xref: /freebsd/contrib/one-true-awk/FIXES (revision d86a0988d2ea55252f08da2dfc8db09d7f8aba1b)
1/****************************************************************
2Copyright (C) Lucent Technologies 1997
3All Rights Reserved
4
5Permission to use, copy, modify, and distribute this software and
6its documentation for any purpose and without fee is hereby
7granted, provided that the above copyright notice appear in all
8copies and that both that the copyright notice and this
9permission notice and warranty disclaimer appear in supporting
10documentation, and that the name Lucent Technologies or any of
11its entities not be used in advertising or publicity pertaining
12to distribution of the software without specific, written prior
13permission.
14
15LUCENT DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE,
16INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS.
17IN NO EVENT SHALL LUCENT OR ANY OF ITS ENTITIES BE LIABLE FOR ANY
18SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
19WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER
20IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION,
21ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF
22THIS SOFTWARE.
23****************************************************************/
24
25This file lists all bug fixes, changes, etc., made since the AWK book
26was sent to the printers in August, 1987.
27
28Aug 7, 2011:
29	split(s, a, //) now behaves the same as split(s, a, "")
30
31Jun 12, 2011:
32	/pat/, \n /pat/ {...} is now legal, though bad style to use.
33
34	added checks to new -v code that permits -vnospace; thanks to
35	ruslan ermilov for spotting this and providing the patch.
36
37	removed fixed limit on number of open files; thanks to aleksey
38	cheusov and christos zoulos.
39
40	fixed day 1 bug that resurrected deleted elements of ARGV when
41	used as filenames (in lib.c).
42
43	minor type fiddles to make gcc -Wall -pedantic happier (but not
44	totally so); turned on -fno-strict-aliasing in makefile.
45
46May 6, 2011:
47	added #ifdef for isblank.
48	now allows -ffoo as well as -f foo arguments.
49	(thanks, ruslan)
50
51May 1, 2011:
52	after advice from todd miller, kevin lo, ruslan ermilov,
53	and arnold robbins, changed srand() to return the previous
54	seed (which is 1 on the first call of srand).  the seed is
55	an Awkfloat internally though converted to unsigned int to
56	pass to the library srand().  thanks, everyone.
57
58	fixed a subtle (and i hope low-probability) overflow error
59	in fldbld, by adding space for one extra \0.  thanks to
60	robert bassett for spotting this one and providing a fix.
61
62	removed the files related to compilation on windows.  i no
63	longer have anything like a current windows environment, so
64	i can't test any of it.
65
66May 23, 2010:
67	fixed long-standing overflow bug in run.c; many thanks to
68	nelson beebe for spotting it and providing the fix.
69
70	fixed bug that didn't parse -vd=1 properly; thanks to santiago
71	vila for spotting it.
72
73Feb 8, 2010:
74	i give up.  replaced isblank with isspace in b.c; there are
75	no consistent header files.
76
77Nov 26, 2009:
78	fixed a long-standing issue with when FS takes effect.  a
79	change to FS is now noticed immediately for subsequent splits.
80
81	changed the name getline() to awkgetline() to avoid yet another
82	name conflict somewhere.
83
84Feb 11, 2009:
85	temporarily for now defined HAS_ISBLANK, since that seems to
86	be the best way through the thicket.  isblank arrived in C99,
87	but seems to be arriving at different systems at different
88	times.
89
90Oct 8, 2008:
91	fixed typo in b.c that set tmpvec wrongly.  no one had ever
92	run into the problem, apparently.  thanks to alistair crooks.
93
94Oct 23, 2007:
95	minor fix in lib.c: increase inputFS to 100, change malloc
96	for fields to n+1.
97
98	fixed memory fault caused by out of order test in setsval.
99
100	thanks to david o'brien, freebsd, for both fixes.
101
102May 1, 2007:
103	fiddle in makefile to fix for BSD make; thanks to igor sobrado.
104
105Mar 31, 2007:
106	fixed some null pointer refs calling adjbuf.
107
108Feb 21, 2007:
109	fixed a bug in matching the null RE in sub and gsub.  thanks to al aho
110	who actually did the fix (in b.c), and to wolfgang seeberg for finding
111	it and providing a very compact test case.
112
113	fixed quotation in b.c; thanks to Hal Pratt and the Princeton Dante
114	Project.
115
116	removed some no-effect asserts in run.c.
117
118	fiddled maketab.c to not complain about bison-generated values.
119
120	removed the obsolete -V argument; fixed --version to print the
121	version and exit.
122
123	fixed wording and an outright error in the usage message; thanks to igor
124	sobrado and jason mcintyre.
125
126	fixed a bug in -d that caused core dump if no program followed.
