/[pcre]/code/trunk/ChangeLog
ViewVC logotype

Diff of /code/trunk/ChangeLog

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 45 by nigel, Sat Feb 24 21:39:25 2007 UTC revision 1433 by ph10, Fri Jan 3 15:15:00 2014 UTC
# Line 1  Line 1 
1  ChangeLog for PCRE  ChangeLog for PCRE
2  ------------------  ------------------
3    
4    Version 8.35-RC1 xx-xxxx-201x
5    -----------------------------
6    
7    1.  A new flag is set, when property checks are present in an XCLASS.
8        When this flag is not set, PCRE can perform certain optimizations
9        such as studying these XCLASS-es.
10    
11    2.  The auto-possessification of character sets were improved: a normal
12        and an extended character set can be compared now. Furthermore
13        the JIT compiler optimizes more character set checks.
14    
15    3.  Got rid of some compiler warnings for potentially uninitialized variables
16        that show up only when compiled with -O2.
17    
18    4.  A pattern such as (?=ab\K) that uses \K in an assertion can set the start
19        of a match later then the end of the match. The pcretest program was not
20        handling the case sensibly - it was outputting from the start to the next
21        binary zero. It now reports this situation in a message, and outputs the
22        text from the end to the start.
23    
24    5.  Fast forward search is improved in JIT. Instead of the first three
25        characters, any three characters with fixed position can be searched.
26        Search order: first, last, middle.
27    
28    6.  Improve character range checks in JIT. Characters are read by an inprecise
29        function now, which returns with an unknown value if the character code is
30        above a certain treshold (e.g: 256). The only limitation is that the value
31        must be bigger than the treshold as well. This function is useful, when
32        the characters above the treshold are handled in the same way.
33    
34    7.  The macros whose names start with RAWUCHAR are placeholders for a future
35        mode in which only the bottom 21 bits of 32-bit data items are used. To
36        make this more memorable for those maintaining the code, the names have
37        been changed to start with UCHAR21, and an extensive comment has been added
38        to their definition.
39    
40    8.  Add missing (new) files sljitNativeTILEGX.c and sljitNativeTILEGX-encoder.c
41        to the export list in Makefile.am (they were accidentally omitted from the
42        8.34 tarball).
43    
44    9.  The informational output from pcretest used the phrase "starting byte set"
45        which is inappropriate for the 16-bit and 32-bit libraries. As the output
46        for "first char" and "need char" really means "non-UTF-char", I've changed
47        "byte" to "char", and slightly reworded the output. The documentation about
48        these values has also been (I hope) clarified.
49    
50    
51    Version 8.34 15-December-2013
52    -----------------------------
53    
54    1.  Add pcre[16|32]_jit_free_unused_memory to forcibly free unused JIT
55        executable memory. Patch inspired by Carsten Klein.
56    
57    2.  ./configure --enable-coverage defined SUPPORT_GCOV in config.h, although
58        this macro is never tested and has no effect, because the work to support
59        coverage involves only compiling and linking options and special targets in
60        the Makefile. The comment in config.h implied that defining the macro would
61        enable coverage support, which is totally false. There was also support for
62        setting this macro in the CMake files (my fault, I just copied it from
63        configure). SUPPORT_GCOV has now been removed.
64    
65    3.  Make a small performance improvement in strlen16() and strlen32() in
66        pcretest.
67    
68    4.  Change 36 for 8.33 left some unreachable statements in pcre_exec.c,
69        detected by the Solaris compiler (gcc doesn't seem to be able to diagnose
70        these cases). There was also one in pcretest.c.
71    
72    5.  Cleaned up a "may be uninitialized" compiler warning in pcre_exec.c.
73    
74    6.  In UTF mode, the code for checking whether a group could match an empty
75        string (which is used for indefinitely repeated groups to allow for
76        breaking an infinite loop) was broken when the group contained a repeated
77        negated single-character class with a character that occupied more than one
78        data item and had a minimum repetition of zero (for example, [^\x{100}]* in
79        UTF-8 mode). The effect was undefined: the group might or might not be
80        deemed as matching an empty string, or the program might have crashed.
81    
82    7.  The code for checking whether a group could match an empty string was not
83        recognizing that \h, \H, \v, \V, and \R must match a character.
84    
85    8.  Implemented PCRE_INFO_MATCH_EMPTY, which yields 1 if the pattern can match
86        an empty string. If it can, pcretest shows this in its information output.
87    
88    9.  Fixed two related bugs that applied to Unicode extended grapheme clusters
89        that were repeated with a maximizing qualifier (e.g. \X* or \X{2,5}) when
90        matched by pcre_exec() without using JIT:
91    
92        (a) If the rest of the pattern did not match after a maximal run of
93            grapheme clusters, the code for backing up to try with fewer of them
94            did not always back up over a full grapheme when characters that do not
95            have the modifier quality were involved, e.g. Hangul syllables.
96    
97        (b) If the match point in a subject started with modifier character, and
98            there was no match, the code could incorrectly back up beyond the match
99            point, and potentially beyond the first character in the subject,
100            leading to a segfault or an incorrect match result.
101    
102    10. A conditional group with an assertion condition could lead to PCRE
103        recording an incorrect first data item for a match if no other first data
104        item was recorded. For example, the pattern (?(?=ab)ab) recorded "a" as a
105        first data item, and therefore matched "ca" after "c" instead of at the
106        start.
107    
108    11. Change 40 for 8.33 (allowing pcregrep to find empty strings) showed up a
109        bug that caused the command "echo a | ./pcregrep -M '|a'" to loop.
110    
111    12. The source of pcregrep now includes z/OS-specific code so that it can be
112        compiled for z/OS as part of the special z/OS distribution.
113    
114    13. Added the -T and -TM options to pcretest.
115    
116    14. The code in pcre_compile.c for creating the table of named capturing groups
117        has been refactored. Instead of creating the table dynamically during the
118        actual compiling pass, the information is remembered during the pre-compile
119        pass (on the stack unless there are more than 20 named groups, in which
120        case malloc() is used) and the whole table is created before the actual
121        compile happens. This has simplified the code (it is now nearly 150 lines
122        shorter) and prepared the way for better handling of references to groups
123        with duplicate names.
124    
125    15. A back reference to a named subpattern when there is more than one of the
126        same name now checks them in the order in which they appear in the pattern.
127        The first one that is set is used for the reference. Previously only the
128        first one was inspected. This change makes PCRE more compatible with Perl.
129    
130    16. Unicode character properties were updated from Unicode 6.3.0.
131    
132    17. The compile-time code for auto-possessification has been refactored, based
133        on a patch by Zoltan Herczeg. It now happens after instead of during
134        compilation. The code is cleaner, and more cases are handled. The option
135        PCRE_NO_AUTO_POSSESS is added for testing purposes, and the -O and /O
136        options in pcretest are provided to set it. It can also be set by
137        (*NO_AUTO_POSSESS) at the start of a pattern.
138    
139    18. The character VT has been added to the default ("C" locale) set of
140        characters that match \s and are generally treated as white space,
141        following this same change in Perl 5.18. There is now no difference between
142        "Perl space" and "POSIX space". Whether VT is treated as white space in
143        other locales depends on the locale.
144    
145    19. The code for checking named groups as conditions, either for being set or
146        for being recursed, has been refactored (this is related to 14 and 15
147        above). Processing unduplicated named groups should now be as fast at
148        numerical groups, and processing duplicated groups should be faster than
149        before.
150    
151    20. Two patches to the CMake build system, by Alexander Barkov:
152    
153          (1) Replace the "source" command by "." in CMakeLists.txt because
154              "source" is a bash-ism.
155    
156          (2) Add missing HAVE_STDINT_H and HAVE_INTTYPES_H to config-cmake.h.in;
157              without these the CMake build does not work on Solaris.
158    
159    21. Perl has changed its handling of \8 and \9. If there is no previously
160        encountered capturing group of those numbers, they are treated as the
161        literal characters 8 and 9 instead of a binary zero followed by the
162        literals. PCRE now does the same.
163    
164    22. Following Perl, added \o{} to specify codepoints in octal, making it
165        possible to specify values greater than 0777 and also making them
166        unambiguous.
167    
168    23. Perl now gives an error for missing closing braces after \x{... instead of
169        treating the string as literal. PCRE now does the same.
170    
171    24. RunTest used to grumble if an inappropriate test was selected explicitly,
172        but just skip it when running all tests. This make it awkward to run ranges
173        of tests when one of them was inappropriate. Now it just skips any
174        inappropriate tests, as it always did when running all tests.
175    
176    25. If PCRE_AUTO_CALLOUT and PCRE_UCP were set for a pattern that contained
177        character types such as \d or \w, too many callouts were inserted, and the
178        data that they returned was rubbish.
179    
180    26. In UCP mode, \s was not matching two of the characters that Perl matches,
181        namely NEL (U+0085) and MONGOLIAN VOWEL SEPARATOR (U+180E), though they
182        were matched by \h. The code has now been refactored so that the lists of
183        the horizontal and vertical whitespace characters used for \h and \v (which
184        are defined only in one place) are now also used for \s.
185    
186    27. Add JIT support for the 64 bit TileGX architecture.
187        Patch by Jiong Wang (Tilera Corporation).
188    
189    28. Possessive quantifiers for classes (both explicit and automatically
190        generated) now use special opcodes instead of wrapping in ONCE brackets.
191    
192    29. Whereas an item such as A{4}+ ignored the possessivenes of the quantifier
193        (because it's meaningless), this was not happening when PCRE_CASELESS was
194        set. Not wrong, but inefficient.
195    
196    30. Updated perltest.pl to add /u (force Unicode mode) when /W (use Unicode
197        properties for \w, \d, etc) is present in a test regex. Otherwise if the
198        test contains no characters greater than 255, Perl doesn't realise it
199        should be using Unicode semantics.
200    
201    31. Upgraded the handling of the POSIX classes [:graph:], [:print:], and
202        [:punct:] when PCRE_UCP is set so as to include the same characters as Perl
203        does in Unicode mode.
204    
205    32. Added the "forbid" facility to pcretest so that putting tests into the
206        wrong test files can sometimes be quickly detected.
207    
208    33. There is now a limit (default 250) on the depth of nesting of parentheses.
209        This limit is imposed to control the amount of system stack used at compile
210        time. It can be changed at build time by --with-parens-nest-limit=xxx or
211        the equivalent in CMake.
212    
213    34. Character classes such as [A-\d] or [a-[:digit:]] now cause compile-time
214        errors. Perl warns for these when in warning mode, but PCRE has no facility
215        for giving warnings.
216    
217    35. Change 34 for 8.13 allowed quantifiers on assertions, because Perl does.
218        However, this was not working for (?!) because it is optimized to (*FAIL),
219        for which PCRE does not allow quantifiers. The optimization is now disabled
220        when a quantifier follows (?!). I can't see any use for this, but it makes
221        things uniform.
222    
223    36. Perl no longer allows group names to start with digits, so I have made this
224        change also in PCRE. It simplifies the code a bit.
225    
226    37. In extended mode, Perl ignores spaces before a + that indicates a
227        possessive quantifier. PCRE allowed a space before the quantifier, but not
228        before the possessive +. It now does.
229    
230    38. The use of \K (reset reported match start) within a repeated possessive
231        group such as (a\Kb)*+ was not working.
232    
233    40. Document that the same character tables must be used at compile time and
234        run time, and that the facility to pass tables to pcre_exec() and
235        pcre_dfa_exec() is for use only with saved/restored patterns.
236    
237    41. Applied Jeff Trawick's patch CMakeLists.txt, which "provides two new
238        features for Builds with MSVC:
239    
240        1. Support pcre.rc and/or pcreposix.rc (as is already done for MinGW
241           builds). The .rc files can be used to set FileDescription and many other
242           attributes.
243    
244        2. Add an option (-DINSTALL_MSVC_PDB) to enable installation of .pdb files.
245           This allows higher-level build scripts which want .pdb files to avoid
246           hard-coding the exact files needed."
247    
248    42. Added support for [[:<:]] and [[:>:]] as used in the BSD POSIX library to
249        mean "start of word" and "end of word", respectively, as a transition aid.
250    
251    43. A minimizing repeat of a class containing codepoints greater than 255 in
252        non-UTF 16-bit or 32-bit modes caused an internal error when PCRE was
253        compiled to use the heap for recursion.
254    
255    44. Got rid of some compiler warnings for unused variables when UTF but not UCP
256        is configured.
257    
258    
259    Version 8.33 28-May-2013
260    ------------------------
261    
262    1.  Added 'U' to some constants that are compared to unsigned integers, to
263        avoid compiler signed/unsigned warnings. Added (int) casts to unsigned
264        variables that are added to signed variables, to ensure the result is
265        signed and can be negated.
266    
267    2.  Applied patch by Daniel Richard G for quashing MSVC warnings to the
268        CMake config files.
269    
270    3.  Revise the creation of config.h.generic so that all boolean macros are
271        #undefined, whereas non-boolean macros are #ifndef/#endif-ed. This makes
272        overriding via -D on the command line possible.
273    
274    4.  Changing the definition of the variable "op" in pcre_exec.c from pcre_uchar
275        to unsigned int is reported to make a quite noticeable speed difference in
276        a specific Windows environment. Testing on Linux did also appear to show
277        some benefit (and it is clearly not harmful). Also fixed the definition of
278        Xop which should be unsigned.
279    
280    5.  Related to (4), changing the definition of the intermediate variable cc
281        in repeated character loops from pcre_uchar to pcre_uint32 also gave speed
282        improvements.
283    
284    6.  Fix forward search in JIT when link size is 3 or greater. Also removed some
285        unnecessary spaces.
286    
287    7.  Adjust autogen.sh and configure.ac to lose warnings given by automake 1.12
288        and later.
289    
290    8.  Fix two buffer over read issues in 16 and 32 bit modes. Affects JIT only.
291    
292    9.  Optimizing fast_forward_start_bits in JIT.
293    
294    10. Adding support for callouts in JIT, and fixing some issues revealed
295        during this work. Namely:
296    
297        (a) Unoptimized capturing brackets incorrectly reset on backtrack.
298    
299        (b) Minimum length was not checked before the matching is started.
300    
301    11. The value of capture_last that is passed to callouts was incorrect in some
302        cases when there was a capture on one path that was subsequently abandoned
303        after a backtrack. Also, the capture_last value is now reset after a
304        recursion, since all captures are also reset in this case.
305    
306    12. The interpreter no longer returns the "too many substrings" error in the
307        case when an overflowing capture is in a branch that is subsequently
308        abandoned after a backtrack.
309    
310    13. In the pathological case when an offset vector of size 2 is used, pcretest
311        now prints out the matched string after a yield of 0 or 1.
312    
313    14. Inlining subpatterns in recursions, when certain conditions are fulfilled.
314        Only supported by the JIT compiler at the moment.
315    
316    15. JIT compiler now supports 32 bit Macs thanks to Lawrence Velazquez.
317    
318    16. Partial matches now set offsets[2] to the "bumpalong" value, that is, the
319        offset of the starting point of the matching process, provided the offsets
320        vector is large enough.
321    
322    17. The \A escape now records a lookbehind value of 1, though its execution
323        does not actually inspect the previous character. This is to ensure that,
324        in partial multi-segment matching, at least one character from the old
325        segment is retained when a new segment is processed. Otherwise, if there
326        are no lookbehinds in the pattern, \A might match incorrectly at the start
327        of a new segment.
328    
329    18. Added some #ifdef __VMS code into pcretest.c to help VMS implementations.
330    
331    19. Redefined some pcre_uchar variables in pcre_exec.c as pcre_uint32; this
332        gives some modest performance improvement in 8-bit mode.
333    
334    20. Added the PCRE-specific property \p{Xuc} for matching characters that can
335        be expressed in certain programming languages using Universal Character
336        Names.
337    
338    21. Unicode validation has been updated in the light of Unicode Corrigendum #9,
339        which points out that "non characters" are not "characters that may not
340        appear in Unicode strings" but rather "characters that are reserved for
341        internal use and have only local meaning".
342    
343    22. When a pattern was compiled with automatic callouts (PCRE_AUTO_CALLOUT) and
344        there was a conditional group that depended on an assertion, if the
345        assertion was false, the callout that immediately followed the alternation
346        in the condition was skipped when pcre_exec() was used for matching.
347    
348    23. Allow an explicit callout to be inserted before an assertion that is the
349        condition for a conditional group, for compatibility with automatic
350        callouts, which always insert a callout at this point.
351    
352    24. In 8.31, (*COMMIT) was confined to within a recursive subpattern. Perl also
353        confines (*SKIP) and (*PRUNE) in the same way, and this has now been done.
354    
355    25. (*PRUNE) is now supported by the JIT compiler.
356    
357    26. Fix infinite loop when /(?<=(*SKIP)ac)a/ is matched against aa.
358    
359    27. Fix the case where there are two or more SKIPs with arguments that may be
360        ignored.
361    
362    28. (*SKIP) is now supported by the JIT compiler.
363    
364    29. (*THEN) is now supported by the JIT compiler.
365    
366    30. Update RunTest with additional test selector options.
367    
368    31. The way PCRE handles backtracking verbs has been changed in two ways.
369    
370        (1) Previously, in something like (*COMMIT)(*SKIP), COMMIT would override
371        SKIP. Now, PCRE acts on whichever backtracking verb is reached first by
372        backtracking. In some cases this makes it more Perl-compatible, but Perl's
373        rather obscure rules do not always do the same thing.
374    
375        (2) Previously, backtracking verbs were confined within assertions. This is
376        no longer the case for positive assertions, except for (*ACCEPT). Again,
377        this sometimes improves Perl compatibility, and sometimes does not.
378    
379    32. A number of tests that were in test 2 because Perl did things differently
380        have been moved to test 1, because either Perl or PCRE has changed, and
381        these tests are now compatible.
382    
383    32. Backtracking control verbs are now handled in the same way in JIT and
384        interpreter.
385    
386    33. An opening parenthesis in a MARK/PRUNE/SKIP/THEN name in a pattern that
387        contained a forward subroutine reference caused a compile error.
388    
389    34. Auto-detect and optimize limited repetitions in JIT.
390    
391    35. Implement PCRE_NEVER_UTF to lock out the use of UTF, in particular,
392        blocking (*UTF) etc.
393    
394    36. In the interpreter, maximizing pattern repetitions for characters and
395        character types now use tail recursion, which reduces stack usage.
396    
397    37. The value of the max lookbehind was not correctly preserved if a compiled
398        and saved regex was reloaded on a host of different endianness.
399    
400    38. Implemented (*LIMIT_MATCH) and (*LIMIT_RECURSION). As part of the extension
401        of the compiled pattern block, expand the flags field from 16 to 32 bits
402        because it was almost full.
403    
404    39. Try madvise first before posix_madvise.
405    
406    40. Change 7 for PCRE 7.9 made it impossible for pcregrep to find empty lines
407        with a pattern such as ^$. It has taken 4 years for anybody to notice! The
408        original change locked out all matches of empty strings. This has been
409        changed so that one match of an empty string per line is recognized.
410        Subsequent searches on the same line (for colouring or for --only-matching,
411        for example) do not recognize empty strings.
412    
413    41. Applied a user patch to fix a number of spelling mistakes in comments.
414    
415    42. Data lines longer than 65536 caused pcretest to crash.
416    
417    43. Clarified the data type for length and startoffset arguments for pcre_exec
418        and pcre_dfa_exec in the function-specific man pages, where they were
419        explicitly stated to be in bytes, never having been updated. I also added
420        some clarification to the pcreapi man page.
421    
422    44. A call to pcre_dfa_exec() with an output vector size less than 2 caused
423        a segmentation fault.
424    
425    
426    Version 8.32 30-November-2012
427    -----------------------------
428    
429    1.  Improved JIT compiler optimizations for first character search and single
430        character iterators.
431    
432    2.  Supporting IBM XL C compilers for PPC architectures in the JIT compiler.
433        Patch by Daniel Richard G.
434    
435    3.  Single character iterator optimizations in the JIT compiler.
436    
437    4.  Improved JIT compiler optimizations for character ranges.
438    
439    5.  Rename the "leave" variable names to "quit" to improve WinCE compatibility.
440        Reported by Giuseppe D'Angelo.
441    
442    6.  The PCRE_STARTLINE bit, indicating that a match can occur only at the start
443        of a line, was being set incorrectly in cases where .* appeared inside
444        atomic brackets at the start of a pattern, or where there was a subsequent
445        *PRUNE or *SKIP.
446    
447    7.  Improved instruction cache flush for POWER/PowerPC.
448        Patch by Daniel Richard G.
449    
450    8.  Fixed a number of issues in pcregrep, making it more compatible with GNU
451        grep:
452    
453        (a) There is now no limit to the number of patterns to be matched.
454    
455        (b) An error is given if a pattern is too long.
456    
457        (c) Multiple uses of --exclude, --exclude-dir, --include, and --include-dir
458            are now supported.
459    
460        (d) --exclude-from and --include-from (multiple use) have been added.
461    
462        (e) Exclusions and inclusions now apply to all files and directories, not
463            just to those obtained from scanning a directory recursively.
464    
465        (f) Multiple uses of -f and --file-list are now supported.
466    
467        (g) In a Windows environment, the default for -d has been changed from
468            "read" (the GNU grep default) to "skip", because otherwise the presence
469            of a directory in the file list provokes an error.
470    
471        (h) The documentation has been revised and clarified in places.
472    
473    9.  Improve the matching speed of capturing brackets.
474    
475    10. Changed the meaning of \X so that it now matches a Unicode extended
476        grapheme cluster.
477    
478    11. Patch by Daniel Richard G to the autoconf files to add a macro for sorting
479        out POSIX threads when JIT support is configured.
480    
481    12. Added support for PCRE_STUDY_EXTRA_NEEDED.
482    
483    13. In the POSIX wrapper regcomp() function, setting re_nsub field in the preg
484        structure could go wrong in environments where size_t is not the same size
485        as int.
