pcre3 (1:8.30-5) unstable; urgency=low

[pcre3.git] / ChangeLog

ChangeLog for PCRE
------------------

Version 8.30 04-February-2012
-----------------------------

1.  Renamed "isnumber" as "is_a_number" because in some Mac environments this
    name is defined in ctype.h.

2.  Fixed a bug in fixed-length calculation for lookbehinds that would show up
    only in quite long subpatterns.

3.  Removed the function pcre_info(), which has been obsolete and deprecated
    since it was replaced by pcre_fullinfo() in February 2000.

4.  For a non-anchored pattern, if (*SKIP) was given with a name that did not
    match a (*MARK), and the match failed at the start of the subject, a
    reference to memory before the start of the subject could occur. This bug
    was introduced by fix 17 of release 8.21.

5.  A reference to an unset group with zero minimum repetition was giving
    totally wrong answers (in non-JavaScript-compatibility mode). For example,
    /(another)?(\1?)test/ matched against "hello world test". This bug was
    introduced in release 8.13.

6.  Add support for 16-bit character strings (a large amount of work involving
    many changes and refactorings).

7.  RunGrepTest failed on msys because \r\n was replaced by whitespace when the
    command "pattern=`printf 'xxx\r\njkl'`" was run. The pattern is now taken
    from a file.

8.  Ovector size of 2 is also supported by JIT based pcre_exec (the ovector size
    rounding is not applied in this particular case).

9.  The invalid Unicode surrogate codepoints U+D800 to U+DFFF are now rejected
    if they appear, or are escaped, in patterns.

10. Get rid of a number of -Wunused-but-set-variable warnings.

11. The pattern /(?=(*:x))(q|)/ matches an empty string, and returns the mark
    "x". The similar pattern /(?=(*:x))((*:y)q|)/ did not return a mark at all.
    Oddly, Perl behaves the same way. PCRE has been fixed so that this pattern
    also returns the mark "x". This bug applied to capturing parentheses,
    non-capturing parentheses, and atomic parentheses. It also applied to some
    assertions.

12. Stephen Kelly's patch to CMakeLists.txt allows it to parse the version
    information out of configure.ac instead of relying on pcre.h.generic, which
    is not stored in the repository.

13. Applied Dmitry V. Levin's patch for a more portable method for linking with
    -lreadline.

14. ZH added PCRE_CONFIG_JITTARGET; added its output to pcretest -C.

15. Applied Graycode's patch to put the top-level frame on the stack rather
    than the heap when not using the stack for recursion. This gives a
    performance improvement in many cases when recursion is not deep.

16. Experimental code added to "pcretest -C" to output the stack frame size.


Version 8.21 12-Dec-2011
------------------------

1.  Updating the JIT compiler.

2.  JIT compiler now supports OP_NCREF, OP_RREF and OP_NRREF. New test cases
    are added as well.

3.  Fix cache-flush issue on PowerPC (It is still an experimental JIT port).
    PCRE_EXTRA_TABLES is not suported by JIT, and should be checked before
    calling _pcre_jit_exec. Some extra comments are added.

4.  (*MARK) settings inside atomic groups that do not contain any capturing
    parentheses, for example, (?>a(*:m)), were not being passed out. This bug
    was introduced by change 18 for 8.20.

5.  Supporting of \x, \U and \u in JavaScript compatibility mode based on the
    ECMA-262 standard.

6.  Lookbehinds such as (?<=a{2}b) that contained a fixed repetition were
    erroneously being rejected as "not fixed length" if PCRE_CASELESS was set.
    This bug was probably introduced by change 9 of 8.13.

7.  While fixing 6 above, I noticed that a number of other items were being
    incorrectly rejected as "not fixed length". This arose partly because newer
    opcodes had not been added to the fixed-length checking code. I have (a)
    corrected the bug and added tests for these items, and (b) arranged for an
    error to occur if an unknown opcode is encountered while checking for fixed
    length instead of just assuming "not fixed length". The items that were
    rejected were: (*ACCEPT), (*COMMIT), (*FAIL), (*MARK), (*PRUNE), (*SKIP),
    (*THEN), \h, \H, \v, \V, and single character negative classes with fixed
    repetitions, e.g. [^a]{3}, with and without PCRE_CASELESS.

8.  A possessively repeated conditional subpattern such as (?(?=c)c|d)++ was
    being incorrectly compiled and would have given unpredicatble results.

9.  A possessively repeated subpattern with minimum repeat count greater than
    one behaved incorrectly. For example, (A){2,}+ behaved as if it was
    (A)(A)++ which meant that, after a subsequent mismatch, backtracking into
    the first (A) could occur when it should not.

10. Add a cast and remove a redundant test from the code.

11. JIT should use pcre_malloc/pcre_free for allocation.

12. Updated pcre-config so that it no longer shows -L/usr/lib, which seems
    best practice nowadays, and helps with cross-compiling. (If the exec_prefix
    is anything other than /usr, -L is still shown).

13. In non-UTF-8 mode, \C is now supported in lookbehinds and DFA matching.

14. Perl does not support \N without a following name in a [] class; PCRE now
    also gives an error.

15. If a forward reference was repeated with an upper limit of around 2000,
    it caused the error "internal error: overran compiling workspace". The
    maximum number of forward references (including repeats) was limited by the
    internal workspace, and dependent on the LINK_SIZE. The code has been
    rewritten so that the workspace expands (via pcre_malloc) if necessary, and
    the default depends on LINK_SIZE. There is a new upper limit (for safety)
    of around 200,000 forward references. While doing this, I also speeded up
    the filling in of repeated forward references.

16. A repeated forward reference in a pattern such as (a)(?2){2}(.) was
    incorrectly expecting the subject to contain another "a" after the start.

17. When (*SKIP:name) is activated without a corresponding (*MARK:name) earlier
    in the match, the SKIP should be ignored. This was not happening; instead
    the SKIP was being treated as NOMATCH. For patterns such as
    /A(*MARK:A)A+(*SKIP:B)Z|AAC/ this meant that the AAC branch was never
    tested.

18. The behaviour of (*MARK), (*PRUNE), and (*THEN) has been reworked and is
    now much more compatible with Perl, in particular in cases where the result
    is a non-match for a non-anchored pattern. For example, if
    /b(*:m)f|a(*:n)w/ is matched against "abc", the non-match returns the name
    "m", where previously it did not return a name. A side effect of this
    change is that for partial matches, the last encountered mark name is
    returned, as for non matches. A number of tests that were previously not
    Perl-compatible have been moved into the Perl-compatible test files. The
    refactoring has had the pleasing side effect of removing one argument from
    the match() function, thus reducing its stack requirements.