127
128Jan 1, 2007:
129	dropped mac.code from makefile; there are few non-MacOSX
130	mac's these days.
131
132Jan 17, 2006:
133	system() not flagged as unsafe in the unadvertised -safe option.
134	found it while enhancing tests before shipping the ;login: article.
135	practice what you preach.
136
137	removed the 9-years-obsolete -mr and -mf flags.
138
139	added -version and --version options.
140
141	core dump on linux with BEGIN {nextfile}, now fixed.
142
143	removed some #ifdef's in run.c and lex.c that appear to no
144	longer be necessary.
145
146Apr 24, 2005:
147	modified lib.c so that values of $0 et al are preserved in the END
148	block, apparently as required by posix.  thanks to havard eidnes
149	for the report and code.
150
151Jan 14, 2005:
152	fixed infinite loop in parsing, originally found by brian tsang.
153	thanks to arnold robbins for a suggestion that started me
154	rethinking it.
155
156Dec 31, 2004:
157	prevent overflow of -f array in main, head off potential error in
158	call of SYNTAX(), test malloc return in lib.c, all with thanks to
159	todd miller.
160
161Dec 22, 2004:
162	cranked up size of NCHARS; coverity thinks it can be overrun with
163	smaller size, and i think that's right.  added some assertions to b.c
164	to catch places where it might overrun.  the RE code is still fragile.
165
166Dec 5, 2004:
167	fixed a couple of overflow problems with ridiculous field numbers:
168	e.g., print $(2^32-1).  thanks to ruslan ermilov, giorgos keramidas
169	and david o'brien at freebsd.org for patches.  this really should
170	be re-done from scratch.
171
172Nov 21, 2004:
173	fixed another 25-year-old RE bug, in split.  it's another failure
174	to (re-)initialize.  thanks to steve fisher for spotting this and
175	providing a good test case.
176
177Nov 22, 2003:
178	fixed a bug in regular expressions that dates (so help me) from 1977;
179	it's been there from the beginning.  an anchored longest match that
180	was longer than the number of states triggered a failure to initialize
181	the machine properly.  many thanks to moinak ghosh for not only finding
182	this one but for providing a fix, in some of the most mysterious
183	code known to man.
184
185	fixed a storage leak in call() that appears to have been there since
186	1983 or so -- a function without an explicit return that assigns a
187	string to a parameter leaked a Cell.  thanks to moinak ghosh for
188	spotting this very subtle one.
189
190Jul 31, 2003:
191	fixed, thanks to andrey chernov and ruslan ermilov, a bug in lex.c
192	that mis-handled the character 255 in input.  (it was being compared
193	to EOF with a signed comparison.)
194
195Jul 29, 2003:
196	fixed (i think) the long-standing botch that included the beginning of
197	line state ^ for RE's in the set of valid characters; this led to a
198	variety of odd problems, including failure to properly match certain
199	regular expressions in non-US locales.  thanks to ruslan for keeping
200	at this one.
201
202Jul 28, 2003:
203	n-th try at getting internationalization right, with thanks to volker
204	kiefel, arnold robbins and ruslan ermilov for advice, though they
205	should not be blamed for the outcome.  according to posix, "."  is the
206	radix character in programs and command line arguments regardless of
207	the locale; otherwise, the locale should prevail for input and output
208	of numbers.  so it's intended to work that way.
209
210	i have rescinded the attempt to use strcoll in expanding shorthands in
211	regular expressions (cclenter).  its properties are much too
212	surprising; for example [a-c] matches aAbBc in locale en_US but abBcC
213	in locale fr_CA.  i can see how this might arise by implementation
214	but i cannot explain it to a human user.  (this behavior can be seen
215	in gawk as well; we're leaning on the same library.)
216
217	the issue appears to be that strcoll is meant for sorting, where
218	merging upper and lower case may make sense (though note that unix
219	sort does not do this by default either).  it is not appropriate
220	for regular expressions, where the goal is to match specific
221	patterns of characters.  in any case, the notations [:lower:], etc.,
222	are available in awk, and they are more likely to work correctly in
223	most locales.
224
225	a moratorium is hereby declared on internationalization changes.
226	i apologize to friends and colleagues in other parts of the world.
227	i would truly like to get this "right", but i don't know what
228	that is, and i do not want to keep making changes until it's clear.
229
230Jul 4, 2003:
231	fixed bug that permitted non-terminated RE, as in "awk /x".
232
233Jun 1, 2003:
234	subtle change to split: if source is empty, number of elems
235	is always 0 and the array is not set.
236
237Mar 21, 2003:
238	added some parens to isblank, in another attempt to make things
239	internationally portable.
240
241Mar 14, 2003:
242	the internationalization changes, somewhat modified, are now
243	reinstated.  in theory awk will now do character comparisons
244	and case conversions in national language, but "." will always
245	be the decimal point separator on input and output regardless
246	of national language.  isblank(){} has an #ifndef.