486    
487    14. Applied user-supplied patch to pcrecpp.cc to allow PCRE_NO_UTF8_CHECK to be
488        set.
489    
490    15. The EBCDIC support had decayed; later updates to the code had included
491        explicit references to (e.g.) \x0a instead of CHAR_LF. There has been a
492        general tidy up of EBCDIC-related issues, and the documentation was also
493        not quite right. There is now a test that can be run on ASCII systems to
494        check some of the EBCDIC-related things (but is it not a full test).
495    
496    16. The new PCRE_STUDY_EXTRA_NEEDED option is now used by pcregrep, resulting
497        in a small tidy to the code.
498    
499    17. Fix JIT tests when UTF is disabled and both 8 and 16 bit mode are enabled.
500    
501    18. If the --only-matching (-o) option in pcregrep is specified multiple
502        times, each one causes appropriate output. For example, -o1 -o2 outputs the
503        substrings matched by the 1st and 2nd capturing parentheses. A separating
504        string can be specified by --om-separator (default empty).
505    
506    19. Improving the first n character searches.
507    
508    20. Turn case lists for horizontal and vertical white space into macros so that
509        they are defined only once.
510    
511    21. This set of changes together give more compatible Unicode case-folding
512        behaviour for characters that have more than one other case when UCP
513        support is available.
514    
515        (a) The Unicode property table now has offsets into a new table of sets of
516            three or more characters that are case-equivalent. The MultiStage2.py
517            script that generates these tables (the pcre_ucd.c file) now scans
518            CaseFolding.txt instead of UnicodeData.txt for character case
519            information.
520    
521        (b) The code for adding characters or ranges of characters to a character
522            class has been abstracted into a generalized function that also handles
523            case-independence. In UTF-mode with UCP support, this uses the new data
524            to handle characters with more than one other case.
525    
526        (c) A bug that is fixed as a result of (b) is that codepoints less than 256
527            whose other case is greater than 256 are now correctly matched
528            caselessly. Previously, the high codepoint matched the low one, but not
529            vice versa.
530    
531        (d) The processing of \h, \H, \v, and \ in character classes now makes use
532            of the new class addition function, using character lists defined as
533            macros alongside the case definitions of 20 above.
534    
535        (e) Caseless back references now work with characters that have more than
536            one other case.
537    
538        (f) General caseless matching of characters with more than one other case
539            is supported.
540    
541    22. Unicode character properties were updated from Unicode 6.2.0
542    
543    23. Improved CMake support under Windows. Patch by Daniel Richard G.
544    
545    24. Add support for 32-bit character strings, and UTF-32
546    
547    25. Major JIT compiler update (code refactoring and bugfixing).
548        Experimental Sparc 32 support is added.
549    
550    26. Applied a modified version of Daniel Richard G's patch to create
551        pcre.h.generic and config.h.generic by "make" instead of in the
552        PrepareRelease script.
553    
554    27. Added a definition for CHAR_NULL (helpful for the z/OS port), and use it in
555        pcre_compile.c when checking for a zero character.
556    
557    28. Introducing a native interface for JIT. Through this interface, the compiled
558        machine code can be directly executed. The purpose of this interface is to
559        provide fast pattern matching, so several sanity checks are not performed.
560        However, feature tests are still performed. The new interface provides
561        1.4x speedup compared to the old one.
562    
563    29. If pcre_exec() or pcre_dfa_exec() was called with a negative value for
564        the subject string length, the error given was PCRE_ERROR_BADOFFSET, which
565        was confusing. There is now a new error PCRE_ERROR_BADLENGTH for this case.
566    
567    30. In 8-bit UTF-8 mode, pcretest failed to give an error for data codepoints
568        greater than 0x7fffffff (which cannot be represented in UTF-8, even under
569        the "old" RFC 2279). Instead, it ended up passing a negative length to
570        pcre_exec().
571    
572    31. Add support for GCC's visibility feature to hide internal functions.
573    
574    32. Running "pcretest -C pcre8" or "pcretest -C pcre16" gave a spurious error
575        "unknown -C option" after outputting 0 or 1.
576    
577    33. There is now support for generating a code coverage report for the test
578        suite in environments where gcc is the compiler and lcov is installed. This
579        is mainly for the benefit of the developers.
580    
581    34. If PCRE is built with --enable-valgrind, certain memory regions are marked
582        unaddressable using valgrind annotations, allowing valgrind to detect
583        invalid memory accesses. This is mainly for the benefit of the developers.
584    
585    25. (*UTF) can now be used to start a pattern in any of the three libraries.
586    
587    26. Give configure error if --enable-cpp but no C++ compiler found.
588    
589    
590    Version 8.31 06-July-2012
591    -------------------------
592    
593    1.  Fixing a wrong JIT test case and some compiler warnings.
594    
595    2.  Removed a bashism from the RunTest script.
596    
597    3.  Add a cast to pcre_exec.c to fix the warning "unary minus operator applied
598        to unsigned type, result still unsigned" that was given by an MS compiler
599        on encountering the code "-sizeof(xxx)".
600    
601    4.  Partial matching support is added to the JIT compiler.
602    
603    5.  Fixed several bugs concerned with partial matching of items that consist
604        of more than one character:
605    
606        (a) /^(..)\1/ did not partially match "aba" because checking references was
607            done on an "all or nothing" basis. This also applied to repeated
608            references.
609    
610        (b) \R did not give a hard partial match if \r was found at the end of the
611            subject.
612    
613        (c) \X did not give a hard partial match after matching one or more
614            characters at the end of the subject.
615    
616        (d) When newline was set to CRLF, a pattern such as /a$/ did not recognize
617            a partial match for the string "\r".
618    
619        (e) When newline was set to CRLF, the metacharacter "." did not recognize
620            a partial match for a CR character at the end of the subject string.
621    
622    6.  If JIT is requested using /S++ or -s++ (instead of just /S+ or -s+) when
623        running pcretest, the text "(JIT)" added to the output whenever JIT is
624        actually used to run the match.
625    
626    7.  Individual JIT compile options can be set in pcretest by following -s+[+]
627        or /S+[+] with a digit between 1 and 7.
628    
629    8.  OP_NOT now supports any UTF character not just single-byte ones.
630    
631    9.  (*MARK) control verb is now supported by the JIT compiler.
632    
633    10. The command "./RunTest list" lists the available tests without actually
634        running any of them. (Because I keep forgetting what they all are.)
635    
636    11. Add PCRE_INFO_MAXLOOKBEHIND.
637    
638    12. Applied a (slightly modified) user-supplied patch that improves performance
639        when the heap is used for recursion (compiled with --disable-stack-for-
640        recursion). Instead of malloc and free for each heap frame each time a
641        logical recursion happens, frames are retained on a chain and re-used where
642        possible. This sometimes gives as much as 30% improvement.
643    
644    13. As documented, (*COMMIT) is now confined to within a recursive subpattern
645        call.
646    
647    14. As documented, (*COMMIT) is now confined to within a positive assertion.
648    
649    15. It is now possible to link pcretest with libedit as an alternative to
650        libreadline.
651    
652    16. (*COMMIT) control verb is now supported by the JIT compiler.
653    
654    17. The Unicode data tables have been updated to Unicode 6.1.0.
655    
656    18. Added --file-list option to pcregrep.
657    
658    19. Added binary file support to pcregrep, including the -a, --binary-files,
659        -I, and --text options.
660    
661    20. The madvise function is renamed for posix_madvise for QNX compatibility
662        reasons. Fixed by Giuseppe D'Angelo.
663    
664    21. Fixed a bug for backward assertions with REVERSE 0 in the JIT compiler.
665    
666    22. Changed the option for creating symbolic links for 16-bit man pages from
667        -s to -sf so that re-installing does not cause issues.
668    
669    23. Support PCRE_NO_START_OPTIMIZE in JIT as (*MARK) support requires it.
670    
671    24. Fixed a very old bug in pcretest that caused errors with restarted DFA
672        matches in certain environments (the workspace was not being correctly
673        retained). Also added to pcre_dfa_exec() a simple plausibility check on
674        some of the workspace data at the beginning of a restart.
675    
676    25. \s*\R was auto-possessifying the \s* when it should not, whereas \S*\R
677        was not doing so when it should - probably a typo introduced by SVN 528
678        (change 8.10/14).
679    
680    26. When PCRE_UCP was not set, \w+\x{c4} was incorrectly auto-possessifying the
681        \w+ when the character tables indicated that \x{c4} was a word character.
682        There were several related cases, all because the tests for doing a table
683        lookup were testing for characters less than 127 instead of 255.
684    
685    27. If a pattern contains capturing parentheses that are not used in a match,
686        their slots in the ovector are set to -1. For those that are higher than
687        any matched groups, this happens at the end of processing. In the case when
688        there were back references that the ovector was too small to contain
689        (causing temporary malloc'd memory to be used during matching), and the
690        highest capturing number was not used, memory off the end of the ovector
691        was incorrectly being set to -1. (It was using the size of the temporary
692        memory instead of the true size.)
693    
694    28. To catch bugs like 27 using valgrind, when pcretest is asked to specify an
695        ovector size, it uses memory at the end of the block that it has got.
696    
697    29. Check for an overlong MARK name and give an error at compile time. The
698        limit is 255 for the 8-bit library and 65535 for the 16-bit library.
699    
700    30. JIT compiler update.
701    
702    31. JIT is now supported on jailbroken iOS devices. Thanks for Ruiger
703        Rill for the patch.
704    
705    32. Put spaces around SLJIT_PRINT_D in the JIT compiler. Required by CXX11.
706    
707    33. Variable renamings in the PCRE-JIT compiler. No functionality change.
708    
709    34. Fixed typos in pcregrep: in two places there was SUPPORT_LIBZ2 instead of
710        SUPPORT_LIBBZ2. This caused a build problem when bzip2 but not gzip (zlib)
711        was enabled.
712    
713    35. Improve JIT code generation for greedy plus quantifier.
714    
715    36. When /((?:a?)*)*c/ or /((?>a?)*)*c/ was matched against "aac", it set group
716        1 to "aa" instead of to an empty string. The bug affected repeated groups
717        that could potentially match an empty string.
718    
719    37. Optimizing single character iterators in JIT.
720    
721    38. Wide characters specified with \uxxxx in JavaScript mode are now subject to
722        the same checks as \x{...} characters in non-JavaScript mode. Specifically,
723        codepoints that are too big for the mode are faulted, and in a UTF mode,
724        disallowed codepoints are also faulted.
725    
726    39. If PCRE was compiled with UTF support, in three places in the DFA
727        matcher there was code that should only have been obeyed in UTF mode, but
728        was being obeyed unconditionally. In 8-bit mode this could cause incorrect
729        processing when bytes with values greater than 127 were present. In 16-bit
730        mode the bug would be provoked by values in the range 0xfc00 to 0xdc00. In
731        both cases the values are those that cannot be the first data item in a UTF
732        character. The three items that might have provoked this were recursions,
733        possessively repeated groups, and atomic groups.
734    
735    40. Ensure that libpcre is explicitly listed in the link commands for pcretest
736        and pcregrep, because some OS require shared objects to be explicitly
737        passed to ld, causing the link step to fail if they are not.
738    
739    41. There were two incorrect #ifdefs in pcre_study.c, meaning that, in 16-bit
740        mode, patterns that started with \h* or \R* might be incorrectly matched.
741    
742    
743    Version 8.30 04-February-2012
744    -----------------------------
745    
746    1.  Renamed "isnumber" as "is_a_number" because in some Mac environments this
747        name is defined in ctype.h.
748    
749    2.  Fixed a bug in fixed-length calculation for lookbehinds that would show up
750        only in quite long subpatterns.
751    
752    3.  Removed the function pcre_info(), which has been obsolete and deprecated
753        since it was replaced by pcre_fullinfo() in February 2000.
754    
755    4.  For a non-anchored pattern, if (*SKIP) was given with a name that did not
756        match a (*MARK), and the match failed at the start of the subject, a
757        reference to memory before the start of the subject could occur. This bug
758        was introduced by fix 17 of release 8.21.
759    
760    5.  A reference to an unset group with zero minimum repetition was giving
761        totally wrong answers (in non-JavaScript-compatibility mode). For example,
762        /(another)?(\1?)test/ matched against "hello world test". This bug was
763        introduced in release 8.13.
764    
765    6.  Add support for 16-bit character strings (a large amount of work involving
766        many changes and refactorings).
767    
768    7.  RunGrepTest failed on msys because \r\n was replaced by whitespace when the
769        command "pattern=`printf 'xxx\r\njkl'`" was run. The pattern is now taken
770        from a file.
771    
772    8.  Ovector size of 2 is also supported by JIT based pcre_exec (the ovector size
773        rounding is not applied in this particular case).
774    
775    9.  The invalid Unicode surrogate codepoints U+D800 to U+DFFF are now rejected
776        if they appear, or are escaped, in patterns.
777    
778    10. Get rid of a number of -Wunused-but-set-variable warnings.
779    
780    11. The pattern /(?=(*:x))(q|)/ matches an empty string, and returns the mark
781        "x". The similar pattern /(?=(*:x))((*:y)q|)/ did not return a mark at all.
782        Oddly, Perl behaves the same way. PCRE has been fixed so that this pattern
783        also returns the mark "x". This bug applied to capturing parentheses,
784        non-capturing parentheses, and atomic parentheses. It also applied to some
785        assertions.
786    
787    12. Stephen Kelly's patch to CMakeLists.txt allows it to parse the version
788        information out of configure.ac instead of relying on pcre.h.generic, which
789        is not stored in the repository.
790    
791    13. Applied Dmitry V. Levin's patch for a more portable method for linking with
792        -lreadline.
793    
794    14. ZH added PCRE_CONFIG_JITTARGET; added its output to pcretest -C.
795    
796    15. Applied Graycode's patch to put the top-level frame on the stack rather
797        than the heap when not using the stack for recursion. This gives a
798        performance improvement in many cases when recursion is not deep.
799    
800    16. Experimental code added to "pcretest -C" to output the stack frame size.
801    
802    
803    Version 8.21 12-Dec-2011
804    ------------------------
805    
806    1.  Updating the JIT compiler.
807    
808    2.  JIT compiler now supports OP_NCREF, OP_RREF and OP_NRREF. New test cases
809        are added as well.
810    
811    3.  Fix cache-flush issue on PowerPC (It is still an experimental JIT port).
812        PCRE_EXTRA_TABLES is not suported by JIT, and should be checked before
813        calling _pcre_jit_exec. Some extra comments are added.
814    
815    4.  (*MARK) settings inside atomic groups that do not contain any capturing
816        parentheses, for example, (?>a(*:m)), were not being passed out. This bug
817        was introduced by change 18 for 8.20.
818    
819    5.  Supporting of \x, \U and \u in JavaScript compatibility mode based on the
820        ECMA-262 standard.
821    
822    6.  Lookbehinds such as (?<=a{2}b) that contained a fixed repetition were
823        erroneously being rejected as "not fixed length" if PCRE_CASELESS was set.
824        This bug was probably introduced by change 9 of 8.13.
825    
826    7.  While fixing 6 above, I noticed that a number of other items were being
827        incorrectly rejected as "not fixed length". This arose partly because newer
828        opcodes had not been added to the fixed-length checking code. I have (a)
829        corrected the bug and added tests for these items, and (b) arranged for an
830        error to occur if an unknown opcode is encountered while checking for fixed
831        length instead of just assuming "not fixed length". The items that were
832        rejected were: (*ACCEPT), (*COMMIT), (*FAIL), (*MARK), (*PRUNE), (*SKIP),
833        (*THEN), \h, \H, \v, \V, and single character negative classes with fixed
834        repetitions, e.g. [^a]{3}, with and without PCRE_CASELESS.
835    
836    8.  A possessively repeated conditional subpattern such as (?(?=c)c|d)++ was
837        being incorrectly compiled and would have given unpredicatble results.
838    
839    9.  A possessively repeated subpattern with minimum repeat count greater than
840        one behaved incorrectly. For example, (A){2,}+ behaved as if it was
841        (A)(A)++ which meant that, after a subsequent mismatch, backtracking into
842        the first (A) could occur when it should not.
843    
844    10. Add a cast and remove a redundant test from the code.
845    
846    11. JIT should use pcre_malloc/pcre_free for allocation.
847    
848    12. Updated pcre-config so that it no longer shows -L/usr/lib, which seems
849        best practice nowadays, and helps with cross-compiling. (If the exec_prefix
850        is anything other than /usr, -L is still shown).
851    
852    13. In non-UTF-8 mode, \C is now supported in lookbehinds and DFA matching.
853    
854    14. Perl does not support \N without a following name in a [] class; PCRE now
855        also gives an error.
856    
857    15. If a forward reference was repeated with an upper limit of around 2000,
858        it caused the error "internal error: overran compiling workspace". The
859        maximum number of forward references (including repeats) was limited by the
860        internal workspace, and dependent on the LINK_SIZE. The code has been
861        rewritten so that the workspace expands (via pcre_malloc) if necessary, and
862        the default depends on LINK_SIZE. There is a new upper limit (for safety)
863        of around 200,000 forward references. While doing this, I also speeded up
864        the filling in of repeated forward references.
865    
866    16. A repeated forward reference in a pattern such as (a)(?2){2}(.) was
867        incorrectly expecting the subject to contain another "a" after the start.
868    
869    17. When (*SKIP:name) is activated without a corresponding (*MARK:name) earlier
870        in the match, the SKIP should be ignored. This was not happening; instead
871        the SKIP was being treated as NOMATCH. For patterns such as
872        /A(*MARK:A)A+(*SKIP:B)Z|AAC/ this meant that the AAC branch was never
873        tested.
874    
875    18. The behaviour of (*MARK), (*PRUNE), and (*THEN) has been reworked and is
876        now much more compatible with Perl, in particular in cases where the result
877        is a non-match for a non-anchored pattern. For example, if
878        /b(*:m)f|a(*:n)w/ is matched against "abc", the non-match returns the name
879        "m", where previously it did not return a name. A side effect of this
880        change is that for partial matches, the last encountered mark name is
881        returned, as for non matches. A number of tests that were previously not
882        Perl-compatible have been moved into the Perl-compatible test files. The
883        refactoring has had the pleasing side effect of removing one argument from
884        the match() function, thus reducing its stack requirements.
885    
886    19. If the /S+ option was used in pcretest to study a pattern using JIT,
887        subsequent uses of /S (without +) incorrectly behaved like /S+.
888    
889    21. Retrieve executable code size support for the JIT compiler and fixing
890        some warnings.
891    
892    22. A caseless match of a UTF-8 character whose other case uses fewer bytes did
893        not work when the shorter character appeared right at the end of the
894        subject string.
895    
896    23. Added some (int) casts to non-JIT modules to reduce warnings on 64-bit
897        systems.
898    
899    24. Added PCRE_INFO_JITSIZE to pass on the value from (21) above, and also
900        output it when the /M option is used in pcretest.
901    
902    25. The CheckMan script was not being included in the distribution. Also, added
903        an explicit "perl" to run Perl scripts from the PrepareRelease script
904        because this is reportedly needed in Windows.
905    
906    26. If study data was being save in a file and studying had not found a set of
907        "starts with" bytes for the pattern, the data written to the file (though
908        never used) was taken from uninitialized memory and so caused valgrind to
909        complain.
910    
911    27. Updated RunTest.bat as provided by Sheri Pierce.
912    
913    28. Fixed a possible uninitialized memory bug in pcre_jit_compile.c.
914    
915    29. Computation of memory usage for the table of capturing group names was
916        giving an unnecessarily large value.
917    
918    
919    Version 8.20 21-Oct-2011
920    ------------------------
921    
922    1.  Change 37 of 8.13 broke patterns like [:a]...[b:] because it thought it had
923        a POSIX class. After further experiments with Perl, which convinced me that
924        Perl has bugs and confusions, a closing square bracket is no longer allowed
925        in a POSIX name. This bug also affected patterns with classes that started
926        with full stops.
927    
928    2.  If a pattern such as /(a)b|ac/ is matched against "ac", there is no
929        captured substring, but while checking the failing first alternative,
930        substring 1 is temporarily captured. If the output vector supplied to
931        pcre_exec() was not big enough for this capture, the yield of the function
932        was still zero ("insufficient space for captured substrings"). This cannot
933        be totally fixed without adding another stack variable, which seems a lot
934        of expense for a edge case. However, I have improved the situation in cases
935        such as /(a)(b)x|abc/ matched against "abc", where the return code
936        indicates that fewer than the maximum number of slots in the ovector have
937        been set.
938    
939    3.  Related to (2) above: when there are more back references in a pattern than
940        slots in the output vector, pcre_exec() uses temporary memory during
941        matching, and copies in the captures as far as possible afterwards. It was
942        using the entire output vector, but this conflicts with the specification
943        that only 2/3 is used for passing back captured substrings. Now it uses
944        only the first 2/3, for compatibility. This is, of course, another edge
945        case.
946    
947    4.  Zoltan Herczeg's just-in-time compiler support has been integrated into the
948        main code base, and can be used by building with --enable-jit. When this is
949        done, pcregrep automatically uses it unless --disable-pcregrep-jit or the
950        runtime --no-jit option is given.
951    
952    5.  When the number of matches in a pcre_dfa_exec() run exactly filled the
953        ovector, the return from the function was zero, implying that there were
954        other matches that did not fit. The correct "exactly full" value is now
955        returned.
956    
957    6.  If a subpattern that was called recursively or as a subroutine contained
958        (*PRUNE) or any other control that caused it to give a non-standard return,
959        invalid errors such as "Error -26 (nested recursion at the same subject
960        position)" or even infinite loops could occur.
961    
962    7.  If a pattern such as /a(*SKIP)c|b(*ACCEPT)|/ was studied, it stopped
963        computing the minimum length on reaching *ACCEPT, and so ended up with the
964        wrong value of 1 rather than 0. Further investigation indicates that
965        computing a minimum subject length in the presence of *ACCEPT is difficult
966        (think back references, subroutine calls), and so I have changed the code
967        so that no minimum is registered for a pattern that contains *ACCEPT.