19. If the /S+ option was used in pcretest to study a pattern using JIT,
    subsequent uses of /S (without +) incorrectly behaved like /S+.

21. Retrieve executable code size support for the JIT compiler and fixing
    some warnings.

22. A caseless match of a UTF-8 character whose other case uses fewer bytes did
    not work when the shorter character appeared right at the end of the
    subject string.

23. Added some (int) casts to non-JIT modules to reduce warnings on 64-bit
    systems.

24. Added PCRE_INFO_JITSIZE to pass on the value from (21) above, and also
    output it when the /M option is used in pcretest.

25. The CheckMan script was not being included in the distribution. Also, added
    an explicit "perl" to run Perl scripts from the PrepareRelease script
    because this is reportedly needed in Windows.

26. If study data was being save in a file and studying had not found a set of
    "starts with" bytes for the pattern, the data written to the file (though
    never used) was taken from uninitialized memory and so caused valgrind to
    complain.

27. Updated RunTest.bat as provided by Sheri Pierce.

28. Fixed a possible uninitialized memory bug in pcre_jit_compile.c.

29. Computation of memory usage for the table of capturing group names was
    giving an unnecessarily large value.


Version 8.20 21-Oct-2011
------------------------

1.  Change 37 of 8.13 broke patterns like [:a]...[b:] because it thought it had
    a POSIX class. After further experiments with Perl, which convinced me that
    Perl has bugs and confusions, a closing square bracket is no longer allowed
    in a POSIX name. This bug also affected patterns with classes that started
    with full stops.

2.  If a pattern such as /(a)b|ac/ is matched against "ac", there is no
    captured substring, but while checking the failing first alternative,
    substring 1 is temporarily captured. If the output vector supplied to
    pcre_exec() was not big enough for this capture, the yield of the function
    was still zero ("insufficient space for captured substrings"). This cannot
    be totally fixed without adding another stack variable, which seems a lot
    of expense for a edge case. However, I have improved the situation in cases
    such as /(a)(b)x|abc/ matched against "abc", where the return code
    indicates that fewer than the maximum number of slots in the ovector have
    been set.

3.  Related to (2) above: when there are more back references in a pattern than
    slots in the output vector, pcre_exec() uses temporary memory during
    matching, and copies in the captures as far as possible afterwards. It was
    using the entire output vector, but this conflicts with the specification
    that only 2/3 is used for passing back captured substrings. Now it uses
    only the first 2/3, for compatibility. This is, of course, another edge
    case.

4.  Zoltan Herczeg's just-in-time compiler support has been integrated into the
    main code base, and can be used by building with --enable-jit. When this is
    done, pcregrep automatically uses it unless --disable-pcregrep-jit or the
    runtime --no-jit option is given.

5.  When the number of matches in a pcre_dfa_exec() run exactly filled the
    ovector, the return from the function was zero, implying that there were
    other matches that did not fit. The correct "exactly full" value is now
    returned.

6.  If a subpattern that was called recursively or as a subroutine contained
    (*PRUNE) or any other control that caused it to give a non-standard return,
    invalid errors such as "Error -26 (nested recursion at the same subject
    position)" or even infinite loops could occur.

7.  If a pattern such as /a(*SKIP)c|b(*ACCEPT)|/ was studied, it stopped
    computing the minimum length on reaching *ACCEPT, and so ended up with the
    wrong value of 1 rather than 0. Further investigation indicates that
    computing a minimum subject length in the presence of *ACCEPT is difficult
    (think back references, subroutine calls), and so I have changed the code
    so that no minimum is registered for a pattern that contains *ACCEPT.

8.  If (*THEN) was present in the first (true) branch of a conditional group,
    it was not handled as intended. [But see 16 below.]

9.  Replaced RunTest.bat and CMakeLists.txt with improved versions provided by
    Sheri Pierce.

10. A pathological pattern such as /(*ACCEPT)a/ was miscompiled, thinking that
    the first byte in a match must be "a".

11. Change 17 for 8.13 increased the recursion depth for patterns like
    /a(?:.)*?a/ drastically. I've improved things by remembering whether a
    pattern contains any instances of (*THEN). If it does not, the old
    optimizations are restored. It would be nice to do this on a per-group
    basis, but at the moment that is not feasible.

12. In some environments, the output of pcretest -C is CRLF terminated. This
    broke RunTest's code that checks for the link size. A single white space
    character after the value is now allowed for.

13. RunTest now checks for the "fr" locale as well as for "fr_FR" and "french".
    For "fr", it uses the Windows-specific input and output files.

14. If (*THEN) appeared in a group that was called recursively or as a
    subroutine, it did not work as intended. [But see next item.]

15. Consider the pattern /A (B(*THEN)C) | D/ where A, B, C, and D are complex
    pattern fragments (but not containing any | characters). If A and B are
    matched, but there is a failure in C so that it backtracks to (*THEN), PCRE
    was behaving differently to Perl. PCRE backtracked into A, but Perl goes to
    D. In other words, Perl considers parentheses that do not contain any |
    characters to be part of a surrounding alternative, whereas PCRE was
    treading (B(*THEN)C) the same as (B(*THEN)C|(*FAIL)) -- which Perl handles
    differently. PCRE now behaves in the same way as Perl, except in the case
    of subroutine/recursion calls such as (?1) which have in any case always
    been different (but PCRE had them first :-).

16. Related to 15 above: Perl does not treat the | in a conditional group as
    creating alternatives. Such a group is treated in the same way as an
    ordinary group without any | characters when processing (*THEN). PCRE has
    been changed to match Perl's behaviour.

17. If a user had set PCREGREP_COLO(U)R to something other than 1:31, the
    RunGrepTest script failed.

18. Change 22 for version 13 caused atomic groups to use more stack. This is
    inevitable for groups that contain captures, but it can lead to a lot of
    stack use in large patterns. The old behaviour has been restored for atomic
    groups that do not contain any capturing parentheses.

19. If the PCRE_NO_START_OPTIMIZE option was set for pcre_compile(), it did not
    suppress the check for a minimum subject length at run time. (If it was
    given to pcre_exec() or pcre_dfa_exec() it did work.)

20. Fixed an ASCII-dependent infelicity in pcretest that would have made it
    fail to work when decoding hex characters in data strings in EBCDIC
    environments.