247
248	this no longer compiles on windows: LC_MESSAGES isn't defined
249	in vc6++.
250
251	fixed subtle behavior in field and record splitting: if FS is
252	a single character and RS is not empty, \n is NOT a separator.
253	this tortuous reading is found in the awk book; behavior now
254	matches gawk and mawk.
255
256Dec 13, 2002:
257	for the moment, the internationalization changes of nov 29 are
258	rolled back -- programs like x = 1.2 don't work in some locales,
259	because the parser is expecting x = 1,2.  until i understand this
260	better, this will have to wait.
261
262Nov 29, 2002:
263	modified b.c (with tiny changes in main and run) to support
264	locales, using strcoll and iswhatever tests for posix character
265	classes.  thanks to ruslan ermilov (ru@freebsd.org) for code.
266	the function isblank doesn't seem to have propagated to any
267	header file near me, so it's there explicitly.  not properly
268	tested on non-ascii character sets by me.
269
270Jun 28, 2002:
271	modified run/format() and tran/getsval() to do a slightly better
272	job on using OFMT for output from print and CONVFMT for other
273	number->string conversions, as promised by posix and done by
274	gawk and mawk.  there are still places where it doesn't work
275	right if CONVFMT is changed; by then the STR attribute of the
276	variable has been irrevocably set.  thanks to arnold robbins for
277	code and examples.
278
279	fixed subtle bug in format that could get core dump.  thanks to
280	Jaromir Dolecek <jdolecek@NetBSD.org> for finding and fixing.
281	minor cleanup in run.c / format() at the same time.
282
283	added some tests for null pointers to debugging printf's, which
284	were never intended for external consumption.  thanks to dave
285	kerns (dkerns@lucent.com) for pointing this out.
286
287	GNU compatibility: an empty regexp matches anything (thanks to
288	dag-erling smorgrav, des@ofug.org).  subject to reversion if
289	this does more harm than good.
290
291	pervasive small changes to make things more const-correct, as
292	reported by gcc's -Wwrite-strings.  as it says in the gcc manual,
293	this may be more nuisance than useful.  provoked by a suggestion
294	and code from arnaud desitter, arnaud@nimbus.geog.ox.ac.uk
295
296	minor documentation changes to note that this now compiles out
297	of the box on Mac OS X.
298
299Feb 10, 2002:
300	changed types in posix chars structure to quiet solaris cc.
301
302Jan 1, 2002:
303	fflush() or fflush("") flushes all files and pipes.
304
305	length(arrayname) returns number of elements; thanks to
306	arnold robbins for suggestion.
307
308	added a makefile.win to make it easier to build on windows.
309	based on dan allen's buildwin.bat.
310
311Nov 16, 2001:
312	added support for posix character class names like [:digit:],
313	which are not exactly shorter than [0-9] and perhaps no more
314	portable.  thanks to dag-erling smorgrav for code.
315
316Feb 16, 2001:
317	removed -m option; no longer needed, and it was actually
318	broken (noted thanks to volker kiefel).
319
320Feb 10, 2001:
321	fixed an appalling bug in gettok: any sequence of digits, +,-, E, e,
322	and period was accepted as a valid number if it started with a period.
323	this would never have happened with the lex version.
324
325	other 1-character botches, now fixed, include a bare $ and a
326	bare " at the end of the input.
327
328Feb 7, 2001:
329	more (const char *) casts in b.c and tran.c to silence warnings.
330
331Nov 15, 2000:
332	fixed a bug introduced in august 1997 that caused expressions
333	like $f[1] to be syntax errors.  thanks to arnold robbins for
334	noticing this and providing a fix.
335
336Oct 30, 2000:
337	fixed some nextfile bugs: not handling all cases.  thanks to
338	arnold robbins for pointing this out.  new regressions added.
339
340	close() is now a function.  it returns whatever the library
341	fclose returns, and -1 for closing a file or pipe that wasn't
342	opened.
343
344Sep 24, 2000:
345	permit \n explicitly in character classes; won't work right
346	if comes in as "[\n]" but ok as /[\n]/, because of multiple
347	processing of \'s.  thanks to arnold robbins.
348
349July 5, 2000:
350	minor fiddles in tran.c to keep compilers happy about uschar.
351	thanks to norman wilson.
352
353May 25, 2000:
354	yet another attempt at making 8-bit input work, with another
355	band-aid in b.c (member()), and some (uschar) casts to head
356	off potential errors in subscripts (like isdigit).  also
357	changed HAT to NCHARS-2.  thanks again to santiago vila.