968    
969    8.  If (*THEN) was present in the first (true) branch of a conditional group,
970        it was not handled as intended. [But see 16 below.]
971    
972    9.  Replaced RunTest.bat and CMakeLists.txt with improved versions provided by
973        Sheri Pierce.
974    
975    10. A pathological pattern such as /(*ACCEPT)a/ was miscompiled, thinking that
976        the first byte in a match must be "a".
977    
978    11. Change 17 for 8.13 increased the recursion depth for patterns like
979        /a(?:.)*?a/ drastically. I've improved things by remembering whether a
980        pattern contains any instances of (*THEN). If it does not, the old
981        optimizations are restored. It would be nice to do this on a per-group
982        basis, but at the moment that is not feasible.
983    
984    12. In some environments, the output of pcretest -C is CRLF terminated. This
985        broke RunTest's code that checks for the link size. A single white space
986        character after the value is now allowed for.
987    
988    13. RunTest now checks for the "fr" locale as well as for "fr_FR" and "french".
989        For "fr", it uses the Windows-specific input and output files.
990    
991    14. If (*THEN) appeared in a group that was called recursively or as a
992        subroutine, it did not work as intended. [But see next item.]
993    
994    15. Consider the pattern /A (B(*THEN)C) | D/ where A, B, C, and D are complex
995        pattern fragments (but not containing any | characters). If A and B are
996        matched, but there is a failure in C so that it backtracks to (*THEN), PCRE
997        was behaving differently to Perl. PCRE backtracked into A, but Perl goes to
998        D. In other words, Perl considers parentheses that do not contain any |
999        characters to be part of a surrounding alternative, whereas PCRE was
1000        treading (B(*THEN)C) the same as (B(*THEN)C|(*FAIL)) -- which Perl handles
1001        differently. PCRE now behaves in the same way as Perl, except in the case
1002        of subroutine/recursion calls such as (?1) which have in any case always
1003        been different (but PCRE had them first :-).
1004    
1005    16. Related to 15 above: Perl does not treat the | in a conditional group as
1006        creating alternatives. Such a group is treated in the same way as an
1007        ordinary group without any | characters when processing (*THEN). PCRE has
1008        been changed to match Perl's behaviour.
1009    
1010    17. If a user had set PCREGREP_COLO(U)R to something other than 1:31, the
1011        RunGrepTest script failed.
1012    
1013    18. Change 22 for version 13 caused atomic groups to use more stack. This is
1014        inevitable for groups that contain captures, but it can lead to a lot of
1015        stack use in large patterns. The old behaviour has been restored for atomic
1016        groups that do not contain any capturing parentheses.
1017    
1018    19. If the PCRE_NO_START_OPTIMIZE option was set for pcre_compile(), it did not
1019        suppress the check for a minimum subject length at run time. (If it was
1020        given to pcre_exec() or pcre_dfa_exec() it did work.)
1021    
1022    20. Fixed an ASCII-dependent infelicity in pcretest that would have made it
1023        fail to work when decoding hex characters in data strings in EBCDIC
1024        environments.
1025    
1026    21. It appears that in at least one Mac OS environment, the isxdigit() function
1027        is implemented as a macro that evaluates to its argument more than once,
1028        contravening the C 90 Standard (I haven't checked a later standard). There
1029        was an instance in pcretest which caused it to go wrong when processing
1030        \x{...} escapes in subject strings. The has been rewritten to avoid using
1031        things like p++ in the argument of isxdigit().
1032    
1033    
1034    Version 8.13 16-Aug-2011
1035    ------------------------
1036    
1037    1.  The Unicode data tables have been updated to Unicode 6.0.0.
1038    
1039    2.  Two minor typos in pcre_internal.h have been fixed.
1040    
1041    3.  Added #include <string.h> to pcre_scanner_unittest.cc, pcrecpp.cc, and
1042        pcrecpp_unittest.cc. They are needed for strcmp(), memset(), and strchr()
1043        in some environments (e.g. Solaris 10/SPARC using Sun Studio 12U2).
1044    
1045    4.  There were a number of related bugs in the code for matching backrefences
1046        caselessly in UTF-8 mode when codes for the characters concerned were
1047        different numbers of bytes. For example, U+023A and U+2C65 are an upper
1048        and lower case pair, using 2 and 3 bytes, respectively. The main bugs were:
1049        (a) A reference to 3 copies of a 2-byte code matched only 2 of a 3-byte
1050        code. (b) A reference to 2 copies of a 3-byte code would not match 2 of a
1051        2-byte code at the end of the subject (it thought there wasn't enough data
1052        left).
1053    
1054    5.  Comprehensive information about what went wrong is now returned by
1055        pcre_exec() and pcre_dfa_exec() when the UTF-8 string check fails, as long
1056        as the output vector has at least 2 elements. The offset of the start of
1057        the failing character and a reason code are placed in the vector.
1058    
1059    6.  When the UTF-8 string check fails for pcre_compile(), the offset that is
1060        now returned is for the first byte of the failing character, instead of the
1061        last byte inspected. This is an incompatible change, but I hope it is small
1062        enough not to be a problem. It makes the returned offset consistent with
1063        pcre_exec() and pcre_dfa_exec().
1064    
1065    7.  pcretest now gives a text phrase as well as the error number when
1066        pcre_exec() or pcre_dfa_exec() fails; if the error is a UTF-8 check
1067        failure, the offset and reason code are output.
1068    
1069    8.  When \R was used with a maximizing quantifier it failed to skip backwards
1070        over a \r\n pair if the subsequent match failed. Instead, it just skipped
1071        back over a single character (\n). This seems wrong (because it treated the
1072        two characters as a single entity when going forwards), conflicts with the
1073        documentation that \R is equivalent to (?>\r\n|\n|...etc), and makes the
1074        behaviour of \R* different to (\R)*, which also seems wrong. The behaviour
1075        has been changed.
1076    
1077    9.  Some internal refactoring has changed the processing so that the handling
1078        of the PCRE_CASELESS and PCRE_MULTILINE options is done entirely at compile
1079        time (the PCRE_DOTALL option was changed this way some time ago: version
1080        7.7 change 16). This has made it possible to abolish the OP_OPT op code,
1081        which was always a bit of a fudge. It also means that there is one less
1082        argument for the match() function, which reduces its stack requirements
1083        slightly. This change also fixes an incompatibility with Perl: the pattern
1084        (?i:([^b]))(?1) should not match "ab", but previously PCRE gave a match.
1085    
1086    10. More internal refactoring has drastically reduced the number of recursive
1087        calls to match() for possessively repeated groups such as (abc)++ when
1088        using pcre_exec().
1089    
1090    11. While implementing 10, a number of bugs in the handling of groups were
1091        discovered and fixed:
1092    
1093        (?<=(a)+) was not diagnosed as invalid (non-fixed-length lookbehind).
1094        (a|)*(?1) gave a compile-time internal error.
1095        ((a|)+)+  did not notice that the outer group could match an empty string.
1096        (^a|^)+   was not marked as anchored.
1097        (.*a|.*)+ was not marked as matching at start or after a newline.
1098    
1099    12. Yet more internal refactoring has removed another argument from the match()
1100        function. Special calls to this function are now indicated by setting a
1101        value in a variable in the "match data" data block.
1102    
1103    13. Be more explicit in pcre_study() instead of relying on "default" for
1104        opcodes that mean there is no starting character; this means that when new
1105        ones are added and accidentally left out of pcre_study(), testing should
1106        pick them up.
1107    
1108    14. The -s option of pcretest has been documented for ages as being an old
1109        synonym of -m (show memory usage). I have changed it to mean "force study
1110        for every regex", that is, assume /S for every regex. This is similar to -i
1111        and -d etc. It's slightly incompatible, but I'm hoping nobody is still
1112        using it. It makes it easier to run collections of tests with and without
1113        study enabled, and thereby test pcre_study() more easily. All the standard
1114        tests are now run with and without -s (but some patterns can be marked as
1115        "never study" - see 20 below).
1116    
1117    15. When (*ACCEPT) was used in a subpattern that was called recursively, the
1118        restoration of the capturing data to the outer values was not happening
1119        correctly.
1120    
1121    16. If a recursively called subpattern ended with (*ACCEPT) and matched an
1122        empty string, and PCRE_NOTEMPTY was set, pcre_exec() thought the whole
1123        pattern had matched an empty string, and so incorrectly returned a no
1124        match.
1125    
1126    17. There was optimizing code for the last branch of non-capturing parentheses,
1127        and also for the obeyed branch of a conditional subexpression, which used
1128        tail recursion to cut down on stack usage. Unfortunately, now that there is
1129        the possibility of (*THEN) occurring in these branches, tail recursion is
1130        no longer possible because the return has to be checked for (*THEN). These
1131        two optimizations have therefore been removed. [But see 8.20/11 above.]
1132    
1133    18. If a pattern containing \R was studied, it was assumed that \R always
1134        matched two bytes, thus causing the minimum subject length to be
1135        incorrectly computed because \R can also match just one byte.
1136    
1137    19. If a pattern containing (*ACCEPT) was studied, the minimum subject length
1138        was incorrectly computed.
1139    
1140    20. If /S is present twice on a test pattern in pcretest input, it now
1141        *disables* studying, thereby overriding the use of -s on the command line
1142        (see 14 above). This is necessary for one or two tests to keep the output
1143        identical in both cases.
1144    
1145    21. When (*ACCEPT) was used in an assertion that matched an empty string and
1146        PCRE_NOTEMPTY was set, PCRE applied the non-empty test to the assertion.
1147    
1148    22. When an atomic group that contained a capturing parenthesis was
1149        successfully matched, but the branch in which it appeared failed, the
1150        capturing was not being forgotten if a higher numbered group was later
1151        captured. For example, /(?>(a))b|(a)c/ when matching "ac" set capturing
1152        group 1 to "a", when in fact it should be unset. This applied to multi-
1153        branched capturing and non-capturing groups, repeated or not, and also to
1154        positive assertions (capturing in negative assertions does not happen
1155        in PCRE) and also to nested atomic groups.
1156    
1157    23. Add the ++ qualifier feature to pcretest, to show the remainder of the
1158        subject after a captured substring, to make it easier to tell which of a
1159        number of identical substrings has been captured.
1160    
1161    24. The way atomic groups are processed by pcre_exec() has been changed so that
1162        if they are repeated, backtracking one repetition now resets captured
1163        values correctly. For example, if ((?>(a+)b)+aabab) is matched against
1164        "aaaabaaabaabab" the value of captured group 2 is now correctly recorded as
1165        "aaa". Previously, it would have been "a". As part of this code
1166        refactoring, the way recursive calls are handled has also been changed.
1167    
1168    25. If an assertion condition captured any substrings, they were not passed
1169        back unless some other capturing happened later. For example, if
1170        (?(?=(a))a) was matched against "a", no capturing was returned.
1171    
1172    26. When studying a pattern that contained subroutine calls or assertions,
1173        the code for finding the minimum length of a possible match was handling
1174        direct recursions such as (xxx(?1)|yyy) but not mutual recursions (where
1175        group 1 called group 2 while simultaneously a separate group 2 called group
1176        1). A stack overflow occurred in this case. I have fixed this by limiting
1177        the recursion depth to 10.
1178    
1179    27. Updated RunTest.bat in the distribution to the version supplied by Tom
1180        Fortmann. This supports explicit test numbers on the command line, and has
1181        argument validation and error reporting.
1182    
1183    28. An instance of \X with an unlimited repeat could fail if at any point the
1184        first character it looked at was a mark character.
1185    
1186    29. Some minor code refactoring concerning Unicode properties and scripts
1187        should reduce the stack requirement of match() slightly.
1188    
1189    30. Added the '=' option to pcretest to check the setting of unused capturing
1190        slots at the end of the pattern, which are documented as being -1, but are
1191        not included in the return count.
1192    
1193    31. If \k was not followed by a braced, angle-bracketed, or quoted name, PCRE
1194        compiled something random. Now it gives a compile-time error (as does
1195        Perl).
1196    
1197    32. A *MARK encountered during the processing of a positive assertion is now
1198        recorded and passed back (compatible with Perl).
1199    
1200    33. If --only-matching or --colour was set on a pcregrep call whose pattern
1201        had alternative anchored branches, the search for a second match in a line
1202        was done as if at the line start. Thus, for example, /^01|^02/ incorrectly
1203        matched the line "0102" twice. The same bug affected patterns that started
1204        with a backwards assertion. For example /\b01|\b02/ also matched "0102"
1205        twice.
1206    
1207    34. Previously, PCRE did not allow quantification of assertions. However, Perl
1208        does, and because of capturing effects, quantifying parenthesized
1209        assertions may at times be useful. Quantifiers are now allowed for
1210        parenthesized assertions.
1211    
1212    35. A minor code tidy in pcre_compile() when checking options for \R usage.
1213    
1214    36. \g was being checked for fancy things in a character class, when it should
1215        just be a literal "g".
1216    
1217    37. PCRE was rejecting [:a[:digit:]] whereas Perl was not. It seems that the
1218        appearance of a nested POSIX class supersedes an apparent external class.
1219        For example, [:a[:digit:]b:] matches "a", "b", ":", or a digit. Also,
1220        unescaped square brackets may also appear as part of class names. For
1221        example, [:a[:abc]b:] gives unknown class "[:abc]b:]". PCRE now behaves
1222        more like Perl. (But see 8.20/1 above.)
1223    
1224    38. PCRE was giving an error for \N with a braced quantifier such as {1,} (this
1225        was because it thought it was \N{name}, which is not supported).
1226    
1227    39. Add minix to OS list not supporting the -S option in pcretest.
1228    
1229    40. PCRE tries to detect cases of infinite recursion at compile time, but it
1230        cannot analyze patterns in sufficient detail to catch mutual recursions
1231        such as ((?1))((?2)). There is now a runtime test that gives an error if a
1232        subgroup is called recursively as a subpattern for a second time at the
1233        same position in the subject string. In previous releases this might have
1234        been caught by the recursion limit, or it might have run out of stack.
1235    
1236    41. A pattern such as /(?(R)a+|(?R)b)/ is quite safe, as the recursion can
1237        happen only once. PCRE was, however incorrectly giving a compile time error
1238        "recursive call could loop indefinitely" because it cannot analyze the
1239        pattern in sufficient detail. The compile time test no longer happens when
1240        PCRE is compiling a conditional subpattern, but actual runaway loops are
1241        now caught at runtime (see 40 above).
1242    
1243    42. It seems that Perl allows any characters other than a closing parenthesis
1244        to be part of the NAME in (*MARK:NAME) and other backtracking verbs. PCRE
1245        has been changed to be the same.
1246    
1247    43. Updated configure.ac to put in more quoting round AC_LANG_PROGRAM etc. so
1248        as not to get warnings when autogen.sh is called. Also changed
1249        AC_PROG_LIBTOOL (deprecated) to LT_INIT (the current macro).
1250    
1251    44. To help people who use pcregrep to scan files containing exceedingly long
1252        lines, the following changes have been made:
1253    
1254        (a) The default value of the buffer size parameter has been increased from
1255            8K to 20K. (The actual buffer used is three times this size.)
1256    
1257        (b) The default can be changed by ./configure --with-pcregrep-bufsize when
1258            PCRE is built.
1259    
1260        (c) A --buffer-size=n option has been added to pcregrep, to allow the size
1261            to be set at run time.
1262    
1263        (d) Numerical values in pcregrep options can be followed by K or M, for
1264            example --buffer-size=50K.
1265    
1266        (e) If a line being scanned overflows pcregrep's buffer, an error is now
1267            given and the return code is set to 2.
1268    
1269    45. Add a pointer to the latest mark to the callout data block.
1270    
1271    46. The pattern /.(*F)/, when applied to "abc" with PCRE_PARTIAL_HARD, gave a
1272        partial match of an empty string instead of no match. This was specific to
1273        the use of ".".
1274    
1275    47. The pattern /f.*/8s, when applied to "for" with PCRE_PARTIAL_HARD, gave a
1276        complete match instead of a partial match. This bug was dependent on both
1277        the PCRE_UTF8 and PCRE_DOTALL options being set.
1278    
1279    48. For a pattern such as /\babc|\bdef/ pcre_study() was failing to set up the
1280        starting byte set, because \b was not being ignored.
1281    
1282    
1283    Version 8.12 15-Jan-2011
1284    ------------------------
1285    
1286    1.  Fixed some typos in the markup of the man pages, and wrote a script that
1287        checks for such things as part of the documentation building process.
1288    
1289    2.  On a big-endian 64-bit system, pcregrep did not correctly process the
1290        --match-limit and --recursion-limit options (added for 8.11). In
1291        particular, this made one of the standard tests fail. (The integer value
1292        went into the wrong half of a long int.)
1293    
1294    3.  If the --colour option was given to pcregrep with -v (invert match), it
1295        did strange things, either producing crazy output, or crashing. It should,
1296        of course, ignore a request for colour when reporting lines that do not
1297        match.
1298    
1299    4.  Another pcregrep bug caused similar problems if --colour was specified with
1300        -M (multiline) and the pattern match finished with a line ending.
1301    
1302    5.  In pcregrep, when a pattern that ended with a literal newline sequence was
1303        matched in multiline mode, the following line was shown as part of the
1304        match. This seems wrong, so I have changed it.
1305    
1306    6.  Another pcregrep bug in multiline mode, when --colour was specified, caused
1307        the check for further matches in the same line (so they could be coloured)
1308        to overrun the end of the current line. If another match was found, it was
1309        incorrectly shown (and then shown again when found in the next line).
1310    
1311    7.  If pcregrep was compiled under Windows, there was a reference to the
1312        function pcregrep_exit() before it was defined. I am assuming this was
1313        the cause of the "error C2371: 'pcregrep_exit' : redefinition;" that was
1314        reported by a user. I've moved the definition above the reference.
1315    
1316    
1317    Version 8.11 10-Dec-2010
1318    ------------------------
1319    
1320    1.  (*THEN) was not working properly if there were untried alternatives prior
1321        to it in the current branch. For example, in ((a|b)(*THEN)(*F)|c..) it
1322        backtracked to try for "b" instead of moving to the next alternative branch
1323        at the same level (in this case, to look for "c"). The Perl documentation
1324        is clear that when (*THEN) is backtracked onto, it goes to the "next
1325        alternative in the innermost enclosing group".
1326    
1327    2.  (*COMMIT) was not overriding (*THEN), as it does in Perl. In a pattern
1328        such as   (A(*COMMIT)B(*THEN)C|D)  any failure after matching A should
1329        result in overall failure. Similarly, (*COMMIT) now overrides (*PRUNE) and
1330        (*SKIP), (*SKIP) overrides (*PRUNE) and (*THEN), and (*PRUNE) overrides
1331        (*THEN).
1332    
1333    3.  If \s appeared in a character class, it removed the VT character from
1334        the class, even if it had been included by some previous item, for example
1335        in [\x00-\xff\s]. (This was a bug related to the fact that VT is not part
1336        of \s, but is part of the POSIX "space" class.)
1337    
1338    4.  A partial match never returns an empty string (because you can always
1339        match an empty string at the end of the subject); however the checking for
1340        an empty string was starting at the "start of match" point. This has been
1341        changed to the "earliest inspected character" point, because the returned
1342        data for a partial match starts at this character. This means that, for
1343        example, /(?<=abc)def/ gives a partial match for the subject "abc"
1344        (previously it gave "no match").
1345    
1346    5.  Changes have been made to the way PCRE_PARTIAL_HARD affects the matching
1347        of $, \z, \Z, \b, and \B. If the match point is at the end of the string,
1348        previously a full match would be given. However, setting PCRE_PARTIAL_HARD
1349        has an implication that the given string is incomplete (because a partial
1350        match is preferred over a full match). For this reason, these items now
1351        give a partial match in this situation. [Aside: previously, the one case
1352        /t\b/ matched against "cat" with PCRE_PARTIAL_HARD set did return a partial
1353        match rather than a full match, which was wrong by the old rules, but is
1354        now correct.]
1355    
1356    6.  There was a bug in the handling of #-introduced comments, recognized when
1357        PCRE_EXTENDED is set, when PCRE_NEWLINE_ANY and PCRE_UTF8 were also set.
1358        If a UTF-8 multi-byte character included the byte 0x85 (e.g. +U0445, whose
1359        UTF-8 encoding is 0xd1,0x85), this was misinterpreted as a newline when
1360        scanning for the end of the comment. (*Character* 0x85 is an "any" newline,
1361        but *byte* 0x85 is not, in UTF-8 mode). This bug was present in several
1362        places in pcre_compile().
1363    
1364    7.  Related to (6) above, when pcre_compile() was skipping #-introduced
1365        comments when looking ahead for named forward references to subpatterns,
1366        the only newline sequence it recognized was NL. It now handles newlines
1367        according to the set newline convention.
1368    
1369    8.  SunOS4 doesn't have strerror() or strtoul(); pcregrep dealt with the
1370        former, but used strtoul(), whereas pcretest avoided strtoul() but did not
1371        cater for a lack of strerror(). These oversights have been fixed.
1372    
1373    9.  Added --match-limit and --recursion-limit to pcregrep.
1374    
1375    10. Added two casts needed to build with Visual Studio when NO_RECURSE is set.
1376    
1377    11. When the -o option was used, pcregrep was setting a return code of 1, even
1378        when matches were found, and --line-buffered was not being honoured.
1379    
1380    12. Added an optional parentheses number to the -o and --only-matching options
1381        of pcregrep.
1382    
1383    13. Imitating Perl's /g action for multiple matches is tricky when the pattern
1384        can match an empty string. The code to do it in pcretest and pcredemo
1385        needed fixing:
1386    
1387        (a) When the newline convention was "crlf", pcretest got it wrong, skipping
1388            only one byte after an empty string match just before CRLF (this case
1389            just got forgotten; "any" and "anycrlf" were OK).