21. It appears that in at least one Mac OS environment, the isxdigit() function
    is implemented as a macro that evaluates to its argument more than once,
    contravening the C 90 Standard (I haven't checked a later standard). There
    was an instance in pcretest which caused it to go wrong when processing
    \x{...} escapes in subject strings. The has been rewritten to avoid using
    things like p++ in the argument of isxdigit().


Version 8.13 16-Aug-2011
------------------------

1.  The Unicode data tables have been updated to Unicode 6.0.0.

2.  Two minor typos in pcre_internal.h have been fixed.

3.  Added #include <string.h> to pcre_scanner_unittest.cc, pcrecpp.cc, and
    pcrecpp_unittest.cc. They are needed for strcmp(), memset(), and strchr()
    in some environments (e.g. Solaris 10/SPARC using Sun Studio 12U2).

4.  There were a number of related bugs in the code for matching backrefences
    caselessly in UTF-8 mode when codes for the characters concerned were
    different numbers of bytes. For example, U+023A and U+2C65 are an upper
    and lower case pair, using 2 and 3 bytes, respectively. The main bugs were:
    (a) A reference to 3 copies of a 2-byte code matched only 2 of a 3-byte
    code. (b) A reference to 2 copies of a 3-byte code would not match 2 of a
    2-byte code at the end of the subject (it thought there wasn't enough data
    left).

5.  Comprehensive information about what went wrong is now returned by
    pcre_exec() and pcre_dfa_exec() when the UTF-8 string check fails, as long
    as the output vector has at least 2 elements. The offset of the start of
    the failing character and a reason code are placed in the vector.

6.  When the UTF-8 string check fails for pcre_compile(), the offset that is
    now returned is for the first byte of the failing character, instead of the
    last byte inspected. This is an incompatible change, but I hope it is small
    enough not to be a problem. It makes the returned offset consistent with
    pcre_exec() and pcre_dfa_exec().

7.  pcretest now gives a text phrase as well as the error number when
    pcre_exec() or pcre_dfa_exec() fails; if the error is a UTF-8 check
    failure, the offset and reason code are output.

8.  When \R was used with a maximizing quantifier it failed to skip backwards
    over a \r\n pair if the subsequent match failed. Instead, it just skipped
    back over a single character (\n). This seems wrong (because it treated the
    two characters as a single entity when going forwards), conflicts with the
    documentation that \R is equivalent to (?>\r\n|\n|...etc), and makes the
    behaviour of \R* different to (\R)*, which also seems wrong. The behaviour
    has been changed.

9.  Some internal refactoring has changed the processing so that the handling
    of the PCRE_CASELESS and PCRE_MULTILINE options is done entirely at compile
    time (the PCRE_DOTALL option was changed this way some time ago: version
    7.7 change 16). This has made it possible to abolish the OP_OPT op code,
    which was always a bit of a fudge. It also means that there is one less
    argument for the match() function, which reduces its stack requirements
    slightly. This change also fixes an incompatibility with Perl: the pattern
    (?i:([^b]))(?1) should not match "ab", but previously PCRE gave a match.

10. More internal refactoring has drastically reduced the number of recursive
    calls to match() for possessively repeated groups such as (abc)++ when
    using pcre_exec().

11. While implementing 10, a number of bugs in the handling of groups were
    discovered and fixed:

    (?<=(a)+) was not diagnosed as invalid (non-fixed-length lookbehind).
    (a|)*(?1) gave a compile-time internal error.
    ((a|)+)+  did not notice that the outer group could match an empty string.
    (^a|^)+   was not marked as anchored.
    (.*a|.*)+ was not marked as matching at start or after a newline.

12. Yet more internal refactoring has removed another argument from the match()
    function. Special calls to this function are now indicated by setting a
    value in a variable in the "match data" data block.

13. Be more explicit in pcre_study() instead of relying on "default" for
    opcodes that mean there is no starting character; this means that when new
    ones are added and accidentally left out of pcre_study(), testing should
    pick them up.

14. The -s option of pcretest has been documented for ages as being an old
    synonym of -m (show memory usage). I have changed it to mean "force study
    for every regex", that is, assume /S for every regex. This is similar to -i
    and -d etc. It's slightly incompatible, but I'm hoping nobody is still
    using it. It makes it easier to run collections of tests with and without
    study enabled, and thereby test pcre_study() more easily. All the standard
    tests are now run with and without -s (but some patterns can be marked as
    "never study" - see 20 below).

15. When (*ACCEPT) was used in a subpattern that was called recursively, the
    restoration of the capturing data to the outer values was not happening
    correctly.

16. If a recursively called subpattern ended with (*ACCEPT) and matched an
    empty string, and PCRE_NOTEMPTY was set, pcre_exec() thought the whole
    pattern had matched an empty string, and so incorrectly returned a no
    match.

17. There was optimizing code for the last branch of non-capturing parentheses,
    and also for the obeyed branch of a conditional subexpression, which used
    tail recursion to cut down on stack usage. Unfortunately, now that there is
    the possibility of (*THEN) occurring in these branches, tail recursion is
    no longer possible because the return has to be checked for (*THEN). These
    two optimizations have therefore been removed. [But see 8.20/11 above.]

18. If a pattern containing \R was studied, it was assumed that \R always
    matched two bytes, thus causing the minimum subject length to be
    incorrectly computed because \R can also match just one byte.

19. If a pattern containing (*ACCEPT) was studied, the minimum subject length
    was incorrectly computed.

20. If /S is present twice on a test pattern in pcretest input, it now
    *disables* studying, thereby overriding the use of -s on the command line
    (see 14 above). This is necessary for one or two tests to keep the output
    identical in both cases.

21. When (*ACCEPT) was used in an assertion that matched an empty string and
    PCRE_NOTEMPTY was set, PCRE applied the non-empty test to the assertion.

22. When an atomic group that contained a capturing parenthesis was
    successfully matched, but the branch in which it appeared failed, the
    capturing was not being forgotten if a higher numbered group was later
    captured. For example, /(?>(a))b|(a)c/ when matching "ac" set capturing
    group 1 to "a", when in fact it should be unset. This applied to multi-
    branched capturing and non-capturing groups, repeated or not, and also to
    positive assertions (capturing in negative assertions does not happen
    in PCRE) and also to nested atomic groups.

23. Add the ++ qualifier feature to pcretest, to show the remainder of the
    subject after a captured substring, to make it easier to tell which of a
    number of identical substrings has been captured.

24. The way atomic groups are processed by pcre_exec() has been changed so that
    if they are repeated, backtracking one repetition now resets captured
    values correctly. For example, if ((?>(a+)b)+aabab) is matched against
    "aaaabaaabaabab" the value of captured group 2 is now correctly recorded as
    "aaa". Previously, it would have been "a". As part of this code
    refactoring, the way recursive calls are handled has also been changed.