358
359	changed maketab.c to ignore apparently out of range definitions
360	instead of halting; new freeBSD generates one.  thanks to
361	jon snader <jsnader@ix.netcom.com> for pointing out the problem.
362
363May 2, 2000:
364	fixed an 8-bit problem in b.c by making several char*'s into
365	unsigned char*'s.  not clear i have them all yet.  thanks to
366	Santiago Vila <sanvila@unex.es> for the bug report.
367
368Apr 21, 2000:
369	finally found and fixed a memory leak in function call; it's
370	been there since functions were added ~1983.  thanks to
371	jon bentley for the test case that found it.
372
373	added test in envinit to catch environment "variables" with
374	names beginning with '='; thanks to Berend Hasselman.
375
376Jul 28, 1999:
377	added test in defn() to catch function foo(foo), which
378	otherwise recurses until core dump.  thanks to arnold
379	robbins for noticing this.
380
381Jun 20, 1999:
382	added *bp in gettok in lex.c; appears possible to exit function
383	without terminating the string.  thanks to russ cox.
384
385Jun 2, 1999:
386	added function stdinit() to run to initialize files[] array,
387	in case stdin, etc., are not constants; some compilers care.
388
389May 10, 1999:
390	replaced the ERROR ... FATAL, etc., macros with functions
391	based on vprintf, to avoid problems caused by overrunning
392	fixed-size errbuf array.  thanks to ralph corderoy for the
393	impetus, and for pointing out a string termination bug in
394	qstring as well.
395
396Apr 21, 1999:
397	fixed bug that caused occasional core dumps with commandline
398	variable with value ending in \.  (thanks to nelson beebe for
399	the test case.)
400
401Apr 16, 1999:
402	with code kindly provided by Bruce Lilly, awk now parses
403	/=/ and similar constructs more sensibly in more places.
404	Bruce also provided some helpful test cases.
405
406Apr 5, 1999:
407	changed true/false to True/False in run.c to make it
408	easier to compile with C++.  Added some casts on malloc
409	and realloc to be honest about casts; ditto.  changed
410	ltype int to long in struct rrow to reduce some 64-bit
411	complaints; other changes scattered throughout for the
412	same purpose.  thanks to Nelson Beebe for these portability
413	improvements.
414
415	removed some horrible pointer-int casting in b.c and elsewhere
416	by adding ptoi and itonp to localize the casts, which are
417	all benign.  fixed one incipient bug that showed up on sgi
418	in 64-bit mode.
419
420	reset lineno for new source file; include filename in error
421	message.  also fixed line number error in continuation lines.
422	(thanks to Nelson Beebe for both of these.)
423
424Mar 24, 1999:
425	Nelson Beebe notes that irix 5.3 yacc dies with a bogus
426	error; use a newer version or switch to bison, since sgi
427	is unlikely to fix it.
428
429Mar 5, 1999:
430	changed isnumber to is_number to avoid the problem caused by
431	versions of ctype.h that include the name isnumber.
432
433	distribution now includes a script for building on a Mac,
434	thanks to Dan Allen.
435
436Feb 20, 1999:
437	fixed memory leaks in run.c (call) and tran.c (setfval).
438	thanks to Stephen Nutt for finding these and providing the fixes.
439
440Jan 13, 1999:
441	replaced srand argument by (unsigned int) in run.c;
442	avoids problem on Mac and potentially on Unix & Windows.
443	thanks to Dan Allen.
444
445	added a few (int) casts to silence useless compiler warnings.
446	e.g., errorflag= in run.c jump().
447
448	added proctab.c to the bundle outout; one less thing
449	to have to compile out of the box.
450
451	added calls to _popen and _pclose to the win95 stub for
452	pipes (thanks to Steve Adams for this helpful suggestion).
453	seems to work, though properties are not well understood
454	by me, and it appears that under some circumstances the
455	pipe output is truncated.  Be careful.
456
457Oct 19, 1998:
458	fixed a couple of bugs in getrec: could fail to update $0
459	after a getline var; because inputFS wasn't initialized,
460	could split $0 on every character, a misleading diversion.
461
462	fixed caching bug in makedfa: LRU was actually removing
463	least often used.
464
465	thanks to ross ridge for finding these, and for providing
466	great bug reports.
467
468May 12, 1998:
469	fixed potential bug in readrec: might fail to update record
470	pointer after growing.  thanks to dan levy for spotting this
471	and suggesting the fix.
472
473Mar 12, 1998:
474	added -V to print version number and die.
475
476Feb 11, 1998:
477	subtle silent bug in lex.c: if the program ended with a number
478	longer than 1 digit, part of the input would be pushed back and
479	parsed again because token buffer wasn't terminated right.
480	example:  awk 'length($0) > 10'.  blush.  at least i found it
481	myself.