1390    
1391        (b) The pcretest code also had a bug, causing it to loop forever in UTF-8
1392            mode when an empty string match preceded an ASCII character followed by
1393            a non-ASCII character. (The code for advancing by one character rather
1394            than one byte was nonsense.)
1395    
1396        (c) The pcredemo.c sample program did not have any code at all to handle
1397            the cases when CRLF is a valid newline sequence.
1398    
1399    14. Neither pcre_exec() nor pcre_dfa_exec() was checking that the value given
1400        as a starting offset was within the subject string. There is now a new
1401        error, PCRE_ERROR_BADOFFSET, which is returned if the starting offset is
1402        negative or greater than the length of the string. In order to test this,
1403        pcretest is extended to allow the setting of negative starting offsets.
1404    
1405    15. In both pcre_exec() and pcre_dfa_exec() the code for checking that the
1406        starting offset points to the beginning of a UTF-8 character was
1407        unnecessarily clumsy. I tidied it up.
1408    
1409    16. Added PCRE_ERROR_SHORTUTF8 to make it possible to distinguish between a
1410        bad UTF-8 sequence and one that is incomplete when using PCRE_PARTIAL_HARD.
1411    
1412    17. Nobody had reported that the --include_dir option, which was added in
1413        release 7.7 should have been called --include-dir (hyphen, not underscore)
1414        for compatibility with GNU grep. I have changed it to --include-dir, but
1415        left --include_dir as an undocumented synonym, and the same for
1416        --exclude-dir, though that is not available in GNU grep, at least as of
1417        release 2.5.4.
1418    
1419    18. At a user's suggestion, the macros GETCHAR and friends (which pick up UTF-8
1420        characters from a string of bytes) have been redefined so as not to use
1421        loops, in order to improve performance in some environments. At the same
1422        time, I abstracted some of the common code into auxiliary macros to save
1423        repetition (this should not affect the compiled code).
1424    
1425    19. If \c was followed by a multibyte UTF-8 character, bad things happened. A
1426        compile-time error is now given if \c is not followed by an ASCII
1427        character, that is, a byte less than 128. (In EBCDIC mode, the code is
1428        different, and any byte value is allowed.)
1429    
1430    20. Recognize (*NO_START_OPT) at the start of a pattern to set the PCRE_NO_
1431        START_OPTIMIZE option, which is now allowed at compile time - but just
1432        passed through to pcre_exec() or pcre_dfa_exec(). This makes it available
1433        to pcregrep and other applications that have no direct access to PCRE
1434        options. The new /Y option in pcretest sets this option when calling
1435        pcre_compile().
1436    
1437    21. Change 18 of release 8.01 broke the use of named subpatterns for recursive
1438        back references. Groups containing recursive back references were forced to
1439        be atomic by that change, but in the case of named groups, the amount of
1440        memory required was incorrectly computed, leading to "Failed: internal
1441        error: code overflow". This has been fixed.
1442    
1443    22. Some patches to pcre_stringpiece.h, pcre_stringpiece_unittest.cc, and
1444        pcretest.c, to avoid build problems in some Borland environments.
1445    
1446    
1447    Version 8.10 25-Jun-2010
1448    ------------------------
1449    
1450    1.  Added support for (*MARK:ARG) and for ARG additions to PRUNE, SKIP, and
1451        THEN.
1452    
1453    2.  (*ACCEPT) was not working when inside an atomic group.
1454    
1455    3.  Inside a character class, \B is treated as a literal by default, but
1456        faulted if PCRE_EXTRA is set. This mimics Perl's behaviour (the -w option
1457        causes the error). The code is unchanged, but I tidied the documentation.
1458    
1459    4.  Inside a character class, PCRE always treated \R and \X as literals,
1460        whereas Perl faults them if its -w option is set. I have changed PCRE so
1461        that it faults them when PCRE_EXTRA is set.
1462    
1463    5.  Added support for \N, which always matches any character other than
1464        newline. (It is the same as "." when PCRE_DOTALL is not set.)
1465    
1466    6.  When compiling pcregrep with newer versions of gcc which may have
1467        FORTIFY_SOURCE set, several warnings "ignoring return value of 'fwrite',
1468        declared with attribute warn_unused_result" were given. Just casting the
1469        result to (void) does not stop the warnings; a more elaborate fudge is
1470        needed. I've used a macro to implement this.
1471    
1472    7.  Minor change to pcretest.c to avoid a compiler warning.
1473    
1474    8.  Added four artifical Unicode properties to help with an option to make
1475        \s etc use properties (see next item). The new properties are: Xan
1476        (alphanumeric), Xsp (Perl space), Xps (POSIX space), and Xwd (word).
1477    
1478    9.  Added PCRE_UCP to make \b, \d, \s, \w, and certain POSIX character classes
1479        use Unicode properties. (*UCP) at the start of a pattern can be used to set
1480        this option. Modified pcretest to add /W to test this facility. Added
1481        REG_UCP to make it available via the POSIX interface.
1482    
1483    10. Added --line-buffered to pcregrep.
1484    
1485    11. In UTF-8 mode, if a pattern that was compiled with PCRE_CASELESS was
1486        studied, and the match started with a letter with a code point greater than
1487        127 whose first byte was different to the first byte of the other case of
1488        the letter, the other case of this starting letter was not recognized
1489        (#976).
1490    
1491    12. If a pattern that was studied started with a repeated Unicode property
1492        test, for example, \p{Nd}+, there was the theoretical possibility of
1493        setting up an incorrect bitmap of starting bytes, but fortunately it could
1494        not have actually happened in practice until change 8 above was made (it
1495        added property types that matched character-matching opcodes).
1496    
1497    13. pcre_study() now recognizes \h, \v, and \R when constructing a bit map of
1498        possible starting bytes for non-anchored patterns.
1499    
1500    14. Extended the "auto-possessify" feature of pcre_compile(). It now recognizes
1501        \R, and also a number of cases that involve Unicode properties, both
1502        explicit and implicit when PCRE_UCP is set.
1503    
1504    15. If a repeated Unicode property match (e.g. \p{Lu}*) was used with non-UTF-8
1505        input, it could crash or give wrong results if characters with values
1506        greater than 0xc0 were present in the subject string. (Detail: it assumed
1507        UTF-8 input when processing these items.)
1508    
1509    16. Added a lot of (int) casts to avoid compiler warnings in systems where
1510        size_t is 64-bit (#991).
1511    
1512    17. Added a check for running out of memory when PCRE is compiled with
1513        --disable-stack-for-recursion (#990).
1514    
1515    18. If the last data line in a file for pcretest does not have a newline on
1516        the end, a newline was missing in the output.
1517    
1518    19. The default pcre_chartables.c file recognizes only ASCII characters (values
1519        less than 128) in its various bitmaps. However, there is a facility for
1520        generating tables according to the current locale when PCRE is compiled. It
1521        turns out that in some environments, 0x85 and 0xa0, which are Unicode space
1522        characters, are recognized by isspace() and therefore were getting set in
1523        these tables, and indeed these tables seem to approximate to ISO 8859. This
1524        caused a problem in UTF-8 mode when pcre_study() was used to create a list
1525        of bytes that can start a match. For \s, it was including 0x85 and 0xa0,
1526        which of course cannot start UTF-8 characters. I have changed the code so
1527        that only real ASCII characters (less than 128) and the correct starting
1528        bytes for UTF-8 encodings are set for characters greater than 127 when in
1529        UTF-8 mode. (When PCRE_UCP is set - see 9 above - the code is different
1530        altogether.)
1531    
1532    20. Added the /T option to pcretest so as to be able to run tests with non-
1533        standard character tables, thus making it possible to include the tests
1534        used for 19 above in the standard set of tests.
1535    
1536    21. A pattern such as (?&t)(?#()(?(DEFINE)(?<t>a)) which has a forward
1537        reference to a subpattern the other side of a comment that contains an
1538        opening parenthesis caused either an internal compiling error, or a
1539        reference to the wrong subpattern.
1540    
1541    
1542    Version 8.02 19-Mar-2010
1543    ------------------------
1544    
1545    1.  The Unicode data tables have been updated to Unicode 5.2.0.
1546    
1547    2.  Added the option --libs-cpp to pcre-config, but only when C++ support is
1548        configured.
1549    
1550    3.  Updated the licensing terms in the pcregexp.pas file, as agreed with the
1551        original author of that file, following a query about its status.
1552    
1553    4.  On systems that do not have stdint.h (e.g. Solaris), check for and include
1554        inttypes.h instead. This fixes a bug that was introduced by change 8.01/8.
1555    
1556    5.  A pattern such as (?&t)*+(?(DEFINE)(?<t>.)) which has a possessive
1557        quantifier applied to a forward-referencing subroutine call, could compile
1558        incorrect code or give the error "internal error: previously-checked
1559        referenced subpattern not found".
1560    
1561    6.  Both MS Visual Studio and Symbian OS have problems with initializing
1562        variables to point to external functions. For these systems, therefore,
1563        pcre_malloc etc. are now initialized to local functions that call the
1564        relevant global functions.
1565    
1566    7.  There were two entries missing in the vectors called coptable and poptable
1567        in pcre_dfa_exec.c. This could lead to memory accesses outsize the vectors.
1568        I've fixed the data, and added a kludgy way of testing at compile time that
1569        the lengths are correct (equal to the number of opcodes).
1570    
1571    8.  Following on from 7, I added a similar kludge to check the length of the
1572        eint vector in pcreposix.c.
1573    
1574    9.  Error texts for pcre_compile() are held as one long string to avoid too
1575        much relocation at load time. To find a text, the string is searched,
1576        counting zeros. There was no check for running off the end of the string,
1577        which could happen if a new error number was added without updating the
1578        string.
1579    
1580    10. \K gave a compile-time error if it appeared in a lookbehind assersion.
1581    
1582    11. \K was not working if it appeared in an atomic group or in a group that
1583        was called as a "subroutine", or in an assertion. Perl 5.11 documents that
1584        \K is "not well defined" if used in an assertion. PCRE now accepts it if
1585        the assertion is positive, but not if it is negative.
1586    
1587    12. Change 11 fortuitously reduced the size of the stack frame used in the
1588        "match()" function of pcre_exec.c by one pointer. Forthcoming
1589        implementation of support for (*MARK) will need an extra pointer on the
1590        stack; I have reserved it now, so that the stack frame size does not
1591        decrease.
1592    
1593    13. A pattern such as (?P<L1>(?P<L2>0)|(?P>L2)(?P>L1)) in which the only other
1594        item in branch that calls a recursion is a subroutine call - as in the
1595        second branch in the above example - was incorrectly given the compile-
1596        time error "recursive call could loop indefinitely" because pcre_compile()
1597        was not correctly checking the subroutine for matching a non-empty string.
1598    
1599    14. The checks for overrunning compiling workspace could trigger after an
1600        overrun had occurred. This is a "should never occur" error, but it can be
1601        triggered by pathological patterns such as hundreds of nested parentheses.
1602        The checks now trigger 100 bytes before the end of the workspace.
1603    
1604    15. Fix typo in configure.ac: "srtoq" should be "strtoq".
1605    
1606    
1607    Version 8.01 19-Jan-2010
1608    ------------------------
1609    
1610    1.  If a pattern contained a conditional subpattern with only one branch (in
1611        particular, this includes all (*DEFINE) patterns), a call to pcre_study()
1612        computed the wrong minimum data length (which is of course zero for such
1613        subpatterns). This could cause incorrect "no match" results.
1614    
1615    2.  For patterns such as (?i)a(?-i)b|c where an option setting at the start of
1616        the pattern is reset in the first branch, pcre_compile() failed with
1617        "internal error: code overflow at offset...". This happened only when
1618        the reset was to the original external option setting. (An optimization
1619        abstracts leading options settings into an external setting, which was the
1620        cause of this.)
1621    
1622    3.  A pattern such as ^(?!a(*SKIP)b) where a negative assertion contained one
1623        of the verbs SKIP, PRUNE, or COMMIT, did not work correctly. When the
1624        assertion pattern did not match (meaning that the assertion was true), it
1625        was incorrectly treated as false if the SKIP had been reached during the
1626        matching. This also applied to assertions used as conditions.
1627    
1628    4.  If an item that is not supported by pcre_dfa_exec() was encountered in an
1629        assertion subpattern, including such a pattern used as a condition,
1630        unpredictable results occurred, instead of the error return
1631        PCRE_ERROR_DFA_UITEM.
1632    
1633    5.  The C++ GlobalReplace function was not working like Perl for the special
1634        situation when an empty string is matched. It now does the fancy magic
1635        stuff that is necessary.
1636    
1637    6.  In pcre_internal.h, obsolete includes to setjmp.h and stdarg.h have been
1638        removed. (These were left over from very, very early versions of PCRE.)
1639    
1640    7.  Some cosmetic changes to the code to make life easier when compiling it
1641        as part of something else:
1642    
1643        (a) Change DEBUG to PCRE_DEBUG.
1644    
1645        (b) In pcre_compile(), rename the member of the "branch_chain" structure
1646            called "current" as "current_branch", to prevent a collision with the
1647            Linux macro when compiled as a kernel module.
1648    
1649        (c) In pcre_study(), rename the function set_bit() as set_table_bit(), to
1650            prevent a collision with the Linux macro when compiled as a kernel
1651            module.
1652    
1653    8.  In pcre_compile() there are some checks for integer overflows that used to
1654        cast potentially large values to (double). This has been changed to that
1655        when building, a check for int64_t is made, and if it is found, it is used
1656        instead, thus avoiding the use of floating point arithmetic. (There is no
1657        other use of FP in PCRE.) If int64_t is not found, the fallback is to
1658        double.
1659    
1660    9.  Added two casts to avoid signed/unsigned warnings from VS Studio Express
1661        2005 (difference between two addresses compared to an unsigned value).
1662    
1663    10. Change the standard AC_CHECK_LIB test for libbz2 in configure.ac to a
1664        custom one, because of the following reported problem in Windows:
1665    
1666          - libbz2 uses the Pascal calling convention (WINAPI) for the functions
1667              under Win32.
1668          - The standard autoconf AC_CHECK_LIB fails to include "bzlib.h",
1669              therefore missing the function definition.
1670          - The compiler thus generates a "C" signature for the test function.
1671          - The linker fails to find the "C" function.
1672          - PCRE fails to configure if asked to do so against libbz2.
1673    
1674    11. When running libtoolize from libtool-2.2.6b as part of autogen.sh, these
1675        messages were output:
1676    
1677          Consider adding `AC_CONFIG_MACRO_DIR([m4])' to configure.ac and
1678          rerunning libtoolize, to keep the correct libtool macros in-tree.
1679          Consider adding `-I m4' to ACLOCAL_AMFLAGS in Makefile.am.
1680    
1681        I have done both of these things.
1682    
1683    12. Although pcre_dfa_exec() does not use nearly as much stack as pcre_exec()
1684        most of the time, it *can* run out if it is given a pattern that contains a
1685        runaway infinite recursion. I updated the discussion in the pcrestack man
1686        page.
1687    
1688    13. Now that we have gone to the x.xx style of version numbers, the minor
1689        version may start with zero. Using 08 or 09 is a bad idea because users
1690        might check the value of PCRE_MINOR in their code, and 08 or 09 may be
1691        interpreted as invalid octal numbers. I've updated the previous comment in
1692        configure.ac, and also added a check that gives an error if 08 or 09 are
1693        used.
1694    
1695    14. Change 8.00/11 was not quite complete: code had been accidentally omitted,
1696        causing partial matching to fail when the end of the subject matched \W
1697        in a UTF-8 pattern where \W was quantified with a minimum of 3.
1698    
1699    15. There were some discrepancies between the declarations in pcre_internal.h
1700        of _pcre_is_newline(), _pcre_was_newline(), and _pcre_valid_utf8() and
1701        their definitions. The declarations used "const uschar *" and the
1702        definitions used USPTR. Even though USPTR is normally defined as "const
1703        unsigned char *" (and uschar is typedeffed as "unsigned char"), it was
1704        reported that: "This difference in casting confuses some C++ compilers, for
1705        example, SunCC recognizes above declarations as different functions and
1706        generates broken code for hbpcre." I have changed the declarations to use
1707        USPTR.
1708    
1709    16. GNU libtool is named differently on some systems. The autogen.sh script now
1710        tries several variants such as glibtoolize (MacOSX) and libtoolize1x
1711        (FreeBSD).
1712    
1713    17. Applied Craig's patch that fixes an HP aCC compile error in pcre 8.00
1714        (strtoXX undefined when compiling pcrecpp.cc). The patch contains this
1715        comment: "Figure out how to create a longlong from a string: strtoll and
1716        equivalent. It's not enough to call AC_CHECK_FUNCS: hpux has a strtoll, for
1717        instance, but it only takes 2 args instead of 3!"
1718    
1719    18. A subtle bug concerned with back references has been fixed by a change of
1720        specification, with a corresponding code fix. A pattern such as
1721        ^(xa|=?\1a)+$ which contains a back reference inside the group to which it
1722        refers, was giving matches when it shouldn't. For example, xa=xaaa would
1723        match that pattern. Interestingly, Perl (at least up to 5.11.3) has the
1724        same bug. Such groups have to be quantified to be useful, or contained
1725        inside another quantified group. (If there's no repetition, the reference
1726        can never match.) The problem arises because, having left the group and
1727        moved on to the rest of the pattern, a later failure that backtracks into
1728        the group uses the captured value from the final iteration of the group
1729        rather than the correct earlier one. I have fixed this in PCRE by forcing
1730        any group that contains a reference to itself to be an atomic group; that
1731        is, there cannot be any backtracking into it once it has completed. This is
1732        similar to recursive and subroutine calls.
1733    
1734    
1735    Version 8.00 19-Oct-09
1736    ----------------------
1737    
1738    1.  The table for translating pcre_compile() error codes into POSIX error codes
1739        was out-of-date, and there was no check on the pcre_compile() error code
1740        being within the table. This could lead to an OK return being given in
1741        error.
1742    
1743    2.  Changed the call to open a subject file in pcregrep from fopen(pathname,
1744        "r") to fopen(pathname, "rb"), which fixed a problem with some of the tests
1745        in a Windows environment.
1746    
1747    3.  The pcregrep --count option prints the count for each file even when it is
1748        zero, as does GNU grep. However, pcregrep was also printing all files when
1749        --files-with-matches was added. Now, when both options are given, it prints
1750        counts only for those files that have at least one match. (GNU grep just
1751        prints the file name in this circumstance, but including the count seems
1752        more useful - otherwise, why use --count?) Also ensured that the
1753        combination -clh just lists non-zero counts, with no names.
1754    
1755    4.  The long form of the pcregrep -F option was incorrectly implemented as
1756        --fixed_strings instead of --fixed-strings. This is an incompatible change,
1757        but it seems right to fix it, and I didn't think it was worth preserving
1758        the old behaviour.
1759    
1760    5.  The command line items --regex=pattern and --regexp=pattern were not
1761        recognized by pcregrep, which required --regex pattern or --regexp pattern
1762        (with a space rather than an '='). The man page documented the '=' forms,
1763        which are compatible with GNU grep; these now work.
1764    
1765    6.  No libpcreposix.pc file was created for pkg-config; there was just
1766        libpcre.pc and libpcrecpp.pc. The omission has been rectified.
1767    
1768    7.  Added #ifndef SUPPORT_UCP into the pcre_ucd.c module, to reduce its size
1769        when UCP support is not needed, by modifying the Python script that
1770        generates it from Unicode data files. This should not matter if the module
1771        is correctly used as a library, but I received one complaint about 50K of
1772        unwanted data. My guess is that the person linked everything into his
1773        program rather than using a library. Anyway, it does no harm.
1774    
1775    8.  A pattern such as /\x{123}{2,2}+/8 was incorrectly compiled; the trigger
1776        was a minimum greater than 1 for a wide character in a possessive
1777        repetition. The same bug could also affect patterns like /(\x{ff}{0,2})*/8
1778        which had an unlimited repeat of a nested, fixed maximum repeat of a wide
1779        character. Chaos in the form of incorrect output or a compiling loop could
1780        result.
1781    
1782    9.  The restrictions on what a pattern can contain when partial matching is
1783        requested for pcre_exec() have been removed. All patterns can now be
1784        partially matched by this function. In addition, if there are at least two
1785        slots in the offset vector, the offset of the earliest inspected character
1786        for the match and the offset of the end of the subject are set in them when
1787        PCRE_ERROR_PARTIAL is returned.
1788    
1789    10. Partial matching has been split into two forms: PCRE_PARTIAL_SOFT, which is
1790        synonymous with PCRE_PARTIAL, for backwards compatibility, and
1791        PCRE_PARTIAL_HARD, which causes a partial match to supersede a full match,
1792        and may be more useful for multi-segment matching.
1793    
1794    11. Partial matching with pcre_exec() is now more intuitive. A partial match
1795        used to be given if ever the end of the subject was reached; now it is
1796        given only if matching could not proceed because another character was
1797        needed. This makes a difference in some odd cases such as Z(*FAIL) with the
1798        string "Z", which now yields "no match" instead of "partial match". In the
1799        case of pcre_dfa_exec(), "no match" is given if every matching path for the
1800        final character ended with (*FAIL).
1801    
1802    12. Restarting a match using pcre_dfa_exec() after a partial match did not work
1803        if the pattern had a "must contain" character that was already found in the
1804        earlier partial match, unless partial matching was again requested. For
1805        example, with the pattern /dog.(body)?/, the "must contain" character is
1806        "g". If the first part-match was for the string "dog", restarting with
1807        "sbody" failed. This bug has been fixed.
1808    
1809    13. The string returned by pcre_dfa_exec() after a partial match has been
1810        changed so that it starts at the first inspected character rather than the
1811        first character of the match. This makes a difference only if the pattern
1812        starts with a lookbehind assertion or \b or \B (\K is not supported by
1813        pcre_dfa_exec()). It's an incompatible change, but it makes the two
1814        matching functions compatible, and I think it's the right thing to do.