25. If an assertion condition captured any substrings, they were not passed
    back unless some other capturing happened later. For example, if
    (?(?=(a))a) was matched against "a", no capturing was returned.

26. When studying a pattern that contained subroutine calls or assertions,
    the code for finding the minimum length of a possible match was handling
    direct recursions such as (xxx(?1)|yyy) but not mutual recursions (where
    group 1 called group 2 while simultaneously a separate group 2 called group
    1). A stack overflow occurred in this case. I have fixed this by limiting
    the recursion depth to 10.

27. Updated RunTest.bat in the distribution to the version supplied by Tom
    Fortmann. This supports explicit test numbers on the command line, and has
    argument validation and error reporting.

28. An instance of \X with an unlimited repeat could fail if at any point the
    first character it looked at was a mark character.

29. Some minor code refactoring concerning Unicode properties and scripts
    should reduce the stack requirement of match() slightly.

30. Added the '=' option to pcretest to check the setting of unused capturing
    slots at the end of the pattern, which are documented as being -1, but are
    not included in the return count.

31. If \k was not followed by a braced, angle-bracketed, or quoted name, PCRE
    compiled something random. Now it gives a compile-time error (as does
    Perl).

32. A *MARK encountered during the processing of a positive assertion is now
    recorded and passed back (compatible with Perl).

33. If --only-matching or --colour was set on a pcregrep call whose pattern
    had alternative anchored branches, the search for a second match in a line
    was done as if at the line start. Thus, for example, /^01|^02/ incorrectly
    matched the line "0102" twice. The same bug affected patterns that started
    with a backwards assertion. For example /\b01|\b02/ also matched "0102"
    twice.

34. Previously, PCRE did not allow quantification of assertions. However, Perl
    does, and because of capturing effects, quantifying parenthesized
    assertions may at times be useful. Quantifiers are now allowed for
    parenthesized assertions.

35. A minor code tidy in pcre_compile() when checking options for \R usage.

36. \g was being checked for fancy things in a character class, when it should
    just be a literal "g".

37. PCRE was rejecting [:a[:digit:]] whereas Perl was not. It seems that the
    appearance of a nested POSIX class supersedes an apparent external class.
    For example, [:a[:digit:]b:] matches "a", "b", ":", or a digit. Also,
    unescaped square brackets may also appear as part of class names. For
    example, [:a[:abc]b:] gives unknown class "[:abc]b:]". PCRE now behaves
    more like Perl. (But see 8.20/1 above.)

38. PCRE was giving an error for \N with a braced quantifier such as {1,} (this
    was because it thought it was \N{name}, which is not supported).

39. Add minix to OS list not supporting the -S option in pcretest.

40. PCRE tries to detect cases of infinite recursion at compile time, but it
    cannot analyze patterns in sufficient detail to catch mutual recursions
    such as ((?1))((?2)). There is now a runtime test that gives an error if a
    subgroup is called recursively as a subpattern for a second time at the
    same position in the subject string. In previous releases this might have
    been caught by the recursion limit, or it might have run out of stack.

41. A pattern such as /(?(R)a+|(?R)b)/ is quite safe, as the recursion can
    happen only once. PCRE was, however incorrectly giving a compile time error
    "recursive call could loop indefinitely" because it cannot analyze the
    pattern in sufficient detail. The compile time test no longer happens when
    PCRE is compiling a conditional subpattern, but actual runaway loops are
    now caught at runtime (see 40 above).

42. It seems that Perl allows any characters other than a closing parenthesis
    to be part of the NAME in (*MARK:NAME) and other backtracking verbs. PCRE
    has been changed to be the same.

43. Updated configure.ac to put in more quoting round AC_LANG_PROGRAM etc. so
    as not to get warnings when autogen.sh is called. Also changed
    AC_PROG_LIBTOOL (deprecated) to LT_INIT (the current macro).

44. To help people who use pcregrep to scan files containing exceedingly long
    lines, the following changes have been made:

    (a) The default value of the buffer size parameter has been increased from
        8K to 20K. (The actual buffer used is three times this size.)

    (b) The default can be changed by ./configure --with-pcregrep-bufsize when
        PCRE is built.

    (c) A --buffer-size=n option has been added to pcregrep, to allow the size
        to be set at run time.

    (d) Numerical values in pcregrep options can be followed by K or M, for
        example --buffer-size=50K.

    (e) If a line being scanned overflows pcregrep's buffer, an error is now
        given and the return code is set to 2.

45. Add a pointer to the latest mark to the callout data block.

46. The pattern /.(*F)/, when applied to "abc" with PCRE_PARTIAL_HARD, gave a
    partial match of an empty string instead of no match. This was specific to
    the use of ".".

47. The pattern /f.*/8s, when applied to "for" with PCRE_PARTIAL_HARD, gave a
    complete match instead of a partial match. This bug was dependent on both
    the PCRE_UTF8 and PCRE_DOTALL options being set.

48. For a pattern such as /\babc|\bdef/ pcre_study() was failing to set up the
    starting byte set, because \b was not being ignored.


Version 8.12 15-Jan-2011
------------------------

1.  Fixed some typos in the markup of the man pages, and wrote a script that
    checks for such things as part of the documentation building process.

2.  On a big-endian 64-bit system, pcregrep did not correctly process the
    --match-limit and --recursion-limit options (added for 8.11). In
    particular, this made one of the standard tests fail. (The integer value
    went into the wrong half of a long int.)

3.  If the --colour option was given to pcregrep with -v (invert match), it
    did strange things, either producing crazy output, or crashing. It should,
    of course, ignore a request for colour when reporting lines that do not
    match.

4.  Another pcregrep bug caused similar problems if --colour was specified with
    -M (multiline) and the pattern match finished with a line ending.

5.  In pcregrep, when a pattern that ended with a literal newline sequence was
    matched in multiline mode, the following line was shown as part of the
    match. This seems wrong, so I have changed it.

6.  Another pcregrep bug in multiline mode, when --colour was specified, caused
    the check for further matches in the same line (so they could be coloured)
    to overrun the end of the current line. If another match was found, it was
    incorrectly shown (and then shown again when found in the next line).

7.  If pcregrep was compiled under Windows, there was a reference to the
    function pcregrep_exit() before it was defined. I am assuming this was
    the cause of the "error C2371: 'pcregrep_exit' : redefinition;" that was
    reported by a user. I've moved the definition above the reference.