482
483Aug 31, 1997:
484	s/adelete/awkdelete/: SGI uses this in malloc.h.
485	thanks to nelson beebe for pointing this one out.
486
487Aug 21, 1997:
488	fixed some bugs in sub and gsub when replacement includes \\.
489	this is a dark, horrible corner, but at least now i believe that
490	the behavior is the same as gawk and the intended posix standard.
491	thanks to arnold robbins for advice here.
492
493Aug 9, 1997:
494	somewhat regretfully, replaced the ancient lex-based lexical
495	analyzer with one written in C.  it's longer, generates less code,
496	and more portable; the old one depended too much on mysterious
497	properties of lex that were not preserved in other environments.
498	in theory these recognize the same language.
499
500	now using strtod to test whether a string is a number, instead of
501	the convoluted original function.  should be more portable and
502	reliable if strtod is implemented right.
503
504	removed now-pointless optimization in makefile that tries to avoid
505	recompilation when awkgram.y is changed but symbols are not.
506
507	removed most fixed-size arrays, though a handful remain, some
508	of which are unchecked.  you have been warned.
509
510Aug 4, 1997:
511	with some trepidation, replaced the ancient code that managed
512	fields and $0 in fixed-size arrays with arrays that grow on
513	demand.  there is still some tension between trying to make this
514	run fast and making it clean; not sure it's right yet.
515
516	the ill-conceived -mr and -mf arguments are now useful only
517	for debugging.  previous dynamic string code removed.
518
519	numerous other minor cleanups along the way.
520
521Jul 30, 1997:
522	using code provided by dan levy (to whom profuse thanks), replaced
523	fixed-size arrays and awkward kludges by a fairly uniform mechanism
524	to grow arrays as needed for printf, sub, gsub, etc.
525
526Jul 23, 1997:
527	falling off the end of a function returns "" and 0, not 0.
528	thanks to arnold robbins.
529
530Jun 17, 1997:
531	replaced several fixed-size arrays by dynamically-created ones
532	in run.c; added overflow tests to some previously unchecked cases.
533	getline, toupper, tolower.
534
535	getline code is still broken in that recursive calls may wind
536	up using the same space.  [fixed later]
537
538	increased RECSIZE to 8192 to push problems further over the horizon.
539
540	added \r to \n as input line separator for programs, not data.
541	damn CRLFs.
542
543	modified format() to permit explicit printf("%c", 0) to include
544	a null byte in output.  thanks to ken stailey for the fix.
545
546	added a "-safe" argument that disables file output (print >,
547	print >>), process creation (cmd|getline, print |, system), and
548	access to the environment (ENVIRON).  this is a first approximation
549	to a "safe" version of awk, but don't rely on it too much.  thanks
550	to joan feigenbaum and matt blaze for the inspiration long ago.
551
552Jul 8, 1996:
553	fixed long-standing bug in sub, gsub(/a/, "\\\\&"); thanks to
554	ralph corderoy.
555
556Jun 29, 1996:
557	fixed awful bug in new field splitting; didn't get all the places
558	where input was done.
559
560Jun 28, 1996:
561	changed field-splitting to conform to posix definition: fields are
562	split using the value of FS at the time of input; it used to be
563	the value when the field or NF was first referred to, a much less
564	predictable definition.  thanks to arnold robbins for encouragement
565	to do the right thing.
566
567May 28, 1996:
568	fixed appalling but apparently unimportant bug in parsing octal
569	numbers in reg exprs.
570
571	explicit hex in reg exprs now limited to 2 chars: \xa, \xaa.
572
573May 27, 1996:
574	cleaned up some declarations so gcc -Wall is now almost silent.
575
576	makefile now includes backup copies of ytab.c and lexyy.c in case
577	one makes before looking; it also avoids recreating lexyy.c unless
578	really needed.
579
580	s/aprintf/awkprint, s/asprintf/awksprintf/ to avoid some name clashes
581	with unwisely-written header files.
582
583	thanks to jeffrey friedl for several of these.
584
585May 26, 1996:
586	an attempt to rationalize the (unsigned) char issue.  almost all
587	instances of unsigned char have been removed; the handful of places
588	in b.c where chars are used as table indices have been hand-crafted.
589	added some latin-1 tests to the regression, but i'm not confident;
590	none of my compilers seem to care much.  thanks to nelson beebe for
591	pointing out some others that do care.
592
593May 2, 1996:
594	removed all register declarations.
595
596	enhanced split(), as in gawk, etc:  split(s, a, "") splits s into
597	a[1]...a[length(s)] with each character a single element.
598
599	made the same changes for field-splitting if FS is "".
600
601	added nextfile, as in gawk: causes immediate advance to next
602	input file. (thanks to arnold robbins for inspiration and code).