1815    
1816    14. Added a pcredemo man page, created automatically from the pcredemo.c file,
1817        so that the demonstration program is easily available in environments where
1818        PCRE has not been installed from source.
1819    
1820    15. Arranged to add -DPCRE_STATIC to cflags in libpcre.pc, libpcreposix.cp,
1821        libpcrecpp.pc and pcre-config when PCRE is not compiled as a shared
1822        library.
1823    
1824    16. Added REG_UNGREEDY to the pcreposix interface, at the request of a user.
1825        It maps to PCRE_UNGREEDY. It is not, of course, POSIX-compatible, but it
1826        is not the first non-POSIX option to be added. Clearly some people find
1827        these options useful.
1828    
1829    17. If a caller to the POSIX matching function regexec() passes a non-zero
1830        value for nmatch with a NULL value for pmatch, the value of
1831        nmatch is forced to zero.
1832    
1833    18. RunGrepTest did not have a test for the availability of the -u option of
1834        the diff command, as RunTest does. It now checks in the same way as
1835        RunTest, and also checks for the -b option.
1836    
1837    19. If an odd number of negated classes containing just a single character
1838        interposed, within parentheses, between a forward reference to a named
1839        subpattern and the definition of the subpattern, compilation crashed with
1840        an internal error, complaining that it could not find the referenced
1841        subpattern. An example of a crashing pattern is /(?&A)(([^m])(?<A>))/.
1842        [The bug was that it was starting one character too far in when skipping
1843        over the character class, thus treating the ] as data rather than
1844        terminating the class. This meant it could skip too much.]
1845    
1846    20. Added PCRE_NOTEMPTY_ATSTART in order to be able to correctly implement the
1847        /g option in pcretest when the pattern contains \K, which makes it possible
1848        to have an empty string match not at the start, even when the pattern is
1849        anchored. Updated pcretest and pcredemo to use this option.
1850    
1851    21. If the maximum number of capturing subpatterns in a recursion was greater
1852        than the maximum at the outer level, the higher number was returned, but
1853        with unset values at the outer level. The correct (outer level) value is
1854        now given.
1855    
1856    22. If (*ACCEPT) appeared inside capturing parentheses, previous releases of
1857        PCRE did not set those parentheses (unlike Perl). I have now found a way to
1858        make it do so. The string so far is captured, making this feature
1859        compatible with Perl.
1860    
1861    23. The tests have been re-organized, adding tests 11 and 12, to make it
1862        possible to check the Perl 5.10 features against Perl 5.10.
1863    
1864    24. Perl 5.10 allows subroutine calls in lookbehinds, as long as the subroutine
1865        pattern matches a fixed length string. PCRE did not allow this; now it
1866        does. Neither allows recursion.
1867    
1868    25. I finally figured out how to implement a request to provide the minimum
1869        length of subject string that was needed in order to match a given pattern.
1870        (It was back references and recursion that I had previously got hung up
1871        on.) This code has now been added to pcre_study(); it finds a lower bound
1872        to the length of subject needed. It is not necessarily the greatest lower
1873        bound, but using it to avoid searching strings that are too short does give
1874        some useful speed-ups. The value is available to calling programs via
1875        pcre_fullinfo().
1876    
1877    26. While implementing 25, I discovered to my embarrassment that pcretest had
1878        not been passing the result of pcre_study() to pcre_dfa_exec(), so the
1879        study optimizations had never been tested with that matching function.
1880        Oops. What is worse, even when it was passed study data, there was a bug in
1881        pcre_dfa_exec() that meant it never actually used it. Double oops. There
1882        were also very few tests of studied patterns with pcre_dfa_exec().
1883    
1884    27. If (?| is used to create subpatterns with duplicate numbers, they are now
1885        allowed to have the same name, even if PCRE_DUPNAMES is not set. However,
1886        on the other side of the coin, they are no longer allowed to have different
1887        names, because these cannot be distinguished in PCRE, and this has caused
1888        confusion. (This is a difference from Perl.)
1889    
1890    28. When duplicate subpattern names are present (necessarily with different
1891        numbers, as required by 27 above), and a test is made by name in a
1892        conditional pattern, either for a subpattern having been matched, or for
1893        recursion in such a pattern, all the associated numbered subpatterns are
1894        tested, and the overall condition is true if the condition is true for any
1895        one of them. This is the way Perl works, and is also more like the way
1896        testing by number works.
1897    
1898    
1899    Version 7.9 11-Apr-09
1900    ---------------------
1901    
1902    1.  When building with support for bzlib/zlib (pcregrep) and/or readline
1903        (pcretest), all targets were linked against these libraries. This included
1904        libpcre, libpcreposix, and libpcrecpp, even though they do not use these
1905        libraries. This caused unwanted dependencies to be created. This problem
1906        has been fixed, and now only pcregrep is linked with bzlib/zlib and only
1907        pcretest is linked with readline.
1908    
1909    2.  The "typedef int BOOL" in pcre_internal.h that was included inside the
1910        "#ifndef FALSE" condition by an earlier change (probably 7.8/18) has been
1911        moved outside it again, because FALSE and TRUE are already defined in AIX,
1912        but BOOL is not.
1913    
1914    3.  The pcre_config() function was treating the PCRE_MATCH_LIMIT and
1915        PCRE_MATCH_LIMIT_RECURSION values as ints, when they should be long ints.
1916    
1917    4.  The pcregrep documentation said spaces were inserted as well as colons (or
1918        hyphens) following file names and line numbers when outputting matching
1919        lines. This is not true; no spaces are inserted. I have also clarified the
1920        wording for the --colour (or --color) option.
1921    
1922    5.  In pcregrep, when --colour was used with -o, the list of matching strings
1923        was not coloured; this is different to GNU grep, so I have changed it to be
1924        the same.
1925    
1926    6.  When --colo(u)r was used in pcregrep, only the first matching substring in
1927        each matching line was coloured. Now it goes on to look for further matches
1928        of any of the test patterns, which is the same behaviour as GNU grep.
1929    
1930    7.  A pattern that could match an empty string could cause pcregrep to loop; it
1931        doesn't make sense to accept an empty string match in pcregrep, so I have
1932        locked it out (using PCRE's PCRE_NOTEMPTY option). By experiment, this
1933        seems to be how GNU grep behaves. [But see later change 40 for release
1934        8.33.]
1935    
1936    8.  The pattern (?(?=.*b)b|^) was incorrectly compiled as "match must be at
1937        start or after a newline", because the conditional assertion was not being
1938        correctly handled. The rule now is that both the assertion and what follows
1939        in the first alternative must satisfy the test.
1940    
1941    9.  If auto-callout was enabled in a pattern with a conditional group whose
1942        condition was an assertion, PCRE could crash during matching, both with
1943        pcre_exec() and pcre_dfa_exec().
1944    
1945    10. The PCRE_DOLLAR_ENDONLY option was not working when pcre_dfa_exec() was
1946        used for matching.
1947    
1948    11. Unicode property support in character classes was not working for
1949        characters (bytes) greater than 127 when not in UTF-8 mode.
1950    
1951    12. Added the -M command line option to pcretest.
1952    
1953    14. Added the non-standard REG_NOTEMPTY option to the POSIX interface.
1954    
1955    15. Added the PCRE_NO_START_OPTIMIZE match-time option.
1956    
1957    16. Added comments and documentation about mis-use of no_arg in the C++
1958        wrapper.
1959    
1960    17. Implemented support for UTF-8 encoding in EBCDIC environments, a patch
1961        from Martin Jerabek that uses macro names for all relevant character and
1962        string constants.
1963    
1964    18. Added to pcre_internal.h two configuration checks: (a) If both EBCDIC and
1965        SUPPORT_UTF8 are set, give an error; (b) If SUPPORT_UCP is set without
1966        SUPPORT_UTF8, define SUPPORT_UTF8. The "configure" script handles both of
1967        these, but not everybody uses configure.
1968    
1969    19. A conditional group that had only one branch was not being correctly
1970        recognized as an item that could match an empty string. This meant that an
1971        enclosing group might also not be so recognized, causing infinite looping
1972        (and probably a segfault) for patterns such as ^"((?(?=[a])[^"])|b)*"$
1973        with the subject "ab", where knowledge that the repeated group can match
1974        nothing is needed in order to break the loop.
1975    
1976    20. If a pattern that was compiled with callouts was matched using pcre_dfa_
1977        exec(), but without supplying a callout function, matching went wrong.
1978    
1979    21. If PCRE_ERROR_MATCHLIMIT occurred during a recursion, there was a memory
1980        leak if the size of the offset vector was greater than 30. When the vector
1981        is smaller, the saved offsets during recursion go onto a local stack
1982        vector, but for larger vectors malloc() is used. It was failing to free
1983        when the recursion yielded PCRE_ERROR_MATCH_LIMIT (or any other "abnormal"
1984        error, in fact).
1985    
1986    22. There was a missing #ifdef SUPPORT_UTF8 round one of the variables in the
1987        heapframe that is used only when UTF-8 support is enabled. This caused no
1988        problem, but was untidy.
1989    
1990    23. Steven Van Ingelgem's patch to CMakeLists.txt to change the name
1991        CMAKE_BINARY_DIR to PROJECT_BINARY_DIR so that it works when PCRE is
1992        included within another project.
1993    
1994    24. Steven Van Ingelgem's patches to add more options to the CMake support,
1995        slightly modified by me:
1996    
1997          (a) PCRE_BUILD_TESTS can be set OFF not to build the tests, including
1998              not building pcregrep.
1999    
2000          (b) PCRE_BUILD_PCREGREP can be see OFF not to build pcregrep, but only
2001              if PCRE_BUILD_TESTS is also set OFF, because the tests use pcregrep.
2002    
2003    25. Forward references, both numeric and by name, in patterns that made use of
2004        duplicate group numbers, could behave incorrectly or give incorrect errors,
2005        because when scanning forward to find the reference group, PCRE was not
2006        taking into account the duplicate group numbers. A pattern such as
2007        ^X(?3)(a)(?|(b)|(q))(Y) is an example.
2008    
2009    26. Changed a few more instances of "const unsigned char *" to USPTR, making
2010        the feature of a custom pointer more persuasive (as requested by a user).
2011    
2012    27. Wrapped the definitions of fileno and isatty for Windows, which appear in
2013        pcretest.c, inside #ifndefs, because it seems they are sometimes already
2014        pre-defined.
2015    
2016    28. Added support for (*UTF8) at the start of a pattern.
2017    
2018    29. Arrange for flags added by the "release type" setting in CMake to be shown
2019        in the configuration summary.
2020    
2021    
2022    Version 7.8 05-Sep-08
2023    ---------------------
2024    
2025    1.  Replaced UCP searching code with optimized version as implemented for Ad
2026        Muncher (http://www.admuncher.com/) by Peter Kankowski. This uses a two-
2027        stage table and inline lookup instead of a function, giving speed ups of 2
2028        to 5 times on some simple patterns that I tested. Permission was given to
2029        distribute the MultiStage2.py script that generates the tables (it's not in
2030        the tarball, but is in the Subversion repository).
2031    
2032    2.  Updated the Unicode datatables to Unicode 5.1.0. This adds yet more
2033        scripts.
2034    
2035    3.  Change 12 for 7.7 introduced a bug in pcre_study() when a pattern contained
2036        a group with a zero qualifier. The result of the study could be incorrect,
2037        or the function might crash, depending on the pattern.
2038    
2039    4.  Caseless matching was not working for non-ASCII characters in back
2040        references. For example, /(\x{de})\1/8i was not matching \x{de}\x{fe}.
2041        It now works when Unicode Property Support is available.
2042    
2043    5.  In pcretest, an escape such as \x{de} in the data was always generating
2044        a UTF-8 string, even in non-UTF-8 mode. Now it generates a single byte in
2045        non-UTF-8 mode. If the value is greater than 255, it gives a warning about
2046        truncation.
2047    
2048    6.  Minor bugfix in pcrecpp.cc (change "" == ... to NULL == ...).
2049    
2050    7.  Added two (int) casts to pcregrep when printing the difference of two
2051        pointers, in case they are 64-bit values.
2052    
2053    8.  Added comments about Mac OS X stack usage to the pcrestack man page and to
2054        test 2 if it fails.
2055    
2056    9.  Added PCRE_CALL_CONVENTION just before the names of all exported functions,
2057        and a #define of that name to empty if it is not externally set. This is to
2058        allow users of MSVC to set it if necessary.
2059    
2060    10. The PCRE_EXP_DEFN macro which precedes exported functions was missing from
2061        the convenience functions in the pcre_get.c source file.
2062    
2063    11. An option change at the start of a pattern that had top-level alternatives
2064        could cause overwriting and/or a crash. This command provoked a crash in
2065        some environments:
2066    
2067          printf "/(?i)[\xc3\xa9\xc3\xbd]|[\xc3\xa9\xc3\xbdA]/8\n" | pcretest
2068    
2069        This potential security problem was recorded as CVE-2008-2371.
2070    
2071    12. For a pattern where the match had to start at the beginning or immediately
2072        after a newline (e.g /.*anything/ without the DOTALL flag), pcre_exec() and
2073        pcre_dfa_exec() could read past the end of the passed subject if there was
2074        no match. To help with detecting such bugs (e.g. with valgrind), I modified
2075        pcretest so that it places the subject at the end of its malloc-ed buffer.
2076    
2077    13. The change to pcretest in 12 above threw up a couple more cases when pcre_
2078        exec() might read past the end of the data buffer in UTF-8 mode.
2079    
2080    14. A similar bug to 7.3/2 existed when the PCRE_FIRSTLINE option was set and
2081        the data contained the byte 0x85 as part of a UTF-8 character within its
2082        first line. This applied both to normal and DFA matching.
2083    
2084    15. Lazy qualifiers were not working in some cases in UTF-8 mode. For example,
2085        /^[^d]*?$/8 failed to match "abc".
2086    
2087    16. Added a missing copyright notice to pcrecpp_internal.h.
2088    
2089    17. Make it more clear in the documentation that values returned from
2090        pcre_exec() in ovector are byte offsets, not character counts.
2091    
2092    18. Tidied a few places to stop certain compilers from issuing warnings.
2093    
2094    19. Updated the Virtual Pascal + BCC files to compile the latest v7.7, as
2095        supplied by Stefan Weber. I made a further small update for 7.8 because
2096        there is a change of source arrangements: the pcre_searchfuncs.c module is
2097        replaced by pcre_ucd.c.
2098    
2099    
2100    Version 7.7 07-May-08
2101    ---------------------
2102    
2103    1.  Applied Craig's patch to sort out a long long problem: "If we can't convert
2104        a string to a long long, pretend we don't even have a long long." This is
2105        done by checking for the strtoq, strtoll, and _strtoi64 functions.
2106    
2107    2.  Applied Craig's patch to pcrecpp.cc to restore ABI compatibility with
2108        pre-7.6 versions, which defined a global no_arg variable instead of putting
2109        it in the RE class. (See also #8 below.)
2110    
2111    3.  Remove a line of dead code, identified by coverity and reported by Nuno
2112        Lopes.
2113    
2114    4.  Fixed two related pcregrep bugs involving -r with --include or --exclude:
2115    
2116        (1) The include/exclude patterns were being applied to the whole pathnames
2117            of files, instead of just to the final components.
2118    
2119        (2) If there was more than one level of directory, the subdirectories were
2120            skipped unless they satisfied the include/exclude conditions. This is
2121            inconsistent with GNU grep (and could even be seen as contrary to the
2122            pcregrep specification - which I improved to make it absolutely clear).
2123            The action now is always to scan all levels of directory, and just
2124            apply the include/exclude patterns to regular files.
2125    
2126    5.  Added the --include_dir and --exclude_dir patterns to pcregrep, and used
2127        --exclude_dir in the tests to avoid scanning .svn directories.
2128    
2129    6.  Applied Craig's patch to the QuoteMeta function so that it escapes the
2130        NUL character as backslash + 0 rather than backslash + NUL, because PCRE
2131        doesn't support NULs in patterns.
2132    
2133    7.  Added some missing "const"s to declarations of static tables in
2134        pcre_compile.c and pcre_dfa_exec.c.
2135    
2136    8.  Applied Craig's patch to pcrecpp.cc to fix a problem in OS X that was
2137        caused by fix #2  above. (Subsequently also a second patch to fix the
2138        first patch. And a third patch - this was a messy problem.)
2139    
2140    9.  Applied Craig's patch to remove the use of push_back().
2141    
2142    10. Applied Alan Lehotsky's patch to add REG_STARTEND support to the POSIX
2143        matching function regexec().
2144    
2145    11. Added support for the Oniguruma syntax \g<name>, \g<n>, \g'name', \g'n',
2146        which, however, unlike Perl's \g{...}, are subroutine calls, not back
2147        references. PCRE supports relative numbers with this syntax (I don't think
2148        Oniguruma does).
2149    
2150    12. Previously, a group with a zero repeat such as (...){0} was completely
2151        omitted from the compiled regex. However, this means that if the group
2152        was called as a subroutine from elsewhere in the pattern, things went wrong
2153        (an internal error was given). Such groups are now left in the compiled
2154        pattern, with a new opcode that causes them to be skipped at execution
2155        time.
2156    
2157    13. Added the PCRE_JAVASCRIPT_COMPAT option. This makes the following changes
2158        to the way PCRE behaves:
2159    
2160        (a) A lone ] character is dis-allowed (Perl treats it as data).
2161    
2162        (b) A back reference to an unmatched subpattern matches an empty string
2163            (Perl fails the current match path).
2164    
2165        (c) A data ] in a character class must be notated as \] because if the
2166            first data character in a class is ], it defines an empty class. (In
2167            Perl it is not possible to have an empty class.) The empty class []
2168            never matches; it forces failure and is equivalent to (*FAIL) or (?!).
2169            The negative empty class [^] matches any one character, independently
2170            of the DOTALL setting.
2171    
2172    14. A pattern such as /(?2)[]a()b](abc)/ which had a forward reference to a
2173        non-existent subpattern following a character class starting with ']' and
2174        containing () gave an internal compiling error instead of "reference to
2175        non-existent subpattern". Fortunately, when the pattern did exist, the
2176        compiled code was correct. (When scanning forwards to check for the
2177        existence of the subpattern, it was treating the data ']' as terminating
2178        the class, so got the count wrong. When actually compiling, the reference
2179        was subsequently set up correctly.)
2180    
2181    15. The "always fail" assertion (?!) is optimzed to (*FAIL) by pcre_compile;
2182        it was being rejected as not supported by pcre_dfa_exec(), even though
2183        other assertions are supported. I have made pcre_dfa_exec() support
2184        (*FAIL).
2185    
2186    16. The implementation of 13c above involved the invention of a new opcode,
2187        OP_ALLANY, which is like OP_ANY but doesn't check the /s flag. Since /s
2188        cannot be changed at match time, I realized I could make a small
2189        improvement to matching performance by compiling OP_ALLANY instead of
2190        OP_ANY for "." when DOTALL was set, and then removing the runtime tests
2191        on the OP_ANY path.
2192    
2193    17. Compiling pcretest on Windows with readline support failed without the
2194        following two fixes: (1) Make the unistd.h include conditional on
2195        HAVE_UNISTD_H; (2) #define isatty and fileno as _isatty and _fileno.
2196    
2197    18. Changed CMakeLists.txt and cmake/FindReadline.cmake to arrange for the
2198        ncurses library to be included for pcretest when ReadLine support is
2199        requested, but also to allow for it to be overridden. This patch came from
2200        Daniel Bergström.
2201    
2202    19. There was a typo in the file ucpinternal.h where f0_rangeflag was defined
2203        as 0x00f00000 instead of 0x00800000. Luckily, this would not have caused
2204        any errors with the current Unicode tables. Thanks to Peter Kankowski for
2205        spotting this.
2206    
2207    
2208    Version 7.6 28-Jan-08
2209    ---------------------
2210    
2211    1.  A character class containing a very large number of characters with
2212        codepoints greater than 255 (in UTF-8 mode, of course) caused a buffer
2213        overflow.
2214    
2215    2.  Patch to cut out the "long long" test in pcrecpp_unittest when
2216        HAVE_LONG_LONG is not defined.
2217    
2218    3.  Applied Christian Ehrlicher's patch to update the CMake build files to
2219        bring them up to date and include new features. This patch includes:
2220    
2221        - Fixed PH's badly added libz and libbz2 support.
2222        - Fixed a problem with static linking.
2223        - Added pcredemo. [But later removed - see 7 below.]
2224        - Fixed dftables problem and added an option.
2225        - Added a number of HAVE_XXX tests, including HAVE_WINDOWS_H and
2226            HAVE_LONG_LONG.
2227        - Added readline support for pcretest.
2228        - Added an listing of the option settings after cmake has run.
2229    
2230    4.  A user submitted a patch to Makefile that makes it easy to create
2231        "pcre.dll" under mingw when using Configure/Make. I added stuff to
2232        Makefile.am that cause it to include this special target, without
2233        affecting anything else. Note that the same mingw target plus all
2234        the other distribution libraries and programs are now supported
2235        when configuring with CMake (see 6 below) instead of with
2236        Configure/Make.
2237    
2238    5.  Applied Craig's patch that moves no_arg into the RE class in the C++ code.
2239        This is an attempt to solve the reported problem "pcrecpp::no_arg is not
2240        exported in the Windows port". It has not yet been confirmed that the patch
2241        solves the problem, but it does no harm.
2242    
2243    6.  Applied Sheri's patch to CMakeLists.txt to add NON_STANDARD_LIB_PREFIX and
2244        NON_STANDARD_LIB_SUFFIX for dll names built with mingw when configured
2245        with CMake, and also correct the comment about stack recursion.
2246    
2247    7.  Remove the automatic building of pcredemo from the ./configure system and
2248        from CMakeLists.txt. The whole idea of pcredemo.c is that it is an example
2249        of a program that users should build themselves after PCRE is installed, so
2250        building it automatically is not really right. What is more, it gave
2251        trouble in some build environments.