Version 8.11 10-Dec-2010
------------------------

1.  (*THEN) was not working properly if there were untried alternatives prior
    to it in the current branch. For example, in ((a|b)(*THEN)(*F)|c..) it
    backtracked to try for "b" instead of moving to the next alternative branch
    at the same level (in this case, to look for "c"). The Perl documentation
    is clear that when (*THEN) is backtracked onto, it goes to the "next
    alternative in the innermost enclosing group".

2.  (*COMMIT) was not overriding (*THEN), as it does in Perl. In a pattern
    such as   (A(*COMMIT)B(*THEN)C|D)  any failure after matching A should
    result in overall failure. Similarly, (*COMMIT) now overrides (*PRUNE) and
    (*SKIP), (*SKIP) overrides (*PRUNE) and (*THEN), and (*PRUNE) overrides
    (*THEN).

3.  If \s appeared in a character class, it removed the VT character from
    the class, even if it had been included by some previous item, for example
    in [\x00-\xff\s]. (This was a bug related to the fact that VT is not part
    of \s, but is part of the POSIX "space" class.)

4.  A partial match never returns an empty string (because you can always
    match an empty string at the end of the subject); however the checking for
    an empty string was starting at the "start of match" point. This has been
    changed to the "earliest inspected character" point, because the returned
    data for a partial match starts at this character. This means that, for
    example, /(?<=abc)def/ gives a partial match for the subject "abc"
    (previously it gave "no match").

5.  Changes have been made to the way PCRE_PARTIAL_HARD affects the matching
    of $, \z, \Z, \b, and \B. If the match point is at the end of the string,
    previously a full match would be given. However, setting PCRE_PARTIAL_HARD
    has an implication that the given string is incomplete (because a partial
    match is preferred over a full match). For this reason, these items now
    give a partial match in this situation. [Aside: previously, the one case
    /t\b/ matched against "cat" with PCRE_PARTIAL_HARD set did return a partial
    match rather than a full match, which was wrong by the old rules, but is
    now correct.]

6.  There was a bug in the handling of #-introduced comments, recognized when
    PCRE_EXTENDED is set, when PCRE_NEWLINE_ANY and PCRE_UTF8 were also set.
    If a UTF-8 multi-byte character included the byte 0x85 (e.g. +U0445, whose
    UTF-8 encoding is 0xd1,0x85), this was misinterpreted as a newline when
    scanning for the end of the comment. (*Character* 0x85 is an "any" newline,
    but *byte* 0x85 is not, in UTF-8 mode). This bug was present in several
    places in pcre_compile().

7.  Related to (6) above, when pcre_compile() was skipping #-introduced
    comments when looking ahead for named forward references to subpatterns,
    the only newline sequence it recognized was NL. It now handles newlines
    according to the set newline convention.

8.  SunOS4 doesn't have strerror() or strtoul(); pcregrep dealt with the
    former, but used strtoul(), whereas pcretest avoided strtoul() but did not
    cater for a lack of strerror(). These oversights have been fixed.

9.  Added --match-limit and --recursion-limit to pcregrep.

10. Added two casts needed to build with Visual Studio when NO_RECURSE is set.

11. When the -o option was used, pcregrep was setting a return code of 1, even
    when matches were found, and --line-buffered was not being honoured.

12. Added an optional parentheses number to the -o and --only-matching options
    of pcregrep.

13. Imitating Perl's /g action for multiple matches is tricky when the pattern
    can match an empty string. The code to do it in pcretest and pcredemo
    needed fixing:

    (a) When the newline convention was "crlf", pcretest got it wrong, skipping
        only one byte after an empty string match just before CRLF (this case
        just got forgotten; "any" and "anycrlf" were OK).

    (b) The pcretest code also had a bug, causing it to loop forever in UTF-8
        mode when an empty string match preceded an ASCII character followed by
        a non-ASCII character. (The code for advancing by one character rather
        than one byte was nonsense.)

    (c) The pcredemo.c sample program did not have any code at all to handle
        the cases when CRLF is a valid newline sequence.

14. Neither pcre_exec() nor pcre_dfa_exec() was checking that the value given
    as a starting offset was within the subject string. There is now a new
    error, PCRE_ERROR_BADOFFSET, which is returned if the starting offset is
    negative or greater than the length of the string. In order to test this,
    pcretest is extended to allow the setting of negative starting offsets.

15. In both pcre_exec() and pcre_dfa_exec() the code for checking that the
    starting offset points to the beginning of a UTF-8 character was
    unnecessarily clumsy. I tidied it up.

16. Added PCRE_ERROR_SHORTUTF8 to make it possible to distinguish between a
    bad UTF-8 sequence and one that is incomplete when using PCRE_PARTIAL_HARD.

17. Nobody had reported that the --include_dir option, which was added in
    release 7.7 should have been called --include-dir (hyphen, not underscore)
    for compatibility with GNU grep. I have changed it to --include-dir, but
    left --include_dir as an undocumented synonym, and the same for
    --exclude-dir, though that is not available in GNU grep, at least as of
    release 2.5.4.

18. At a user's suggestion, the macros GETCHAR and friends (which pick up UTF-8
    characters from a string of bytes) have been redefined so as not to use
    loops, in order to improve performance in some environments. At the same
    time, I abstracted some of the common code into auxiliary macros to save
    repetition (this should not affect the compiled code).

19. If \c was followed by a multibyte UTF-8 character, bad things happened. A
    compile-time error is now given if \c is not followed by an ASCII
    character, that is, a byte less than 128. (In EBCDIC mode, the code is
    different, and any byte value is allowed.)

20. Recognize (*NO_START_OPT) at the start of a pattern to set the PCRE_NO_
    START_OPTIMIZE option, which is now allowed at compile time - but just
    passed through to pcre_exec() or pcre_dfa_exec(). This makes it available
    to pcregrep and other applications that have no direct access to PCRE
    options. The new /Y option in pcretest sets this option when calling
    pcre_compile().

21. Change 18 of release 8.01 broke the use of named subpatterns for recursive
    back references. Groups containing recursive back references were forced to
    be atomic by that change, but in the case of named groups, the amount of
    memory required was incorrectly computed, leading to "Failed: internal
    error: code overflow". This has been fixed.

22. Some patches to pcre_stringpiece.h, pcre_stringpiece_unittest.cc, and
    pcretest.c, to avoid build problems in some Borland environments.


Version 8.10 25-Jun-2010
------------------------

1.  Added support for (*MARK:ARG) and for ARG additions to PRUNE, SKIP, and
    THEN.

2.  (*ACCEPT) was not working when inside an atomic group.