603
604	small fixes to regexpr code:  can now handle []], [[], and
605	variants;  [] is now a syntax error, rather than matching
606	everything;  [z-a] is now empty, not z.  far from complete
607	or correct, however.  (thanks to jeffrey friedl for pointing out
608	some awful behaviors.)
609
610Apr 29, 1996:
611	replaced uchar by uschar everywhere; apparently some compilers
612	usurp this name and this causes conflicts.
613
614	fixed call to time in run.c (bltin); arg is time_t *.
615
616	replaced horrible pointer/long punning in b.c by a legitimate
617	union.  should be safer on 64-bit machines and cleaner everywhere.
618	(thanks to nelson beebe for pointing out some of these problems.)
619
620	replaced nested comments by #if 0...#endif in run.c, lib.c.
621
622	removed getsval, setsval, execute macros from run.c and lib.c.
623	machines are 100x faster than they were when these macros were
624	first used.
625
626	revised filenames: awk.g.y => awkgram.y, awk.lx.l => awklex.l,
627	y.tab.[ch] => ytab.[ch], lex.yy.c => lexyy.c, all in the aid of
628	portability to nameless systems.
629
630	"make bundle" now includes yacc and lex output files for recipients
631	who don't have yacc or lex.
632
633Aug 15, 1995:
634	initialized Cells in setsymtab more carefully; some fields
635	were not set.  (thanks to purify, all of whose complaints i
636	think i now understand.)
637
638	fixed at least one error in gsub that looked at -1-th element
639	of an array when substituting for a null match (e.g., $).
640
641	delete arrayname is now legal; it clears the elements but leaves
642	the array, which may not be the right behavior.
643
644	modified makefile: my current make can't cope with the test used
645	to avoid unnecessary yacc invocations.
646
647Jul 17, 1995:
648	added dynamically growing strings to awk.lx.l and b.c
649	to permit regular expressions to be much bigger.
650	the state arrays can still overflow.
651
652Aug 24, 1994:
653	detect duplicate arguments in function definitions (mdm).
654
655May 11, 1994:
656	trivial fix to printf to limit string size in sub().
657
658Apr 22, 1994:
659	fixed yet another subtle self-assignment problem:
660	$1 = $2; $1 = $1 clobbered $1.
661
662	Regression tests now use private echo, to avoid quoting problems.
663
664Feb 2, 1994:
665	changed error() to print line number as %d, not %g.
666
667Jul 23, 1993:
668	cosmetic changes: increased sizes of some arrays,
669	reworded some error messages.
670
671	added CONVFMT as in posix (just replaced OFMT in getsval)
672
673	FILENAME is now "" until the first thing that causes a file
674	to be opened.
675
676Nov 28, 1992:
677	deleted yyunput and yyoutput from proto.h;
678	different versions of lex give these different declarations.
679
680May 31, 1992:
681	added -mr N and -mf N options: more record and fields.
682	these really ought to adjust automatically.
683
684	cleaned up some error messages; "out of space" now means
685	malloc returned NULL in all cases.
686
687	changed rehash so that if it runs out, it just returns;
688	things will continue to run slow, but maybe a bit longer.
689
690Apr 24, 1992:
691	remove redundant close of stdin when using -f -.
692
693	got rid of core dump with -d; awk -d just prints date.
694
695Apr 12, 1992:
696	added explicit check for /dev/std(in,out,err) in redirection.
697	unlike gawk, no /dev/fd/n yet.
698
699	added (file/pipe) builtin.  hard to test satisfactorily.
700	not posix.
701
702Feb 20, 1992:
703	recompile after abortive changes;  should be unchanged.
704
705Dec 2, 1991:
706	die-casting time:  converted to ansi C, installed that.
707
708Nov 30, 1991:
709	fixed storage leak in freefa, failing to recover [N]CCL.
710	thanks to Bill Jones (jones@cs.usask.ca)
711
712Nov 19, 1991:
713	use RAND_MAX instead of literal in builtin().
714
715Nov 12, 1991:
716	cranked up some fixed-size arrays in b.c, and added a test for
717	overflow in penter.  thanks to mark larsen.
718
719Sep 24, 1991:
720	increased buffer in gsub.  a very crude fix to a general problem.
721	and again on Sep 26.
722
723Aug 18, 1991:
724	enforce variable name syntax for commandline variables: has to
725	start with letter or _.
726
727Jul 27, 1991:
728	allow newline after ; in for statements.
729
730Jul 21, 1991:
731	fixed so that in self-assignment like $1=$1, side effects
732	like recomputing $0 take place.  (this is getting subtle.)
733
734Jun 30, 1991:
735	better test for detecting too-long output record.
736
737Jun 2, 1991:
738	better defense against very long printf strings.
739	made break and continue illegal outside of loops.