2252    
2253    8.  Further tidies to CMakeLists.txt from Sheri and Christian.
2254    
2255    
2256    Version 7.5 10-Jan-08
2257    ---------------------
2258    
2259    1.  Applied a patch from Craig: "This patch makes it possible to 'ignore'
2260        values in parens when parsing an RE using the C++ wrapper."
2261    
2262    2.  Negative specials like \S did not work in character classes in UTF-8 mode.
2263        Characters greater than 255 were excluded from the class instead of being
2264        included.
2265    
2266    3.  The same bug as (2) above applied to negated POSIX classes such as
2267        [:^space:].
2268    
2269    4.  PCRECPP_STATIC was referenced in pcrecpp_internal.h, but nowhere was it
2270        defined or documented. It seems to have been a typo for PCRE_STATIC, so
2271        I have changed it.
2272    
2273    5.  The construct (?&) was not diagnosed as a syntax error (it referenced the
2274        first named subpattern) and a construct such as (?&a) would reference the
2275        first named subpattern whose name started with "a" (in other words, the
2276        length check was missing). Both these problems are fixed. "Subpattern name
2277        expected" is now given for (?&) (a zero-length name), and this patch also
2278        makes it give the same error for \k'' (previously it complained that that
2279        was a reference to a non-existent subpattern).
2280    
2281    6.  The erroneous patterns (?+-a) and (?-+a) give different error messages;
2282        this is right because (?- can be followed by option settings as well as by
2283        digits. I have, however, made the messages clearer.
2284    
2285    7.  Patterns such as (?(1)a|b) (a pattern that contains fewer subpatterns
2286        than the number used in the conditional) now cause a compile-time error.
2287        This is actually not compatible with Perl, which accepts such patterns, but
2288        treats the conditional as always being FALSE (as PCRE used to), but it
2289        seems to me that giving a diagnostic is better.
2290    
2291    8.  Change "alphameric" to the more common word "alphanumeric" in comments
2292        and messages.
2293    
2294    9.  Fix two occurrences of "backslash" in comments that should have been
2295        "backspace".
2296    
2297    10. Remove two redundant lines of code that can never be obeyed (their function
2298        was moved elsewhere).
2299    
2300    11. The program that makes PCRE's Unicode character property table had a bug
2301        which caused it to generate incorrect table entries for sequences of
2302        characters that have the same character type, but are in different scripts.
2303        It amalgamated them into a single range, with the script of the first of
2304        them. In other words, some characters were in the wrong script. There were
2305        thirteen such cases, affecting characters in the following ranges:
2306    
2307          U+002b0 - U+002c1
2308          U+0060c - U+0060d
2309          U+0061e - U+00612
2310          U+0064b - U+0065e
2311          U+0074d - U+0076d
2312          U+01800 - U+01805
2313          U+01d00 - U+01d77
2314          U+01d9b - U+01dbf
2315          U+0200b - U+0200f
2316          U+030fc - U+030fe
2317          U+03260 - U+0327f
2318          U+0fb46 - U+0fbb1
2319          U+10450 - U+1049d
2320    
2321    12. The -o option (show only the matching part of a line) for pcregrep was not
2322        compatible with GNU grep in that, if there was more than one match in a
2323        line, it showed only the first of them. It now behaves in the same way as
2324        GNU grep.
2325    
2326    13. If the -o and -v options were combined for pcregrep, it printed a blank
2327        line for every non-matching line. GNU grep prints nothing, and pcregrep now
2328        does the same. The return code can be used to tell if there were any
2329        non-matching lines.
2330    
2331    14. Added --file-offsets and --line-offsets to pcregrep.
2332    
2333    15. The pattern (?=something)(?R) was not being diagnosed as a potentially
2334        infinitely looping recursion. The bug was that positive lookaheads were not
2335        being skipped when checking for a possible empty match (negative lookaheads
2336        and both kinds of lookbehind were skipped).
2337    
2338    16. Fixed two typos in the Windows-only code in pcregrep.c, and moved the
2339        inclusion of <windows.h> to before rather than after the definition of
2340        INVALID_FILE_ATTRIBUTES (patch from David Byron).
2341    
2342    17. Specifying a possessive quantifier with a specific limit for a Unicode
2343        character property caused pcre_compile() to compile bad code, which led at
2344        runtime to PCRE_ERROR_INTERNAL (-14). Examples of patterns that caused this
2345        are: /\p{Zl}{2,3}+/8 and /\p{Cc}{2}+/8. It was the possessive "+" that
2346        caused the error; without that there was no problem.
2347    
2348    18. Added --enable-pcregrep-libz and --enable-pcregrep-libbz2.
2349    
2350    19. Added --enable-pcretest-libreadline.
2351    
2352    20. In pcrecpp.cc, the variable 'count' was incremented twice in
2353        RE::GlobalReplace(). As a result, the number of replacements returned was
2354        double what it should be. I removed one of the increments, but Craig sent a
2355        later patch that removed the other one (the right fix) and added unit tests
2356        that check the return values (which was not done before).
2357    
2358    21. Several CMake things:
2359    
2360        (1) Arranged that, when cmake is used on Unix, the libraries end up with
2361            the names libpcre and libpcreposix, not just pcre and pcreposix.
2362    
2363        (2) The above change means that pcretest and pcregrep are now correctly
2364            linked with the newly-built libraries, not previously installed ones.
2365    
2366        (3) Added PCRE_SUPPORT_LIBREADLINE, PCRE_SUPPORT_LIBZ, PCRE_SUPPORT_LIBBZ2.
2367    
2368    22. In UTF-8 mode, with newline set to "any", a pattern such as .*a.*=.b.*
2369        crashed when matching a string such as a\x{2029}b (note that \x{2029} is a
2370        UTF-8 newline character). The key issue is that the pattern starts .*;
2371        this means that the match must be either at the beginning, or after a
2372        newline. The bug was in the code for advancing after a failed match and
2373        checking that the new position followed a newline. It was not taking
2374        account of UTF-8 characters correctly.
2375    
2376    23. PCRE was behaving differently from Perl in the way it recognized POSIX
2377        character classes. PCRE was not treating the sequence [:...:] as a
2378        character class unless the ... were all letters. Perl, however, seems to
2379        allow any characters between [: and :], though of course it rejects as
2380        unknown any "names" that contain non-letters, because all the known class
2381        names consist only of letters. Thus, Perl gives an error for [[:1234:]],
2382        for example, whereas PCRE did not - it did not recognize a POSIX character
2383        class. This seemed a bit dangerous, so the code has been changed to be
2384        closer to Perl. The behaviour is not identical to Perl, because PCRE will
2385        diagnose an unknown class for, for example, [[:l\ower:]] where Perl will
2386        treat it as [[:lower:]]. However, PCRE does now give "unknown" errors where
2387        Perl does, and where it didn't before.
2388    
2389    24. Rewrite so as to remove the single use of %n from pcregrep because in some
2390        Windows environments %n is disabled by default.
2391    
2392    
2393    Version 7.4 21-Sep-07
2394    ---------------------
2395    
2396    1.  Change 7.3/28 was implemented for classes by looking at the bitmap. This
2397        means that a class such as [\s] counted as "explicit reference to CR or
2398        LF". That isn't really right - the whole point of the change was to try to
2399        help when there was an actual mention of one of the two characters. So now
2400        the change happens only if \r or \n (or a literal CR or LF) character is
2401        encountered.
2402    
2403    2.  The 32-bit options word was also used for 6 internal flags, but the numbers
2404        of both had grown to the point where there were only 3 bits left.
2405        Fortunately, there was spare space in the data structure, and so I have
2406        moved the internal flags into a new 16-bit field to free up more option
2407        bits.
2408    
2409    3.  The appearance of (?J) at the start of a pattern set the DUPNAMES option,
2410        but did not set the internal JCHANGED flag - either of these is enough to
2411        control the way the "get" function works - but the PCRE_INFO_JCHANGED
2412        facility is supposed to tell if (?J) was ever used, so now (?J) at the
2413        start sets both bits.
2414    
2415    4.  Added options (at build time, compile time, exec time) to change \R from
2416        matching any Unicode line ending sequence to just matching CR, LF, or CRLF.
2417    
2418    5.  doc/pcresyntax.html was missing from the distribution.
2419    
2420    6.  Put back the definition of PCRE_ERROR_NULLWSLIMIT, for backward
2421        compatibility, even though it is no longer used.
2422    
2423    7.  Added macro for snprintf to pcrecpp_unittest.cc and also for strtoll and
2424        strtoull to pcrecpp.cc to select the available functions in WIN32 when the
2425        windows.h file is present (where different names are used). [This was
2426        reversed later after testing - see 16 below.]
2427    
2428    8.  Changed all #include <config.h> to #include "config.h". There were also
2429        some further <pcre.h> cases that I changed to "pcre.h".
2430    
2431    9.  When pcregrep was used with the --colour option, it missed the line ending
2432        sequence off the lines that it output.
2433    
2434    10. It was pointed out to me that arrays of string pointers cause lots of
2435        relocations when a shared library is dynamically loaded. A technique of
2436        using a single long string with a table of offsets can drastically reduce
2437        these. I have refactored PCRE in four places to do this. The result is
2438        dramatic:
2439    
2440          Originally:                          290
2441          After changing UCP table:            187
2442          After changing error message table:   43
2443          After changing table of "verbs"       36
2444          After changing table of Posix names   22
2445    
2446        Thanks to the folks working on Gregex for glib for this insight.
2447    
2448    11. --disable-stack-for-recursion caused compiling to fail unless -enable-
2449        unicode-properties was also set.
2450    
2451    12. Updated the tests so that they work when \R is defaulted to ANYCRLF.
2452    
2453    13. Added checks for ANY and ANYCRLF to pcrecpp.cc where it previously
2454        checked only for CRLF.
2455    
2456    14. Added casts to pcretest.c to avoid compiler warnings.
2457    
2458    15. Added Craig's patch to various pcrecpp modules to avoid compiler warnings.
2459    
2460    16. Added Craig's patch to remove the WINDOWS_H tests, that were not working,
2461        and instead check for _strtoi64 explicitly, and avoid the use of snprintf()
2462        entirely. This removes changes made in 7 above.
2463    
2464    17. The CMake files have been updated, and there is now more information about
2465        building with CMake in the NON-UNIX-USE document.
2466    
2467    
2468    Version 7.3 28-Aug-07
2469    ---------------------
2470    
2471     1. In the rejigging of the build system that eventually resulted in 7.1, the
2472        line "#include <pcre.h>" was included in pcre_internal.h. The use of angle
2473        brackets there is not right, since it causes compilers to look for an
2474        installed pcre.h, not the version that is in the source that is being
2475        compiled (which of course may be different). I have changed it back to:
2476    
2477          #include "pcre.h"
2478    
2479        I have a vague recollection that the change was concerned with compiling in
2480        different directories, but in the new build system, that is taken care of
2481        by the VPATH setting the Makefile.
2482    
2483     2. The pattern .*$ when run in not-DOTALL UTF-8 mode with newline=any failed
2484        when the subject happened to end in the byte 0x85 (e.g. if the last
2485        character was \x{1ec5}). *Character* 0x85 is one of the "any" newline
2486        characters but of course it shouldn't be taken as a newline when it is part
2487        of another character. The bug was that, for an unlimited repeat of . in
2488        not-DOTALL UTF-8 mode, PCRE was advancing by bytes rather than by
2489        characters when looking for a newline.
2490    
2491     3. A small performance improvement in the DOTALL UTF-8 mode .* case.
2492    
2493     4. Debugging: adjusted the names of opcodes for different kinds of parentheses
2494        in debug output.
2495    
2496     5. Arrange to use "%I64d" instead of "%lld" and "%I64u" instead of "%llu" for
2497        long printing in the pcrecpp unittest when running under MinGW.
2498    
2499     6. ESC_K was left out of the EBCDIC table.
2500    
2501     7. Change 7.0/38 introduced a new limit on the number of nested non-capturing
2502        parentheses; I made it 1000, which seemed large enough. Unfortunately, the
2503        limit also applies to "virtual nesting" when a pattern is recursive, and in
2504        this case 1000 isn't so big. I have been able to remove this limit at the
2505        expense of backing off one optimization in certain circumstances. Normally,
2506        when pcre_exec() would call its internal match() function recursively and
2507        immediately return the result unconditionally, it uses a "tail recursion"
2508        feature to save stack. However, when a subpattern that can match an empty
2509        string has an unlimited repetition quantifier, it no longer makes this
2510        optimization. That gives it a stack frame in which to save the data for
2511        checking that an empty string has been matched. Previously this was taken
2512        from the 1000-entry workspace that had been reserved. So now there is no
2513        explicit limit, but more stack is used.
2514    
2515     8. Applied Daniel's patches to solve problems with the import/export magic
2516        syntax that is required for Windows, and which was going wrong for the
2517        pcreposix and pcrecpp parts of the library. These were overlooked when this
2518        problem was solved for the main library.
2519    
2520     9. There were some crude static tests to avoid integer overflow when computing
2521        the size of patterns that contain repeated groups with explicit upper
2522        limits. As the maximum quantifier is 65535, the maximum group length was
2523        set at 30,000 so that the product of these two numbers did not overflow a
2524        32-bit integer. However, it turns out that people want to use groups that
2525        are longer than 30,000 bytes (though not repeat them that many times).
2526        Change 7.0/17 (the refactoring of the way the pattern size is computed) has
2527        made it possible to implement the integer overflow checks in a much more
2528        dynamic way, which I have now done. The artificial limitation on group
2529        length has been removed - we now have only the limit on the total length of
2530        the compiled pattern, which depends on the LINK_SIZE setting.
2531    
2532    10. Fixed a bug in the documentation for get/copy named substring when
2533        duplicate names are permitted. If none of the named substrings are set, the
2534        functions return PCRE_ERROR_NOSUBSTRING (7); the doc said they returned an
2535        empty string.
2536    
2537    11. Because Perl interprets \Q...\E at a high level, and ignores orphan \E
2538        instances, patterns such as [\Q\E] or [\E] or even [^\E] cause an error,
2539        because the ] is interpreted as the first data character and the
2540        terminating ] is not found. PCRE has been made compatible with Perl in this
2541        regard. Previously, it interpreted [\Q\E] as an empty class, and [\E] could
2542        cause memory overwriting.
2543    
2544    10. Like Perl, PCRE automatically breaks an unlimited repeat after an empty
2545        string has been matched (to stop an infinite loop). It was not recognizing
2546        a conditional subpattern that could match an empty string if that
2547        subpattern was within another subpattern. For example, it looped when
2548        trying to match  (((?(1)X|))*)  but it was OK with  ((?(1)X|)*)  where the
2549        condition was not nested. This bug has been fixed.
2550    
2551    12. A pattern like \X?\d or \P{L}?\d in non-UTF-8 mode could cause a backtrack
2552        past the start of the subject in the presence of bytes with the top bit
2553        set, for example "\x8aBCD".
2554    
2555    13. Added Perl 5.10 experimental backtracking controls (*FAIL), (*F), (*PRUNE),
2556        (*SKIP), (*THEN), (*COMMIT), and (*ACCEPT).
2557    
2558    14. Optimized (?!) to (*FAIL).
2559    
2560    15. Updated the test for a valid UTF-8 string to conform to the later RFC 3629.
2561        This restricts code points to be within the range 0 to 0x10FFFF, excluding
2562        the "low surrogate" sequence 0xD800 to 0xDFFF. Previously, PCRE allowed the
2563        full range 0 to 0x7FFFFFFF, as defined by RFC 2279. Internally, it still
2564        does: it's just the validity check that is more restrictive.
2565    
2566    16. Inserted checks for integer overflows during escape sequence (backslash)
2567        processing, and also fixed erroneous offset values for syntax errors during
2568        backslash processing.
2569    
2570    17. Fixed another case of looking too far back in non-UTF-8 mode (cf 12 above)
2571        for patterns like [\PPP\x8a]{1,}\x80 with the subject "A\x80".
2572    
2573    18. An unterminated class in a pattern like (?1)\c[ with a "forward reference"
2574        caused an overrun.
2575    
2576    19. A pattern like (?:[\PPa*]*){8,} which had an "extended class" (one with
2577        something other than just ASCII characters) inside a group that had an
2578        unlimited repeat caused a loop at compile time (while checking to see
2579        whether the group could match an empty string).
2580    
2581    20. Debugging a pattern containing \p or \P could cause a crash. For example,
2582        [\P{Any}] did so. (Error in the code for printing property names.)
2583    
2584    21. An orphan \E inside a character class could cause a crash.
2585    
2586    22. A repeated capturing bracket such as (A)? could cause a wild memory
2587        reference during compilation.
2588    
2589    23. There are several functions in pcre_compile() that scan along a compiled
2590        expression for various reasons (e.g. to see if it's fixed length for look
2591        behind). There were bugs in these functions when a repeated \p or \P was
2592        present in the pattern. These operators have additional parameters compared
2593        with \d, etc, and these were not being taken into account when moving along
2594        the compiled data. Specifically:
2595    
2596        (a) A item such as \p{Yi}{3} in a lookbehind was not treated as fixed
2597            length.
2598    
2599        (b) An item such as \pL+ within a repeated group could cause crashes or
2600            loops.
2601    
2602        (c) A pattern such as \p{Yi}+(\P{Yi}+)(?1) could give an incorrect
2603            "reference to non-existent subpattern" error.
2604    
2605        (d) A pattern like (\P{Yi}{2}\277)? could loop at compile time.
2606    
2607    24. A repeated \S or \W in UTF-8 mode could give wrong answers when multibyte
2608        characters were involved (for example /\S{2}/8g with "A\x{a3}BC").
2609    
2610    25. Using pcregrep in multiline, inverted mode (-Mv) caused it to loop.
2611    
2612    26. Patterns such as [\P{Yi}A] which include \p or \P and just one other
2613        character were causing crashes (broken optimization).
2614    
2615    27. Patterns such as (\P{Yi}*\277)* (group with possible zero repeat containing
2616        \p or \P) caused a compile-time loop.
2617    
2618    28. More problems have arisen in unanchored patterns when CRLF is a valid line
2619        break. For example, the unstudied pattern [\r\n]A does not match the string
2620        "\r\nA" because change 7.0/46 below moves the current point on by two
2621        characters after failing to match at the start. However, the pattern \nA
2622        *does* match, because it doesn't start till \n, and if [\r\n]A is studied,
2623        the same is true. There doesn't seem any very clean way out of this, but
2624        what I have chosen to do makes the common cases work: PCRE now takes note
2625        of whether there can be an explicit match for \r or \n anywhere in the
2626        pattern, and if so, 7.0/46 no longer applies. As part of this change,
2627        there's a new PCRE_INFO_HASCRORLF option for finding out whether a compiled
2628        pattern has explicit CR or LF references.
2629    
2630    29. Added (*CR) etc for changing newline setting at start of pattern.
2631    
2632    
2633    Version 7.2 19-Jun-07
2634    ---------------------
2635    
2636     1. If the fr_FR locale cannot be found for test 3, try the "french" locale,
2637        which is apparently normally available under Windows.
2638    
2639     2. Re-jig the pcregrep tests with different newline settings in an attempt
2640        to make them independent of the local environment's newline setting.
2641    
2642     3. Add code to configure.ac to remove -g from the CFLAGS default settings.
2643    
2644     4. Some of the "internals" tests were previously cut out when the link size
2645        was not 2, because the output contained actual offsets. The recent new
2646        "Z" feature of pcretest means that these can be cut out, making the tests
2647        usable with all link sizes.
2648    
2649     5. Implemented Stan Switzer's goto replacement for longjmp() when not using
2650        stack recursion. This gives a massive performance boost under BSD, but just
2651        a small improvement under Linux. However, it saves one field in the frame
2652        in all cases.
2653    
2654     6. Added more features from the forthcoming Perl 5.10:
2655    
2656        (a) (?-n) (where n is a string of digits) is a relative subroutine or
2657            recursion call. It refers to the nth most recently opened parentheses.
2658    
2659        (b) (?+n) is also a relative subroutine call; it refers to the nth next
2660            to be opened parentheses.
2661    
2662        (c) Conditions that refer to capturing parentheses can be specified
2663            relatively, for example, (?(-2)... or (?(+3)...
2664    
2665        (d) \K resets the start of the current match so that everything before
2666            is not part of it.
2667    
2668        (e) \k{name} is synonymous with \k<name> and \k'name' (.NET compatible).
2669    
2670        (f) \g{name} is another synonym - part of Perl 5.10's unification of
2671            reference syntax.
2672    
2673        (g) (?| introduces a group in which the numbering of parentheses in each
2674            alternative starts with the same number.
2675    
2676        (h) \h, \H, \v, and \V match horizontal and vertical whitespace.
2677    
2678     7. Added two new calls to pcre_fullinfo(): PCRE_INFO_OKPARTIAL and
2679        PCRE_INFO_JCHANGED.
2680    
2681     8. A pattern such as  (.*(.)?)*  caused pcre_exec() to fail by either not
2682        terminating or by crashing. Diagnosed by Viktor Griph; it was in the code
2683        for detecting groups that can match an empty string.
2684    
2685     9. A pattern with a very large number of alternatives (more than several
2686        hundred) was running out of internal workspace during the pre-compile
2687        phase, where pcre_compile() figures out how much memory will be needed. A
2688        bit of new cunning has reduced the workspace needed for groups with
2689        alternatives. The 1000-alternative test pattern now uses 12 bytes of
2690        workspace instead of running out of the 4096 that are available.
2691    
2692    10. Inserted some missing (unsigned int) casts to get rid of compiler warnings.
2693    
2694    11. Applied patch from Google to remove an optimization that didn't quite work.