3.  Inside a character class, \B is treated as a literal by default, but
    faulted if PCRE_EXTRA is set. This mimics Perl's behaviour (the -w option
    causes the error). The code is unchanged, but I tidied the documentation.

4.  Inside a character class, PCRE always treated \R and \X as literals,
    whereas Perl faults them if its -w option is set. I have changed PCRE so
    that it faults them when PCRE_EXTRA is set.

5.  Added support for \N, which always matches any character other than
    newline. (It is the same as "." when PCRE_DOTALL is not set.)

6.  When compiling pcregrep with newer versions of gcc which may have
    FORTIFY_SOURCE set, several warnings "ignoring return value of 'fwrite',
    declared with attribute warn_unused_result" were given. Just casting the
    result to (void) does not stop the warnings; a more elaborate fudge is
    needed. I've used a macro to implement this.

7.  Minor change to pcretest.c to avoid a compiler warning.

8.  Added four artifical Unicode properties to help with an option to make
    \s etc use properties (see next item). The new properties are: Xan
    (alphanumeric), Xsp (Perl space), Xps (POSIX space), and Xwd (word).

9.  Added PCRE_UCP to make \b, \d, \s, \w, and certain POSIX character classes
    use Unicode properties. (*UCP) at the start of a pattern can be used to set
    this option. Modified pcretest to add /W to test this facility. Added
    REG_UCP to make it available via the POSIX interface.

10. Added --line-buffered to pcregrep.

11. In UTF-8 mode, if a pattern that was compiled with PCRE_CASELESS was
    studied, and the match started with a letter with a code point greater than
    127 whose first byte was different to the first byte of the other case of
    the letter, the other case of this starting letter was not recognized
    (#976).

12. If a pattern that was studied started with a repeated Unicode property
    test, for example, \p{Nd}+, there was the theoretical possibility of
    setting up an incorrect bitmap of starting bytes, but fortunately it could
    not have actually happened in practice until change 8 above was made (it
    added property types that matched character-matching opcodes).

13. pcre_study() now recognizes \h, \v, and \R when constructing a bit map of
    possible starting bytes for non-anchored patterns.

14. Extended the "auto-possessify" feature of pcre_compile(). It now recognizes
    \R, and also a number of cases that involve Unicode properties, both
    explicit and implicit when PCRE_UCP is set.

15. If a repeated Unicode property match (e.g. \p{Lu}*) was used with non-UTF-8
    input, it could crash or give wrong results if characters with values
    greater than 0xc0 were present in the subject string. (Detail: it assumed
    UTF-8 input when processing these items.)

16. Added a lot of (int) casts to avoid compiler warnings in systems where
    size_t is 64-bit (#991).

17. Added a check for running out of memory when PCRE is compiled with
    --disable-stack-for-recursion (#990).

18. If the last data line in a file for pcretest does not have a newline on
    the end, a newline was missing in the output.

19. The default pcre_chartables.c file recognizes only ASCII characters (values
    less than 128) in its various bitmaps. However, there is a facility for
    generating tables according to the current locale when PCRE is compiled. It
    turns out that in some environments, 0x85 and 0xa0, which are Unicode space
    characters, are recognized by isspace() and therefore were getting set in
    these tables, and indeed these tables seem to approximate to ISO 8859. This
    caused a problem in UTF-8 mode when pcre_study() was used to create a list
    of bytes that can start a match. For \s, it was including 0x85 and 0xa0,
    which of course cannot start UTF-8 characters. I have changed the code so
    that only real ASCII characters (less than 128) and the correct starting
    bytes for UTF-8 encodings are set for characters greater than 127 when in
    UTF-8 mode. (When PCRE_UCP is set - see 9 above - the code is different
    altogether.)

20. Added the /T option to pcretest so as to be able to run tests with non-
    standard character tables, thus making it possible to include the tests
    used for 19 above in the standard set of tests.

21. A pattern such as (?&t)(?#()(?(DEFINE)(?<t>a)) which has a forward
    reference to a subpattern the other side of a comment that contains an
    opening parenthesis caused either an internal compiling error, or a
    reference to the wrong subpattern.


Version 8.02 19-Mar-2010
------------------------

1.  The Unicode data tables have been updated to Unicode 5.2.0.

2.  Added the option --libs-cpp to pcre-config, but only when C++ support is
    configured.

3.  Updated the licensing terms in the pcregexp.pas file, as agreed with the
    original author of that file, following a query about its status.

4.  On systems that do not have stdint.h (e.g. Solaris), check for and include
    inttypes.h instead. This fixes a bug that was introduced by change 8.01/8.

5.  A pattern such as (?&t)*+(?(DEFINE)(?<t>.)) which has a possessive
    quantifier applied to a forward-referencing subroutine call, could compile
    incorrect code or give the error "internal error: previously-checked
    referenced subpattern not found".

6.  Both MS Visual Studio and Symbian OS have problems with initializing
    variables to point to external functions. For these systems, therefore,
    pcre_malloc etc. are now initialized to local functions that call the
    relevant global functions.

7.  There were two entries missing in the vectors called coptable and poptable
    in pcre_dfa_exec.c. This could lead to memory accesses outsize the vectors.
    I've fixed the data, and added a kludgy way of testing at compile time that
    the lengths are correct (equal to the number of opcodes).

8.  Following on from 7, I added a similar kludge to check the length of the
    eint vector in pcreposix.c.

9.  Error texts for pcre_compile() are held as one long string to avoid too
    much relocation at load time. To find a text, the string is searched,
    counting zeros. There was no check for running off the end of the string,
    which could happen if a new error number was added without updating the
    string.

10. \K gave a compile-time error if it appeared in a lookbehind assersion.

11. \K was not working if it appeared in an atomic group or in a group that
    was called as a "subroutine", or in an assertion. Perl 5.11 documents that
    \K is "not well defined" if used in an assertion. PCRE now accepts it if
    the assertion is positive, but not if it is negative.

12. Change 11 fortuitously reduced the size of the stack frame used in the
    "match()" function of pcre_exec.c by one pointer. Forthcoming
    implementation of support for (*MARK) will need an extra pointer on the
    stack; I have reserved it now, so that the stack frame size does not
    decrease.

13. A pattern such as (?P<L1>(?P<L2>0)|(?P>L2)(?P>L1)) in which the only other
    item in branch that calls a recursion is a subroutine call - as in the
    second branch in the above example - was incorrectly given the compile-
    time error "recursive call could loop indefinitely" because pcre_compile()
    was not correctly checking the subroutine for matching a non-empty string.