740
741May 13, 1991:
742	removed extra arg on gettemp, tempfree.  minor error message rewording.
743
744May 6, 1991:
745	fixed silly bug in hex parsing in hexstr().
746	removed an apparently unnecessary test in isnumber().
747	warn about weird printf conversions.
748	fixed unchecked array overwrite in relex().
749
750	changed for (i in array) to access elements in sorted order.
751	then unchanged it -- it really does run slower in too many cases.
752	left the code in place, commented out.
753
754Feb 10, 1991:
755	check error status on all writes, to avoid banging on full disks.
756
757Jan 28, 1991:
758	awk -f - reads the program from stdin.
759
760Jan 11, 1991:
761	failed to set numeric state on $0 in cmd|getline context in run.c.
762
763Nov 2, 1990:
764	fixed sleazy test for integrality in getsval;  use modf.
765
766Oct 29, 1990:
767	fixed sleazy buggy code in lib.c that looked (incorrectly) for
768	too long input lines.
769
770Oct 14, 1990:
771	fixed the bug on p. 198 in which it couldn't deduce that an
772	argument was an array in some contexts.  replaced the error
773	message in intest() by code that damn well makes it an array.
774
775Oct 8, 1990:
776	fixed horrible bug:  types and values were not preserved in
777	some kinds of self-assignment. (in assign().)
778
779Aug 24, 1990:
780	changed NCHARS to 256 to handle 8-bit characters in strings
781	presented to match(), etc.
782
783Jun 26, 1990:
784	changed struct rrow (awk.h) to use long instead of int for lval,
785	since cfoll() stores a pointer in it.  now works better when int's
786	are smaller than pointers!
787
788May 6, 1990:
789	AVA fixed the grammar so that ! is uniformly of the same precedence as
790	unary + and -.  This renders illegal some constructs like !x=y, which
791	now has to be parenthesized as !(x=y), and makes others work properly:
792	!x+y is (!x)+y, and x!y is x !y, not two pattern-action statements.
793	(These problems were pointed out by Bob Lenk of Posix.)
794
795	Added \x to regular expressions (already in strings).
796	Limited octal to octal digits; \8 and \9 are not octal.
797	Centralized the code for parsing escapes in regular expressions.
798	Added a bunch of tests to T.re and T.sub to verify some of this.
799
800Feb 9, 1990:
801	fixed null pointer dereference bug in main.c:  -F[nothing].  sigh.
802
803	restored srand behavior:  it returns the current seed.
804
805Jan 18, 1990:
806	srand now returns previous seed value (0 to start).
807
808Jan 5, 1990:
809	fix potential problem in tran.c -- something was freed,
810	then used in freesymtab.
811
812Oct 18, 1989:
813	another try to get the max number of open files set with
814	relatively machine-independent code.
815
816	small fix to input() in case of multiple reads after EOF.
817
818Oct 11, 1989:
819	FILENAME is now defined in the BEGIN block -- too many old
820	programs broke.
821
822	"-" means stdin in getline as well as on the commandline.
823
824	added a bunch of casts to the code to tell the truth about
825	char * vs. unsigned char *, a right royal pain.  added a
826	setlocale call to the front of main, though probably no one
827	has it usefully implemented yet.
828
829Aug 24, 1989:
830	removed redundant relational tests against nullnode if parse
831	tree already had a relational at that point.
832
833Aug 11, 1989:
834	fixed bug:  commandline variable assignment has to look like
835	var=something.  (consider the man page for =, in file =.1)
836
837	changed number of arguments to functions to static arrays
838	to avoid repeated malloc calls.
839
840Aug 2, 1989:
841	restored -F (space) separator
842
843Jul 30, 1989:
844	added -v x=1 y=2 ... for immediate commandline variable assignment;
845	done before the BEGIN block for sure.  they have to precede the
846	program if the program is on the commandline.
847	Modified Aug 2 to require a separate -v for each assignment.
848
849Jul 10, 1989:
850	fixed ref-thru-zero bug in environment code in tran.c
851
852Jun 23, 1989:
853	add newline to usage message.
854
855Jun 14, 1989:
856	added some missing ansi printf conversion letters: %i %X %E %G.
857	no sensible meaning for h or L, so they may not do what one expects.
858
859	made %* conversions work.
860
861	changed x^y so that if n is a positive integer, it's done
862	by explicit multiplication, thus achieving maximum accuracy.
863	(this should be done by pow() but it seems not to be locally.)
864	done to x ^= y as well.
865
866Jun 4, 1989:
867	ENVIRON array contains environment: if shell variable V=thing,
868		ENVIRON["V"] is "thing"
869
870	multiple -f arguments permitted.  error reporting is naive.
871	(they were permitted before, but only the last was used.)