2695        The report of the bug said:
2696    
2697          pcrecpp::RE("a*").FullMatch("aaa") matches, while
2698          pcrecpp::RE("a*?").FullMatch("aaa") does not, and
2699          pcrecpp::RE("a*?\\z").FullMatch("aaa") does again.
2700    
2701    12. If \p or \P was used in non-UTF-8 mode on a character greater than 127
2702        it matched the wrong number of bytes.
2703    
2704    
2705    Version 7.1 24-Apr-07
2706    ---------------------
2707    
2708     1. Applied Bob Rossi and Daniel G's patches to convert the build system to one
2709        that is more "standard", making use of automake and other Autotools. There
2710        is some re-arrangement of the files and adjustment of comments consequent
2711        on this.
2712    
2713     2. Part of the patch fixed a problem with the pcregrep tests. The test of -r
2714        for recursive directory scanning broke on some systems because the files
2715        are not scanned in any specific order and on different systems the order
2716        was different. A call to "sort" has been inserted into RunGrepTest for the
2717        approprate test as a short-term fix. In the longer term there may be an
2718        alternative.
2719    
2720     3. I had an email from Eric Raymond about problems translating some of PCRE's
2721        man pages to HTML (despite the fact that I distribute HTML pages, some
2722        people do their own conversions for various reasons). The problems
2723        concerned the use of low-level troff macros .br and .in. I have therefore
2724        removed all such uses from the man pages (some were redundant, some could
2725        be replaced by .nf/.fi pairs). The 132html script that I use to generate
2726        HTML has been updated to handle .nf/.fi and to complain if it encounters
2727        .br or .in.
2728    
2729     4. Updated comments in configure.ac that get placed in config.h.in and also
2730        arranged for config.h to be included in the distribution, with the name
2731        config.h.generic, for the benefit of those who have to compile without
2732        Autotools (compare pcre.h, which is now distributed as pcre.h.generic).
2733    
2734     5. Updated the support (such as it is) for Virtual Pascal, thanks to Stefan
2735        Weber: (1) pcre_internal.h was missing some function renames; (2) updated
2736        makevp.bat for the current PCRE, using the additional files
2737        makevp_c.txt, makevp_l.txt, and pcregexp.pas.
2738    
2739     6. A Windows user reported a minor discrepancy with test 2, which turned out
2740        to be caused by a trailing space on an input line that had got lost in his
2741        copy. The trailing space was an accident, so I've just removed it.
2742    
2743     7. Add -Wl,-R... flags in pcre-config.in for *BSD* systems, as I'm told
2744        that is needed.
2745    
2746     8. Mark ucp_table (in ucptable.h) and ucp_gentype (in pcre_ucp_searchfuncs.c)
2747        as "const" (a) because they are and (b) because it helps the PHP
2748        maintainers who have recently made a script to detect big data structures
2749        in the php code that should be moved to the .rodata section. I remembered
2750        to update Builducptable as well, so it won't revert if ucptable.h is ever
2751        re-created.
2752    
2753     9. Added some extra #ifdef SUPPORT_UTF8 conditionals into pcretest.c,
2754        pcre_printint.src, pcre_compile.c, pcre_study.c, and pcre_tables.c, in
2755        order to be able to cut out the UTF-8 tables in the latter when UTF-8
2756        support is not required. This saves 1.5-2K of code, which is important in
2757        some applications.
2758    
2759        Later: more #ifdefs are needed in pcre_ord2utf8.c and pcre_valid_utf8.c
2760        so as not to refer to the tables, even though these functions will never be
2761        called when UTF-8 support is disabled. Otherwise there are problems with a
2762        shared library.
2763    
2764    10. Fixed two bugs in the emulated memmove() function in pcre_internal.h:
2765    
2766        (a) It was defining its arguments as char * instead of void *.
2767    
2768        (b) It was assuming that all moves were upwards in memory; this was true
2769            a long time ago when I wrote it, but is no longer the case.
2770    
2771        The emulated memove() is provided for those environments that have neither
2772        memmove() nor bcopy(). I didn't think anyone used it these days, but that
2773        is clearly not the case, as these two bugs were recently reported.
2774    
2775    11. The script PrepareRelease is now distributed: it calls 132html, CleanTxt,
2776        and Detrail to create the HTML documentation, the .txt form of the man
2777        pages, and it removes trailing spaces from listed files. It also creates
2778        pcre.h.generic and config.h.generic from pcre.h and config.h. In the latter
2779        case, it wraps all the #defines with #ifndefs. This script should be run
2780        before "make dist".
2781    
2782    12. Fixed two fairly obscure bugs concerned with quantified caseless matching
2783        with Unicode property support.
2784    
2785        (a) For a maximizing quantifier, if the two different cases of the
2786            character were of different lengths in their UTF-8 codings (there are
2787            some cases like this - I found 11), and the matching function had to
2788            back up over a mixture of the two cases, it incorrectly assumed they
2789            were both the same length.
2790    
2791        (b) When PCRE was configured to use the heap rather than the stack for
2792            recursion during matching, it was not correctly preserving the data for
2793            the other case of a UTF-8 character when checking ahead for a match
2794            while processing a minimizing repeat. If the check also involved
2795            matching a wide character, but failed, corruption could cause an
2796            erroneous result when trying to check for a repeat of the original
2797            character.
2798    
2799    13. Some tidying changes to the testing mechanism:
2800    
2801        (a) The RunTest script now detects the internal link size and whether there
2802            is UTF-8 and UCP support by running ./pcretest -C instead of relying on
2803            values substituted by "configure". (The RunGrepTest script already did
2804            this for UTF-8.) The configure.ac script no longer substitutes the
2805            relevant variables.
2806    
2807        (b) The debugging options /B and /D in pcretest show the compiled bytecode
2808            with length and offset values. This means that the output is different
2809            for different internal link sizes. Test 2 is skipped for link sizes
2810            other than 2 because of this, bypassing the problem. Unfortunately,
2811            there was also a test in test 3 (the locale tests) that used /B and
2812            failed for link sizes other than 2. Rather than cut the whole test out,
2813            I have added a new /Z option to pcretest that replaces the length and
2814            offset values with spaces. This is now used to make test 3 independent
2815            of link size. (Test 2 will be tidied up later.)
2816    
2817    14. If erroroffset was passed as NULL to pcre_compile, it provoked a
2818        segmentation fault instead of returning the appropriate error message.
2819    
2820    15. In multiline mode when the newline sequence was set to "any", the pattern
2821        ^$ would give a match between the \r and \n of a subject such as "A\r\nB".
2822        This doesn't seem right; it now treats the CRLF combination as the line
2823        ending, and so does not match in that case. It's only a pattern such as ^$
2824        that would hit this one: something like ^ABC$ would have failed after \r
2825        and then tried again after \r\n.
2826    
2827    16. Changed the comparison command for RunGrepTest from "diff -u" to "diff -ub"
2828        in an attempt to make files that differ only in their line terminators
2829        compare equal. This works on Linux.
2830    
2831    17. Under certain error circumstances pcregrep might try to free random memory
2832        as it exited. This is now fixed, thanks to valgrind.
2833    
2834    19. In pcretest, if the pattern /(?m)^$/g<any> was matched against the string
2835        "abc\r\n\r\n", it found an unwanted second match after the second \r. This
2836        was because its rules for how to advance for /g after matching an empty
2837        string at the end of a line did not allow for this case. They now check for
2838        it specially.
2839    
2840    20. pcretest is supposed to handle patterns and data of any length, by
2841        extending its buffers when necessary. It was getting this wrong when the
2842        buffer for a data line had to be extended.
2843    
2844    21. Added PCRE_NEWLINE_ANYCRLF which is like ANY, but matches only CR, LF, or
2845        CRLF as a newline sequence.
2846    
2847    22. Code for handling Unicode properties in pcre_dfa_exec() wasn't being cut
2848        out by #ifdef SUPPORT_UCP. This did no harm, as it could never be used, but
2849        I have nevertheless tidied it up.
2850    
2851    23. Added some casts to kill warnings from HP-UX ia64 compiler.
2852    
2853    24. Added a man page for pcre-config.
2854    
2855    
2856    Version 7.0 19-Dec-06
2857    ---------------------
2858    
2859     1. Fixed a signed/unsigned compiler warning in pcre_compile.c, shown up by
2860        moving to gcc 4.1.1.
2861    
2862     2. The -S option for pcretest uses setrlimit(); I had omitted to #include
2863        sys/time.h, which is documented as needed for this function. It doesn't
2864        seem to matter on Linux, but it showed up on some releases of OS X.
2865    
2866     3. It seems that there are systems where bytes whose values are greater than
2867        127 match isprint() in the "C" locale. The "C" locale should be the
2868        default when a C program starts up. In most systems, only ASCII printing
2869        characters match isprint(). This difference caused the output from pcretest
2870        to vary, making some of the tests fail. I have changed pcretest so that:
2871    
2872        (a) When it is outputting text in the compiled version of a pattern, bytes
2873            other than 32-126 are always shown as hex escapes.
2874    
2875        (b) When it is outputting text that is a matched part of a subject string,
2876            it does the same, unless a different locale has been set for the match
2877            (using the /L modifier). In this case, it uses isprint() to decide.
2878    
2879     4. Fixed a major bug that caused incorrect computation of the amount of memory
2880        required for a compiled pattern when options that changed within the
2881        pattern affected the logic of the preliminary scan that determines the
2882        length. The relevant options are -x, and -i in UTF-8 mode. The result was
2883        that the computed length was too small. The symptoms of this bug were
2884        either the PCRE error "internal error: code overflow" from pcre_compile(),
2885        or a glibc crash with a message such as "pcretest: free(): invalid next
2886        size (fast)". Examples of patterns that provoked this bug (shown in
2887        pcretest format) are:
2888    
2889          /(?-x: )/x
2890          /(?x)(?-x: \s*#\s*)/
2891          /((?i)[\x{c0}])/8
2892          /(?i:[\x{c0}])/8
2893    
2894        HOWEVER: Change 17 below makes this fix obsolete as the memory computation
2895        is now done differently.
2896    
2897     5. Applied patches from Google to: (a) add a QuoteMeta function to the C++
2898        wrapper classes; (b) implement a new function in the C++ scanner that is
2899        more efficient than the old way of doing things because it avoids levels of
2900        recursion in the regex matching; (c) add a paragraph to the documentation
2901        for the FullMatch() function.
2902    
2903     6. The escape sequence \n was being treated as whatever was defined as
2904        "newline". Not only was this contrary to the documentation, which states
2905        that \n is character 10 (hex 0A), but it also went horribly wrong when
2906        "newline" was defined as CRLF. This has been fixed.
2907    
2908     7. In pcre_dfa_exec.c the value of an unsigned integer (the variable called c)
2909        was being set to -1 for the "end of line" case (supposedly a value that no
2910        character can have). Though this value is never used (the check for end of
2911        line is "zero bytes in current character"), it caused compiler complaints.
2912        I've changed it to 0xffffffff.
2913    
2914     8. In pcre_version.c, the version string was being built by a sequence of
2915        C macros that, in the event of PCRE_PRERELEASE being defined as an empty
2916        string (as it is for production releases) called a macro with an empty
2917        argument. The C standard says the result of this is undefined. The gcc
2918        compiler treats it as an empty string (which was what was wanted) but it is
2919        reported that Visual C gives an error. The source has been hacked around to
2920        avoid this problem.
2921    
2922     9. On the advice of a Windows user, included <io.h> and <fcntl.h> in Windows
2923        builds of pcretest, and changed the call to _setmode() to use _O_BINARY
2924        instead of 0x8000. Made all the #ifdefs test both _WIN32 and WIN32 (not all
2925        of them did).
2926    
2927    10. Originally, pcretest opened its input and output without "b"; then I was
2928        told that "b" was needed in some environments, so it was added for release
2929        5.0 to both the input and output. (It makes no difference on Unix-like
2930        systems.) Later I was told that it is wrong for the input on Windows. I've
2931        now abstracted the modes into two macros, to make it easier to fiddle with
2932        them, and removed "b" from the input mode under Windows.
2933    
2934    11. Added pkgconfig support for the C++ wrapper library, libpcrecpp.
2935    
2936    12. Added -help and --help to pcretest as an official way of being reminded
2937        of the options.
2938    
2939    13. Removed some redundant semicolons after macro calls in pcrecpparg.h.in
2940        and pcrecpp.cc because they annoy compilers at high warning levels.
2941    
2942    14. A bit of tidying/refactoring in pcre_exec.c in the main bumpalong loop.
2943    
2944    15. Fixed an occurrence of == in configure.ac that should have been = (shell
2945        scripts are not C programs :-) and which was not noticed because it works
2946        on Linux.
2947    
2948    16. pcretest is supposed to handle any length of pattern and data line (as one
2949        line or as a continued sequence of lines) by extending its input buffer if
2950        necessary. This feature was broken for very long pattern lines, leading to
2951        a string of junk being passed to pcre_compile() if the pattern was longer
2952        than about 50K.
2953    
2954    17. I have done a major re-factoring of the way pcre_compile() computes the
2955        amount of memory needed for a compiled pattern. Previously, there was code
2956        that made a preliminary scan of the pattern in order to do this. That was
2957        OK when PCRE was new, but as the facilities have expanded, it has become
2958        harder and harder to keep it in step with the real compile phase, and there
2959        have been a number of bugs (see for example, 4 above). I have now found a
2960        cunning way of running the real compile function in a "fake" mode that
2961        enables it to compute how much memory it would need, while actually only
2962        ever using a few hundred bytes of working memory and without too many
2963        tests of the mode. This should make future maintenance and development
2964        easier. A side effect of this work is that the limit of 200 on the nesting
2965        depth of parentheses has been removed (though this was never a serious
2966        limitation, I suspect). However, there is a downside: pcre_compile() now
2967        runs more slowly than before (30% or more, depending on the pattern). I
2968        hope this isn't a big issue. There is no effect on runtime performance.
2969    
2970    18. Fixed a minor bug in pcretest: if a pattern line was not terminated by a
2971        newline (only possible for the last line of a file) and it was a
2972        pattern that set a locale (followed by /Lsomething), pcretest crashed.
2973    
2974    19. Added additional timing features to pcretest. (1) The -tm option now times
2975        matching only, not compiling. (2) Both -t and -tm can be followed, as a
2976        separate command line item, by a number that specifies the number of
2977        repeats to use when timing. The default is 50000; this gives better
2978        precision, but takes uncomfortably long for very large patterns.
2979    
2980    20. Extended pcre_study() to be more clever in cases where a branch of a
2981        subpattern has no definite first character. For example, (a*|b*)[cd] would
2982        previously give no result from pcre_study(). Now it recognizes that the
2983        first character must be a, b, c, or d.
2984    
2985    21. There was an incorrect error "recursive call could loop indefinitely" if
2986        a subpattern (or the entire pattern) that was being tested for matching an
2987        empty string contained only one non-empty item after a nested subpattern.
2988        For example, the pattern (?>\x{100}*)\d(?R) provoked this error
2989        incorrectly, because the \d was being skipped in the check.
2990    
2991    22. The pcretest program now has a new pattern option /B and a command line
2992        option -b, which is equivalent to adding /B to every pattern. This causes
2993        it to show the compiled bytecode, without the additional information that
2994        -d shows. The effect of -d is now the same as -b with -i (and similarly, /D
2995        is the same as /B/I).
2996    
2997    23. A new optimization is now able automatically to treat some sequences such
2998        as a*b as a*+b. More specifically, if something simple (such as a character
2999        or a simple class like \d) has an unlimited quantifier, and is followed by
3000        something that cannot possibly match the quantified thing, the quantifier
3001        is automatically "possessified".
3002    
3003    24. A recursive reference to a subpattern whose number was greater than 39
3004        went wrong under certain circumstances in UTF-8 mode. This bug could also
3005        have affected the operation of pcre_study().
3006    
3007    25. Realized that a little bit of performance could be had by replacing
3008        (c & 0xc0) == 0xc0 with c >= 0xc0 when processing UTF-8 characters.
3009    
3010    26. Timing data from pcretest is now shown to 4 decimal places instead of 3.
3011    
3012    27. Possessive quantifiers such as a++ were previously implemented by turning
3013        them into atomic groups such as ($>a+). Now they have their own opcodes,
3014        which improves performance. This includes the automatically created ones
3015        from 23 above.
3016    
3017    28. A pattern such as (?=(\w+))\1: which simulates an atomic group using a
3018        lookahead was broken if it was not anchored. PCRE was mistakenly expecting
3019        the first matched character to be a colon. This applied both to named and
3020        numbered groups.
3021    
3022    29. The ucpinternal.h header file was missing its idempotency #ifdef.
3023    
3024    30. I was sent a "project" file called libpcre.a.dev which I understand makes
3025        building PCRE on Windows easier, so I have included it in the distribution.
3026    
3027    31. There is now a check in pcretest against a ridiculously large number being
3028        returned by pcre_exec() or pcre_dfa_exec(). If this happens in a /g or /G
3029        loop, the loop is abandoned.
3030    
3031    32. Forward references to subpatterns in conditions such as (?(2)...) where
3032        subpattern 2 is defined later cause pcre_compile() to search forwards in
3033        the pattern for the relevant set of parentheses. This search went wrong
3034        when there were unescaped parentheses in a character class, parentheses
3035        escaped with \Q...\E, or parentheses in a #-comment in /x mode.
3036    
3037    33. "Subroutine" calls and backreferences were previously restricted to
3038        referencing subpatterns earlier in the regex. This restriction has now
3039        been removed.
3040    
3041    34. Added a number of extra features that are going to be in Perl 5.10. On the
3042        whole, these are just syntactic alternatives for features that PCRE had
3043        previously implemented using the Python syntax or my own invention. The
3044        other formats are all retained for compatibility.
3045    
3046        (a) Named groups can now be defined as (?<name>...) or (?'name'...) as well
3047            as (?P<name>...). The new forms, as well as being in Perl 5.10, are
3048            also .NET compatible.
3049    
3050        (b) A recursion or subroutine call to a named group can now be defined as
3051            (?&name) as well as (?P>name).
3052    
3053        (c) A backreference to a named group can now be defined as \k<name> or
3054            \k'name' as well as (?P=name). The new forms, as well as being in Perl
3055            5.10, are also .NET compatible.
3056    
3057        (d) A conditional reference to a named group can now use the syntax
3058            (?(<name>) or (?('name') as well as (?(name).
3059    
3060        (e) A "conditional group" of the form (?(DEFINE)...) can be used to define
3061            groups (named and numbered) that are never evaluated inline, but can be
3062            called as "subroutines" from elsewhere. In effect, the DEFINE condition
3063            is always false. There may be only one alternative in such a group.
3064    
3065        (f) A test for recursion can be given as (?(R1).. or (?(R&name)... as well
3066            as the simple (?(R). The condition is true only if the most recent
3067            recursion is that of the given number or name. It does not search out
3068            through the entire recursion stack.
3069    
3070        (g) The escape \gN or \g{N} has been added, where N is a positive or
3071            negative number, specifying an absolute or relative reference.
3072    
3073    35. Tidied to get rid of some further signed/unsigned compiler warnings and
3074        some "unreachable code" warnings.
3075    
3076    36. Updated the Unicode property tables to Unicode version 5.0.0. Amongst other
3077        things, this adds five new scripts.
3078    
3079    37. Perl ignores orphaned \E escapes completely. PCRE now does the same.
3080        There were also incompatibilities regarding the handling of \Q..\E inside
3081        character classes, for example with patterns like [\Qa\E-\Qz\E] where the
3082        hyphen was adjacent to \Q or \E. I hope I've cleared all this up now.
3083    
3084    38. Like Perl, PCRE detects when an indefinitely repeated parenthesized group
3085        matches an empty string, and forcibly breaks the loop. There were bugs in
3086        this code in non-simple cases. For a pattern such as  ^(a()*)*  matched
3087        against  aaaa  the result was just "a" rather than "aaaa", for example. Two
3088        separate and independent bugs (that affected different cases) have been
3089        fixed.
3090    
3091    39. Refactored the code to abolish the use of different opcodes for small
3092        capturing bracket numbers. This is a tidy that I avoided doing when I
3093        removed the limit on the number of capturing brackets for 3.5 back in 2001.
3094        The new approach is not only tidier, it makes it possible to reduce the
3095        memory needed to fix the previous bug (38).
3096    
3097    40. Implemented PCRE_NEWLINE_ANY to recognize any of the Unicode newline
3098        sequences (http://unicode.org/unicode/reports/tr18/) as "newline" when
3099        processing dot, circumflex, or dollar metacharacters, or #-comments in /x
3100        mode.
3101    
3102    41. Add \R to match any Unicode newline sequence, as suggested in the Unicode
3103        report.
3104    
3105    42. Applied patch, originally from Ari Pollak, modified by Google, to allow
3106        copy construction and assignment in the C++ wrapper.
3107    
3108    43. Updated pcregrep to support "--newline=any". In the process, I fixed a
3109        couple of bugs that could have given wrong results in the "--newline=crlf"
3110        case.
3111    
3112    44. Added a number of casts and did some reorganization of signed/unsigned int
3113        variables following suggestions from Dair Grant. Also renamed the variable
3114        "this" as "item" because it is a C++ keyword.
3115    
3116    45. Arranged for dftables to add
3117    
3118          #include "pcre_internal.h"
3119    
3120        to pcre_chartables.c because without it, gcc 4.x may remove the array
3121        definition from the final binary if PCRE is built into a static library and
3122        dead code stripping is activated.
3123    
3124    46. For an unanchored pattern, if a match attempt fails at the start of a
3125        newline sequence, and the newline setting is CRLF or ANY, and the next two
3126        characters are CRLF, advance by two characters instead of one.
3127    
3128    
3129    Version 6.7 04-Jul-06
3130    ---------------------
3131    
3132     1. In order to handle tests when input lines are enormously long, pcretest has
3133        been re-factored so that it automatically extends its buffers when
3134        necessary. The code is crude, but this _is_ just a test program. The
3135        default size has been increased from 32K to 50K.