14. The checks for overrunning compiling workspace could trigger after an
    overrun had occurred. This is a "should never occur" error, but it can be
    triggered by pathological patterns such as hundreds of nested parentheses.
    The checks now trigger 100 bytes before the end of the workspace.

15. Fix typo in configure.ac: "srtoq" should be "strtoq".


Version 8.01 19-Jan-2010
------------------------

1.  If a pattern contained a conditional subpattern with only one branch (in
    particular, this includes all (*DEFINE) patterns), a call to pcre_study()
    computed the wrong minimum data length (which is of course zero for such
    subpatterns). This could cause incorrect "no match" results.

2.  For patterns such as (?i)a(?-i)b|c where an option setting at the start of
    the pattern is reset in the first branch, pcre_compile() failed with
    "internal error: code overflow at offset...". This happened only when
    the reset was to the original external option setting. (An optimization
    abstracts leading options settings into an external setting, which was the
    cause of this.)

3.  A pattern such as ^(?!a(*SKIP)b) where a negative assertion contained one
    of the verbs SKIP, PRUNE, or COMMIT, did not work correctly. When the
    assertion pattern did not match (meaning that the assertion was true), it
    was incorrectly treated as false if the SKIP had been reached during the
    matching. This also applied to assertions used as conditions.

4.  If an item that is not supported by pcre_dfa_exec() was encountered in an
    assertion subpattern, including such a pattern used as a condition,
    unpredictable results occurred, instead of the error return
    PCRE_ERROR_DFA_UITEM.

5.  The C++ GlobalReplace function was not working like Perl for the special
    situation when an empty string is matched. It now does the fancy magic
    stuff that is necessary.

6.  In pcre_internal.h, obsolete includes to setjmp.h and stdarg.h have been
    removed. (These were left over from very, very early versions of PCRE.)

7.  Some cosmetic changes to the code to make life easier when compiling it
    as part of something else:

    (a) Change DEBUG to PCRE_DEBUG.

    (b) In pcre_compile(), rename the member of the "branch_chain" structure
        called "current" as "current_branch", to prevent a collision with the
        Linux macro when compiled as a kernel module.

    (c) In pcre_study(), rename the function set_bit() as set_table_bit(), to
        prevent a collision with the Linux macro when compiled as a kernel
        module.

8.  In pcre_compile() there are some checks for integer overflows that used to
    cast potentially large values to (double). This has been changed to that
    when building, a check for int64_t is made, and if it is found, it is used
    instead, thus avoiding the use of floating point arithmetic. (There is no
    other use of FP in PCRE.) If int64_t is not found, the fallback is to
    double.

9.  Added two casts to avoid signed/unsigned warnings from VS Studio Express
    2005 (difference between two addresses compared to an unsigned value).

10. Change the standard AC_CHECK_LIB test for libbz2 in configure.ac to a
    custom one, because of the following reported problem in Windows:

      - libbz2 uses the Pascal calling convention (WINAPI) for the functions
          under Win32.
      - The standard autoconf AC_CHECK_LIB fails to include "bzlib.h",
          therefore missing the function definition.
      - The compiler thus generates a "C" signature for the test function.
      - The linker fails to find the "C" function.
      - PCRE fails to configure if asked to do so against libbz2.

11. When running libtoolize from libtool-2.2.6b as part of autogen.sh, these
    messages were output:

      Consider adding `AC_CONFIG_MACRO_DIR([m4])' to configure.ac and
      rerunning libtoolize, to keep the correct libtool macros in-tree.
      Consider adding `-I m4' to ACLOCAL_AMFLAGS in Makefile.am.

    I have done both of these things.

12. Although pcre_dfa_exec() does not use nearly as much stack as pcre_exec()
    most of the time, it *can* run out if it is given a pattern that contains a
    runaway infinite recursion. I updated the discussion in the pcrestack man
    page.

13. Now that we have gone to the x.xx style of version numbers, the minor
    version may start with zero. Using 08 or 09 is a bad idea because users
    might check the value of PCRE_MINOR in their code, and 08 or 09 may be
    interpreted as invalid octal numbers. I've updated the previous comment in
    configure.ac, and also added a check that gives an error if 08 or 09 are
    used.

14. Change 8.00/11 was not quite complete: code had been accidentally omitted,
    causing partial matching to fail when the end of the subject matched \W
    in a UTF-8 pattern where \W was quantified with a minimum of 3.

15. There were some discrepancies between the declarations in pcre_internal.h
    of _pcre_is_newline(), _pcre_was_newline(), and _pcre_valid_utf8() and
    their definitions. The declarations used "const uschar *" and the
    definitions used USPTR. Even though USPTR is normally defined as "const
    unsigned char *" (and uschar is typedeffed as "unsigned char"), it was
    reported that: "This difference in casting confuses some C++ compilers, for
    example, SunCC recognizes above declarations as different functions and
    generates broken code for hbpcre." I have changed the declarations to use
    USPTR.

16. GNU libtool is named differently on some systems. The autogen.sh script now
    tries several variants such as glibtoolize (MacOSX) and libtoolize1x
    (FreeBSD).

17. Applied Craig's patch that fixes an HP aCC compile error in pcre 8.00
    (strtoXX undefined when compiling pcrecpp.cc). The patch contains this
    comment: "Figure out how to create a longlong from a string: strtoll and
    equivalent. It's not enough to call AC_CHECK_FUNCS: hpux has a strtoll, for
    instance, but it only takes 2 args instead of 3!"

18. A subtle bug concerned with back references has been fixed by a change of
    specification, with a corresponding code fix. A pattern such as
    ^(xa|=?\1a)+$ which contains a back reference inside the group to which it
    refers, was giving matches when it shouldn't. For example, xa=xaaa would
    match that pattern. Interestingly, Perl (at least up to 5.11.3) has the
    same bug. Such groups have to be quantified to be useful, or contained
    inside another quantified group. (If there's no repetition, the reference
    can never match.) The problem arises because, having left the group and
    moved on to the rest of the pattern, a later failure that backtracks into
    the group uses the captured value from the final iteration of the group
    rather than the correct earlier one. I have fixed this in PCRE by forcing
    any group that contains a reference to itself to be an atomic group; that
    is, there cannot be any backtracking into it once it has completed. This is
    similar to recursive and subroutine calls.


Version 8.00 19-Oct-09
----------------------

1.  The table for translating pcre_compile() error codes into POSIX error codes
    was out-of-date, and there was no check on the pcre_compile() error code
    being within the table. This could lead to an OK return being given in
    error.

