold D. Robbins * eval.c (r_get_lhs): case Node_subscript. Check if array name is actually a function, fatal error if so. Sun May 5 10:11:52 1996 Arnold D. Robbins * io.c (redirect): call flush_io() before creating a new output pipe, per bug report from Brian Kernighan (bwk@research.bell-labs.com). Fri Mar 15 06:38:33 1996 Arnold D. Robbins * Makefile.in (install): use $(INSTALL_PROGRAM), not $(INSTALL). (local-distclean): add `*~' to list of files to be removed. (CFLAGS): now contains just @CFLAGS@. (COMPFLAGS): replaces use of CFLAGS, has CFLAGS plus all the other stuff. Wed Mar 13 14:19:38 1996 Arnold D. Robbins * io.c (mmap_get_record): fixed to not place sentinel at end of mmap'ed object. Won't work if file is exact multiple of disk block size. See comments in code for more info. Thanks to Rick Adams (rick@uunet.uu.net) for help in testing. Sun Mar 10 22:50:23 1996 Arnold D. Robbins * io.c (do_close): notice if we were called as `close(FILENAME)' and arrange to close the current input file. This turns out to be easy to do, just call `nextfile(TRUE)'. Based on bug report from Pascal A. Dupuis, . Thu Mar 7 08:08:51 1996 Arnold D. Robbins * field.c (init_fields, grow_fields, set_field, rebuild_record): Nuke the `nodes' array everywhere. Anytime a field is unref'ed, allocate a new node that is a copy of Nnull_string. This avoids subtle memory management problems when doing a lot of assignment to fields, and tweaking of NF. Make sure that fields_arr[0] always has a type of Node_val! * field.c (set_NF): If NF is decremented, clear fields between NF and parse_high_water, otherwise if NF incremented, clear fields between parse_high_water and NF. * eval.c (nodetype2str): new function, used for diagnostics. eval.c (interpret): use nodetype2str when finding invalid node. Mon Mar 4 09:02:28 1996 Arnold D. Robbins * builtin.c (do_toupper, do_tolower): use isascii along with isupper/islower before changing case, in case characters have the high bit set. This is a hack. Mon Feb 26 22:24:44 1996 Arnold D. Robbins * builtin.c (sub_common): if no match, and called from gensub, don't free the temporary string, since the tmp_number then writes over it. Sun Feb 25 23:13:01 1996 Arnold D. Robbins * builtin.c (format_tree): fixed %c to treat user input as numeric also by adding test for MAYBE_NUM. Tue Feb 20 12:25:50 1996 Arnold D. Robbins * configure.in: Added AC_FUNC_MMAP call and add madvise to list of functions to look for. * awk.h [IOP_ISMAPPED]: new flag value for mmap support and new `getrec' structure member in struct iobuf. * io.c (iop_alloc, iop_close): changed to map/unmap input file into memory if possible. (mmap_get_record): new function to actually retrieve the record from mmaped file. Thu Feb 1 08:56:46 1996 Arnold D. Robbins * builtin.c (do_substr): fixed lint message to use indx+1 when start position is past end of string. Sun Jan 28 07:00:56 1996 Arnold D. Robbins * builtin.c (do_substr): rationalized handling of missing length argument, as well as various accompanying lint warnings. Previous code was slightly bogus. Talk about your Day 1 bugs. Thu Jan 25 14:09:11 1996 Arnold D. Robbins * builtin.c (do_substr): if length exceeds length of actual string, do computation of needed substring length *after* the lint warning. Wed Jan 24 10:06:16 1996 Arnold D. Robbins * Makefile.in (gawk): Add $(CFLAGS) to link line. (Makefile): target depends on the Makefile.in files. (OTHERS): Added TAGS and tags to the distribution. (local-distclean): New rule. (distclean): Use it. (maintainer-clean): Don't `make distclean' before running submakes, since that removes makefiles needed for the submakes. * builtin.c (do_strftime): Remove hard coded limit on length of result. Based on code from Paul Eggert (eggert@twinsun.com). Mon Jan 22 13:16:37 1996 Arnold D. Robbins * main.c (usage): takes new fp parameter which is either stdout for `--help' (per the GNU Coding Standards) or stderr if an error occurs. Fix all calls. (version): prints to stdout per the coding stds. (copyleft): prints to stdout now, not stderr, and exits. Fri Jan 19 08:10:29 1996 Arnold D. Robbins * regex.h [RE_GNU_AWK]: added RE_CONTEXT_INDEP_OPS to set of bits we turn off for regular operation. Breaks things like /^+[0-9]+/ to match a literal `+' at the beginning of, say, a phone number. Wed Jan 10 23:19:36 1996 Arnold D. Robbins * 3.0.0 polished up and release tar file made. Wed Dec 27 11:46:16 1995 Arnold D. Robbins * 2.94.0 released to porting group (no, I haven't been good about this file; I'll do better once 3.0 is released). Mon Aug 28 23:04:30 1995 Arnold D. Robbins * awk.h updated for NeXT - bracket TRUE/FALSE * io.c (get_a_record): removed shadowing of 'start' in * Makefile.in and doc/Makefile.in: fixed to use gawk.1 and gawk.texi, instead of gawk.1.in and gawk.texi.in. Mon Aug 25 11:04:30 1995 Arnold D. Robbins * 2.90.0 released to porting group. Fri Aug 18 12:43:31 1995 Arnold D. Robbins * ChangeLog created.