872
873	fixed a really stupid botch in the debugging macro dprintf
874
875	fixed order of evaluation of commandline assignments to match
876	what the book claims:  an argument of the form x=e is evaluated
877	at the time it would have been opened if it were a filename (p 63).
878	this invalidates the suggested answer to ex 4-1 (p 195).
879
880	removed some code that permitted -F (space) fieldseparator,
881	since it didn't quite work right anyway.  (restored aug 2)
882
883Apr 27, 1989:
884	Line number now accumulated correctly for comment lines.
885
886Apr 26, 1989:
887	Debugging output now includes a version date,
888	if one compiles it into the source each time.
889
890Apr 9, 1989:
891	Changed grammar to prohibit constants as 3rd arg of sub and gsub;
892	prevents class of overwriting-a-constant errors.  (Last one?)
893	This invalidates the "banana" example on page 43 of the book.
894
895	Added \a ("alert"), \v (vertical tab), \xhhh (hexadecimal),
896	as in ANSI, for strings.  Rescinded the sloppiness that permitted
897	non-octal digits in \ooo.  Warning:  not all compilers and libraries
898	will be able to deal with \x correctly.
899
900Jan 9, 1989:
901	Fixed bug that caused tempcell list to contain a duplicate.
902	The fix is kludgy.
903
904Dec 17, 1988:
905	Catches some more commandline errors in main.
906	Removed redundant decl of modf in run.c (confuses some compilers).
907	Warning:  there's no single declaration of malloc, etc., in awk.h
908	that seems to satisfy all compilers.
909
910Dec 7, 1988:
911	Added a bit of code to error printing to avoid printing nulls.
912	(Not clear that it actually would.)
913
914Nov 27, 1988:
915	With fear and trembling, modified the grammar to permit
916	multiple pattern-action statements on one line without
917	an explicit separator.  By definition, this capitulation
918	to the ghost of ancient implementations remains undefined
919	and thus subject to change without notice or apology.
920	DO NOT COUNT ON IT.
921
922Oct 30, 1988:
923	Fixed bug in call() that failed to recover storage.
924
925	A warning is now generated if there are more arguments
926	in the call than in the definition (in lieu of fixing
927	another storage leak).
928
929Oct 20, 1988:
930	Fixed %c:  if expr is numeric, use numeric value;
931	otherwise print 1st char of string value.  still
932	doesn't work if the value is 0 -- won't print \0.
933
934	Added a few more checks for running out of malloc.
935
936Oct 12, 1988:
937	Fixed bug in call() that freed local arrays twice.
938
939	Fixed to handle deletion of non-existent array right;
940	complains about attempt to delete non-array element.
941
942Sep 30, 1988:
943	Now guarantees to evaluate all arguments of built-in
944	functions, as in C;  the appearance is that arguments
945	are evaluated before the function is called.  Places
946	affected are sub (gsub was ok), substr, printf, and
947	all the built-in arithmetic functions in bltin().
948	A warning is generated if a bltin() is called with
949	the wrong number of arguments.
950
951	This requires changing makeprof on p167 of the book.
952
953Aug 23, 1988:
954	setting FILENAME in BEGIN caused core dump, apparently
955	because it was freeing space not allocated by malloc.
956
957July 24, 1988:
958	fixed egregious error in toupper/tolower functions.
959	still subject to rescinding, however.
960
961July 2, 1988:
962	flush stdout before opening file or pipe
963
964July 2, 1988:
965	performance bug in b.c/cgoto(): not freeing some sets of states.
966	partial fix only right now, and the number of states increased
967	to make it less obvious.
968
969June 1, 1988:
970	check error status on close
971
972May 28, 1988:
973	srand returns seed value it's using.
974	see 1/18/90
975
976May 22, 1988:
977	Removed limit on depth of function calls.
978
979May 10, 1988:
980	Fixed lib.c to permit _ in commandline variable names.
981
982Mar 25, 1988:
983	main.c fixed to recognize -- as terminator of command-
984	line options.  Illegal options flagged.
985	Error reporting slightly cleaned up.
986
987Dec 2, 1987:
988	Newer C compilers apply a strict scope rule to extern
989	declarations within functions.  Two extern declarations in
990	lib.c and tran.c have been moved to obviate this problem.
991
992Oct xx, 1987:
993	Reluctantly added toupper and tolower functions.
994	Subject to rescinding without notice.
995
996Sep 17, 1987:
997	Error-message printer had printf(s) instead of
998	printf("%s",s);  got core dumps when the message
999	included a %.
1000
1001Sep 12, 1987:
1002	Very long printf strings caused core dump;
1003	fixed aprintf, asprintf, format to catch them.
1004	Can still get a core dump in printf itself.
1005
1006
1007