3136    
3137     2. The code in pcre_study() was using the value of the re argument before
3138        testing it for NULL. (Of course, in any sensible call of the function, it
3139        won't be NULL.)
3140    
3141     3. The memmove() emulation function in pcre_internal.h, which is used on
3142        systems that lack both memmove() and bcopy() - that is, hardly ever -
3143        was missing a "static" storage class specifier.
3144    
3145     4. When UTF-8 mode was not set, PCRE looped when compiling certain patterns
3146        containing an extended class (one that cannot be represented by a bitmap
3147        because it contains high-valued characters or Unicode property items, e.g.
3148        [\pZ]). Almost always one would set UTF-8 mode when processing such a
3149        pattern, but PCRE should not loop if you do not (it no longer does).
3150        [Detail: two cases were found: (a) a repeated subpattern containing an
3151        extended class; (b) a recursive reference to a subpattern that followed a
3152        previous extended class. It wasn't skipping over the extended class
3153        correctly when UTF-8 mode was not set.]
3154    
3155     5. A negated single-character class was not being recognized as fixed-length
3156        in lookbehind assertions such as (?<=[^f]), leading to an incorrect
3157        compile error "lookbehind assertion is not fixed length".
3158    
3159     6. The RunPerlTest auxiliary script was showing an unexpected difference
3160        between PCRE and Perl for UTF-8 tests. It turns out that it is hard to
3161        write a Perl script that can interpret lines of an input file either as
3162        byte characters or as UTF-8, which is what "perltest" was being required to
3163        do for the non-UTF-8 and UTF-8 tests, respectively. Essentially what you
3164        can't do is switch easily at run time between having the "use utf8;" pragma
3165        or not. In the end, I fudged it by using the RunPerlTest script to insert
3166        "use utf8;" explicitly for the UTF-8 tests.
3167    
3168     7. In multiline (/m) mode, PCRE was matching ^ after a terminating newline at
3169        the end of the subject string, contrary to the documentation and to what
3170        Perl does. This was true of both matching functions. Now it matches only at
3171        the start of the subject and immediately after *internal* newlines.
3172    
3173     8. A call of pcre_fullinfo() from pcretest to get the option bits was passing
3174        a pointer to an int instead of a pointer to an unsigned long int. This
3175        caused problems on 64-bit systems.
3176    
3177     9. Applied a patch from the folks at Google to pcrecpp.cc, to fix "another
3178        instance of the 'standard' template library not being so standard".
3179    
3180    10. There was no check on the number of named subpatterns nor the maximum
3181        length of a subpattern name. The product of these values is used to compute
3182        the size of the memory block for a compiled pattern. By supplying a very
3183        long subpattern name and a large number of named subpatterns, the size
3184        computation could be caused to overflow. This is now prevented by limiting
3185        the length of names to 32 characters, and the number of named subpatterns
3186        to 10,000.
3187    
3188    11. Subpatterns that are repeated with specific counts have to be replicated in
3189        the compiled pattern. The size of memory for this was computed from the
3190        length of the subpattern and the repeat count. The latter is limited to
3191        65535, but there was no limit on the former, meaning that integer overflow
3192        could in principle occur. The compiled length of a repeated subpattern is
3193        now limited to 30,000 bytes in order to prevent this.
3194    
3195    12. Added the optional facility to have named substrings with the same name.
3196    
3197    13. Added the ability to use a named substring as a condition, using the
3198        Python syntax: (?(name)yes|no). This overloads (?(R)... and names that
3199        are numbers (not recommended). Forward references are permitted.
3200    
3201    14. Added forward references in named backreferences (if you see what I mean).
3202    
3203    15. In UTF-8 mode, with the PCRE_DOTALL option set, a quantified dot in the
3204        pattern could run off the end of the subject. For example, the pattern
3205        "(?s)(.{1,5})"8 did this with the subject "ab".
3206    
3207    16. If PCRE_DOTALL or PCRE_MULTILINE were set, pcre_dfa_exec() behaved as if
3208        PCRE_CASELESS was set when matching characters that were quantified with ?
3209        or *.
3210    
3211    17. A character class other than a single negated character that had a minimum
3212        but no maximum quantifier - for example [ab]{6,} - was not handled
3213        correctly by pce_dfa_exec(). It would match only one character.
3214    
3215    18. A valid (though odd) pattern that looked like a POSIX character
3216        class but used an invalid character after [ (for example [[,abc,]]) caused
3217        pcre_compile() to give the error "Failed: internal error: code overflow" or
3218        in some cases to crash with a glibc free() error. This could even happen if
3219        the pattern terminated after [[ but there just happened to be a sequence of
3220        letters, a binary zero, and a closing ] in the memory that followed.
3221    
3222    19. Perl's treatment of octal escapes in the range \400 to \777 has changed
3223        over the years. Originally (before any Unicode support), just the bottom 8
3224        bits were taken. Thus, for example, \500 really meant \100. Nowadays the
3225        output from "man perlunicode" includes this:
3226    
3227          The regular expression compiler produces polymorphic opcodes.  That
3228          is, the pattern adapts to the data and automatically switches to
3229          the Unicode character scheme when presented with Unicode data--or
3230          instead uses a traditional byte scheme when presented with byte
3231          data.
3232    
3233        Sadly, a wide octal escape does not cause a switch, and in a string with
3234        no other multibyte characters, these octal escapes are treated as before.
3235        Thus, in Perl, the pattern  /\500/ actually matches \100 but the pattern
3236        /\500|\x{1ff}/ matches \500 or \777 because the whole thing is treated as a
3237        Unicode string.
3238    
3239        I have not perpetrated such confusion in PCRE. Up till now, it took just
3240        the bottom 8 bits, as in old Perl. I have now made octal escapes with
3241        values greater than \377 illegal in non-UTF-8 mode. In UTF-8 mode they
3242        translate to the appropriate multibyte character.
3243    
3244    29. Applied some refactoring to reduce the number of warnings from Microsoft
3245        and Borland compilers. This has included removing the fudge introduced
3246        seven years ago for the OS/2 compiler (see 2.02/2 below) because it caused
3247        a warning about an unused variable.
3248    
3249    21. PCRE has not included VT (character 0x0b) in the set of whitespace
3250        characters since release 4.0, because Perl (from release 5.004) does not.
3251        [Or at least, is documented not to: some releases seem to be in conflict
3252        with the documentation.] However, when a pattern was studied with
3253        pcre_study() and all its branches started with \s, PCRE still included VT
3254        as a possible starting character. Of course, this did no harm; it just
3255        caused an unnecessary match attempt.
3256    
3257    22. Removed a now-redundant internal flag bit that recorded the fact that case
3258        dependency changed within the pattern. This was once needed for "required
3259        byte" processing, but is no longer used. This recovers a now-scarce options
3260        bit. Also moved the least significant internal flag bit to the most-
3261        significant bit of the word, which was not previously used (hangover from
3262        the days when it was an int rather than a uint) to free up another bit for
3263        the future.
3264    
3265    23. Added support for CRLF line endings as well as CR and LF. As well as the
3266        default being selectable at build time, it can now be changed at runtime
3267        via the PCRE_NEWLINE_xxx flags. There are now options for pcregrep to
3268        specify that it is scanning data with non-default line endings.
3269    
3270    24. Changed the definition of CXXLINK to make it agree with the definition of
3271        LINK in the Makefile, by replacing LDFLAGS to CXXFLAGS.
3272    
3273    25. Applied Ian Taylor's patches to avoid using another stack frame for tail
3274        recursions. This makes a big different to stack usage for some patterns.
3275    
3276    26. If a subpattern containing a named recursion or subroutine reference such
3277        as (?P>B) was quantified, for example (xxx(?P>B)){3}, the calculation of
3278        the space required for the compiled pattern went wrong and gave too small a
3279        value. Depending on the environment, this could lead to "Failed: internal
3280        error: code overflow at offset 49" or "glibc detected double free or
3281        corruption" errors.
3282    
3283    27. Applied patches from Google (a) to support the new newline modes and (b) to
3284        advance over multibyte UTF-8 characters in GlobalReplace.
3285    
3286    28. Change free() to pcre_free() in pcredemo.c. Apparently this makes a
3287        difference for some implementation of PCRE in some Windows version.
3288    
3289    29. Added some extra testing facilities to pcretest:
3290    
3291        \q<number>   in a data line sets the "match limit" value
3292        \Q<number>   in a data line sets the "match recursion limt" value
3293        -S <number>  sets the stack size, where <number> is in megabytes
3294    
3295        The -S option isn't available for Windows.
3296    
3297    
3298    Version 6.6 06-Feb-06
3299    ---------------------
3300    
3301     1. Change 16(a) for 6.5 broke things, because PCRE_DATA_SCOPE was not defined
3302        in pcreposix.h. I have copied the definition from pcre.h.
3303    
3304     2. Change 25 for 6.5 broke compilation in a build directory out-of-tree
3305        because pcre.h is no longer a built file.
3306    
3307     3. Added Jeff Friedl's additional debugging patches to pcregrep. These are
3308        not normally included in the compiled code.
3309    
3310    
3311    Version 6.5 01-Feb-06
3312    ---------------------
3313    
3314     1. When using the partial match feature with pcre_dfa_exec(), it was not
3315        anchoring the second and subsequent partial matches at the new starting
3316        point. This could lead to incorrect results. For example, with the pattern
3317        /1234/, partially matching against "123" and then "a4" gave a match.
3318    
3319     2. Changes to pcregrep:
3320    
3321        (a) All non-match returns from pcre_exec() were being treated as failures
3322            to match the line. Now, unless the error is PCRE_ERROR_NOMATCH, an
3323            error message is output. Some extra information is given for the
3324            PCRE_ERROR_MATCHLIMIT and PCRE_ERROR_RECURSIONLIMIT errors, which are
3325            probably the only errors that are likely to be caused by users (by
3326            specifying a regex that has nested indefinite repeats, for instance).
3327            If there are more than 20 of these errors, pcregrep is abandoned.
3328    
3329        (b) A binary zero was treated as data while matching, but terminated the
3330            output line if it was written out. This has been fixed: binary zeroes
3331            are now no different to any other data bytes.
3332    
3333        (c) Whichever of the LC_ALL or LC_CTYPE environment variables is set is
3334            used to set a locale for matching. The --locale=xxxx long option has
3335            been added (no short equivalent) to specify a locale explicitly on the
3336            pcregrep command, overriding the environment variables.
3337    
3338        (d) When -B was used with -n, some line numbers in the output were one less
3339            than they should have been.
3340    
3341        (e) Added the -o (--only-matching) option.
3342    
3343        (f) If -A or -C was used with -c (count only), some lines of context were
3344            accidentally printed for the final match.
3345    
3346        (g) Added the -H (--with-filename) option.
3347    
3348        (h) The combination of options -rh failed to suppress file names for files
3349            that were found from directory arguments.
3350    
3351        (i) Added the -D (--devices) and -d (--directories) options.
3352    
3353        (j) Added the -F (--fixed-strings) option.
3354    
3355        (k) Allow "-" to be used as a file name for -f as well as for a data file.
3356    
3357        (l) Added the --colo(u)r option.
3358    
3359        (m) Added Jeffrey Friedl's -S testing option, but within #ifdefs so that it
3360            is not present by default.
3361    
3362     3. A nasty bug was discovered in the handling of recursive patterns, that is,
3363        items such as (?R) or (?1), when the recursion could match a number of
3364        alternatives. If it matched one of the alternatives, but subsequently,
3365        outside the recursion, there was a failure, the code tried to back up into
3366        the recursion. However, because of the way PCRE is implemented, this is not
3367        possible, and the result was an incorrect result from the match.
3368    
3369        In order to prevent this happening, the specification of recursion has
3370        been changed so that all such subpatterns are automatically treated as
3371        atomic groups. Thus, for example, (?R) is treated as if it were (?>(?R)).
3372    
3373     4. I had overlooked the fact that, in some locales, there are characters for
3374        which isalpha() is true but neither isupper() nor islower() are true. In
3375        the fr_FR locale, for instance, the \xAA and \xBA characters (ordmasculine
3376        and ordfeminine) are like this. This affected the treatment of \w and \W
3377        when they appeared in character classes, but not when they appeared outside
3378        a character class. The bit map for "word" characters is now created
3379        separately from the results of isalnum() instead of just taking it from the
3380        upper, lower, and digit maps. (Plus the underscore character, of course.)
3381    
3382     5. The above bug also affected the handling of POSIX character classes such as
3383        [[:alpha:]] and [[:alnum:]]. These do not have their own bit maps in PCRE's
3384        permanent tables. Instead, the bit maps for such a class were previously
3385        created as the appropriate unions of the upper, lower, and digit bitmaps.
3386        Now they are created by subtraction from the [[:word:]] class, which has
3387        its own bitmap.
3388    
3389     6. The [[:blank:]] character class matches horizontal, but not vertical space.
3390        It is created by subtracting the vertical space characters (\x09, \x0a,
3391        \x0b, \x0c) from the [[:space:]] bitmap. Previously, however, the
3392        subtraction was done in the overall bitmap for a character class, meaning
3393        that a class such as [\x0c[:blank:]] was incorrect because \x0c would not
3394        be recognized. This bug has been fixed.
3395    
3396     7. Patches from the folks at Google:
3397    
3398          (a) pcrecpp.cc: "to handle a corner case that may or may not happen in
3399          real life, but is still worth protecting against".
3400    
3401          (b) pcrecpp.cc: "corrects a bug when negative radixes are used with
3402          regular expressions".
3403    
3404          (c) pcre_scanner.cc: avoid use of std::count() because not all systems
3405          have it.
3406    
3407          (d) Split off pcrecpparg.h from pcrecpp.h and had the former built by
3408          "configure" and the latter not, in order to fix a problem somebody had
3409          with compiling the Arg class on HP-UX.
3410    
3411          (e) Improve the error-handling of the C++ wrapper a little bit.
3412    
3413          (f) New tests for checking recursion limiting.
3414    
3415     8. The pcre_memmove() function, which is used only if the environment does not
3416        have a standard memmove() function (and is therefore rarely compiled),
3417        contained two bugs: (a) use of int instead of size_t, and (b) it was not
3418        returning a result (though PCRE never actually uses the result).
3419    
3420     9. In the POSIX regexec() interface, if nmatch is specified as a ridiculously
3421        large number - greater than INT_MAX/(3*sizeof(int)) - REG_ESPACE is
3422        returned instead of calling malloc() with an overflowing number that would
3423        most likely cause subsequent chaos.
3424    
3425    10. The debugging option of pcretest was not showing the NO_AUTO_CAPTURE flag.
3426    
3427    11. The POSIX flag REG_NOSUB is now supported. When a pattern that was compiled
3428        with this option is matched, the nmatch and pmatch options of regexec() are
3429        ignored.
3430    
3431    12. Added REG_UTF8 to the POSIX interface. This is not defined by POSIX, but is
3432        provided in case anyone wants to the the POSIX interface with UTF-8
3433        strings.
3434    
3435    13. Added CXXLDFLAGS to the Makefile parameters to provide settings only on the
3436        C++ linking (needed for some HP-UX environments).
3437    
3438    14. Avoid compiler warnings in get_ucpname() when compiled without UCP support
3439        (unused parameter) and in the pcre_printint() function (omitted "default"
3440        switch label when the default is to do nothing).
3441    
3442    15. Added some code to make it possible, when PCRE is compiled as a C++
3443        library, to replace subject pointers for pcre_exec() with a smart pointer
3444        class, thus making it possible to process discontinuous strings.
3445    
3446    16. The two macros PCRE_EXPORT and PCRE_DATA_SCOPE are confusing, and perform
3447        much the same function. They were added by different people who were trying
3448        to make PCRE easy to compile on non-Unix systems. It has been suggested
3449        that PCRE_EXPORT be abolished now that there is more automatic apparatus
3450        for compiling on Windows systems. I have therefore replaced it with
3451        PCRE_DATA_SCOPE. This is set automatically for Windows; if not set it
3452        defaults to "extern" for C or "extern C" for C++, which works fine on
3453        Unix-like systems. It is now possible to override the value of PCRE_DATA_
3454        SCOPE with something explicit in config.h. In addition:
3455    
3456        (a) pcreposix.h still had just "extern" instead of either of these macros;
3457            I have replaced it with PCRE_DATA_SCOPE.
3458    
3459        (b) Functions such as _pcre_xclass(), which are internal to the library,
3460            but external in the C sense, all had PCRE_EXPORT in their definitions.
3461            This is apparently wrong for the Windows case, so I have removed it.
3462            (It makes no difference on Unix-like systems.)
3463    
3464    17. Added a new limit, MATCH_LIMIT_RECURSION, which limits the depth of nesting
3465        of recursive calls to match(). This is different to MATCH_LIMIT because
3466        that limits the total number of calls to match(), not all of which increase
3467        the depth of recursion. Limiting the recursion depth limits the amount of
3468        stack (or heap if NO_RECURSE is set) that is used. The default can be set
3469        when PCRE is compiled, and changed at run time. A patch from Google adds
3470        this functionality to the C++ interface.
3471    
3472    18. Changes to the handling of Unicode character properties:
3473    
3474        (a) Updated the table to Unicode 4.1.0.
3475    
3476        (b) Recognize characters that are not in the table as "Cn" (undefined).
3477    
3478        (c) I revised the way the table is implemented to a much improved format
3479            which includes recognition of ranges. It now supports the ranges that
3480            are defined in UnicodeData.txt, and it also amalgamates other
3481            characters into ranges. This has reduced the number of entries in the
3482            table from around 16,000 to around 3,000, thus reducing its size
3483            considerably. I realized I did not need to use a tree structure after
3484            all - a binary chop search is just as efficient. Having reduced the
3485            number of entries, I extended their size from 6 bytes to 8 bytes to
3486            allow for more data.
3487    
3488        (d) Added support for Unicode script names via properties such as \p{Han}.
3489    
3490    19. In UTF-8 mode, a backslash followed by a non-Ascii character was not
3491        matching that character.
3492    
3493    20. When matching a repeated Unicode property with a minimum greater than zero,
3494        (for example \pL{2,}), PCRE could look past the end of the subject if it
3495        reached it while seeking the minimum number of characters. This could
3496        happen only if some of the characters were more than one byte long, because
3497        there is a check for at least the minimum number of bytes.
3498    
3499    21. Refactored the implementation of \p and \P so as to be more general, to
3500        allow for more different types of property in future. This has changed the
3501        compiled form incompatibly. Anybody with saved compiled patterns that use
3502        \p or \P will have to recompile them.
3503    
3504    22. Added "Any" and "L&" to the supported property types.
3505    
3506    23. Recognize \x{...} as a code point specifier, even when not in UTF-8 mode,
3507        but give a compile time error if the value is greater than 0xff.
3508    
3509    24. The man pages for pcrepartial, pcreprecompile, and pcre_compile2 were
3510        accidentally not being installed or uninstalled.
3511    
3512    25. The pcre.h file was built from pcre.h.in, but the only changes that were
3513        made were to insert the current release number. This seemed silly, because
3514        it made things harder for people building PCRE on systems that don't run
3515        "configure". I have turned pcre.h into a distributed file, no longer built
3516        by "configure", with the version identification directly included. There is
3517        no longer a pcre.h.in file.
3518    
3519        However, this change necessitated a change to the pcre-config script as
3520        well. It is built from pcre-config.in, and one of the substitutions was the
3521        release number. I have updated configure.ac so that ./configure now finds
3522        the release number by grepping pcre.h.
3523    
3524    26. Added the ability to run the tests under valgrind.
3525    
3526    
3527    Version 6.4 05-Sep-05
3528    ---------------------
3529    
3530     1. Change 6.0/10/(l) to pcregrep introduced a bug that caused separator lines
3531        "--" to be printed when multiple files were scanned, even when none of the
3532        -A, -B, or -C options were used. This is not compatible with Gnu grep, so I
3533        consider it to be a bug, and have restored the previous behaviour.
3534    
3535     2. A couple of code tidies to get rid of compiler warnings.
3536    
3537     3. The pcretest program used to cheat by referring to symbols in the library
3538        whose names begin with _pcre_. These are internal symbols that are not
3539        really supposed to be visible externally, and in some environments it is
3540        possible to suppress them. The cheating is now confined to including
3541        certain files from the library's source, which is a bit cleaner.
3542    
3543     4. Renamed pcre.in as pcre.h.in to go with pcrecpp.h.in; it also makes the
3544        file's purpose clearer.
3545    
3546     5. Reorganized pcre_ucp_findchar().
3547    
3548    
3549    Version 6.3 15-Aug-05
3550    ---------------------
3551    
3552     1. The file libpcre.pc.in did not have general read permission in the tarball.
3553    
3554     2. There were some problems when building without C++ support:
3555    
3556        (a) If C++ support was not built, "make install" and "make test" still
3557            tried to test it.
3558    
3559        (b) There were problems when the value of CXX was explicitly set. Some
3560            changes have been made to try to fix these, and ...
3561    
3562        (c) --disable-cpp can now be used to explicitly disable C++ support.
3563    
3564        (d) The use of @CPP_OBJ@ directly caused a blank line preceded by a
3565            backslash in a target when C++ was disabled. This confuses some
3566            versions of "make", apparently. Using an intermediate variable solves
3567            this. (Same for CPP_LOBJ.)
3568    
3569     3. $(LINK_FOR_BUILD) now includes $(CFLAGS_FOR_BUILD) and $(LINK)
3570        (non-Windows) now includes $(CFLAGS) because these flags are sometimes
3571        necessary on certain architectures.
3572    
3573     4. Added a setting of -export-symbols-regex to the link command to remove
3574        those symbols that are exported in the C sense, but actually are local
3575        within the library, and not documented. Their names all begin with
3576        "_pcre_". This is not a perfect job, because (a) we have to except some
3577        symbols that pcretest ("illegally") uses, and (b) the facility isn't always
3578        available (and never for static libraries). I have made a note to try to
3579        find a way round (a) in the future.
3580