2.  Changed the call to open a subject file in pcregrep from fopen(pathname,
    "r") to fopen(pathname, "rb"), which fixed a problem with some of the tests
    in a Windows environment.

3.  The pcregrep --count option prints the count for each file even when it is
    zero, as does GNU grep. However, pcregrep was also printing all files when
    --files-with-matches was added. Now, when both options are given, it prints
    counts only for those files that have at least one match. (GNU grep just
    prints the file name in this circumstance, but including the count seems
    more useful - otherwise, why use --count?) Also ensured that the
    combination -clh just lists non-zero counts, with no names.

4.  The long form of the pcregrep -F option was incorrectly implemented as
    --fixed_strings instead of --fixed-strings. This is an incompatible change,
    but it seems right to fix it, and I didn't think it was worth preserving
    the old behaviour.

5.  The command line items --regex=pattern and --regexp=pattern were not
    recognized by pcregrep, which required --regex pattern or --regexp pattern
    (with a space rather than an '='). The man page documented the '=' forms,
    which are compatible with GNU grep; these now work.

6.  No libpcreposix.pc file was created for pkg-config; there was just
    libpcre.pc and libpcrecpp.pc. The omission has been rectified.

7.  Added #ifndef SUPPORT_UCP into the pcre_ucd.c module, to reduce its size
    when UCP support is not needed, by modifying the Python script that
    generates it from Unicode data files. This should not matter if the module
    is correctly used as a library, but I received one complaint about 50K of
    unwanted data. My guess is that the person linked everything into his
    program rather than using a library. Anyway, it does no harm.

8.  A pattern such as /\x{123}{2,2}+/8 was incorrectly compiled; the trigger
    was a minimum greater than 1 for a wide character in a possessive
    repetition. The same bug could also affect patterns like /(\x{ff}{0,2})*/8
    which had an unlimited repeat of a nested, fixed maximum repeat of a wide
    character. Chaos in the form of incorrect output or a compiling loop could
    result.

9.  The restrictions on what a pattern can contain when partial matching is
    requested for pcre_exec() have been removed. All patterns can now be
    partially matched by this function. In addition, if there are at least two
    slots in the offset vector, the offset of the earliest inspected character
    for the match and the offset of the end of the subject are set in them when
    PCRE_ERROR_PARTIAL is returned.

10. Partial matching has been split into two forms: PCRE_PARTIAL_SOFT, which is
    synonymous with PCRE_PARTIAL, for backwards compatibility, and
    PCRE_PARTIAL_HARD, which causes a partial match to supersede a full match,
    and may be more useful for multi-segment matching.

11. Partial matching with pcre_exec() is now more intuitive. A partial match
    used to be given if ever the end of the subject was reached; now it is
    given only if matching could not proceed because another character was
    needed. This makes a difference in some odd cases such as Z(*FAIL) with the
    string "Z", which now yields "no match" instead of "partial match". In the
    case of pcre_dfa_exec(), "no match" is given if every matching path for the
    final character ended with (*FAIL).

12. Restarting a match using pcre_dfa_exec() after a partial match did not work
    if the pattern had a "must contain" character that was already found in the
    earlier partial match, unless partial matching was again requested. For
    example, with the pattern /dog.(body)?/, the "must contain" character is
    "g". If the first part-match was for the string "dog", restarting with
    "sbody" failed. This bug has been fixed.

13. The string returned by pcre_dfa_exec() after a partial match has been
    changed so that it starts at the first inspected character rather than the
    first character of the match. This makes a difference only if the pattern
    starts with a lookbehind assertion or \b or \B (\K is not supported by
    pcre_dfa_exec()). It's an incompatible change, but it makes the two
    matching functions compatible, and I think it's the right thing to do.

14. Added a pcredemo man page, created automatically from the pcredemo.c file,
    so that the demonstration program is easily available in environments where
    PCRE has not been installed from source.

15. Arranged to add -DPCRE_STATIC to cflags in libpcre.pc, libpcreposix.cp,
    libpcrecpp.pc and pcre-config when PCRE is not compiled as a shared
    library.

16. Added REG_UNGREEDY to the pcreposix interface, at the request of a user.
    It maps to PCRE_UNGREEDY. It is not, of course, POSIX-compatible, but it
    is not the first non-POSIX option to be added. Clearly some people find
    these options useful.

17. If a caller to the POSIX matching function regexec() passes a non-zero
    value for nmatch with a NULL value for pmatch, the value of
    nmatch is forced to zero.

18. RunGrepTest did not have a test for the availability of the -u option of
    the diff command, as RunTest does. It now checks in the same way as
    RunTest, and also checks for the -b option.

19. If an odd number of negated classes containing just a single character
    interposed, within parentheses, between a forward reference to a named
    subpattern and the definition of the subpattern, compilation crashed with
    an internal error, complaining that it could not find the referenced
    subpattern. An example of a crashing pattern is /(?&A)(([^m])(?<A>))/.
    [The bug was that it was starting one character too far in when skipping
    over the character class, thus treating the ] as data rather than
    terminating the class. This meant it could skip too much.]

20. Added PCRE_NOTEMPTY_ATSTART in order to be able to correctly implement the
    /g option in pcretest when the pattern contains \K, which makes it possible
    to have an empty string match not at the start, even when the pattern is
    anchored. Updated pcretest and pcredemo to use this option.

21. If the maximum number of capturing subpatterns in a recursion was greater
    than the maximum at the outer level, the higher number was returned, but
    with unset values at the outer level. The correct (outer level) value is
    now given.

22. If (*ACCEPT) appeared inside capturing parentheses, previous releases of
    PCRE did not set those parentheses (unlike Perl). I have now found a way to
    make it do so. The string so far is captured, making this feature
    compatible with Perl.

23. The tests have been re-organized, adding tests 11 and 12, to make it
    possible to check the Perl 5.10 features against Perl 5.10.

24. Perl 5.10 allows subroutine calls in lookbehinds, as long as the subroutine
    pattern matches a fixed length string. PCRE did not allow this; now it
    does. Neither allows recursion.

25. I finally figured out how to implement a request to provide the minimum
    length of subject string that was needed in order to match a given pattern.
    (It was back references and recursion that I had previously got hung up
    on.) This code has now been added to pcre_study(); it finds a lower bound
    to the length of subject needed. It is not necessarily the greatest lower
    bound, but using it to avoid searching strings that are too short does give
    some useful speed-ups. The value is available to calling programs via
    pcre_fullinfo().

26. While implementing 25, I discovered to my embarrassment that pcretest had
    not been passing the result of pcre_study() to pcre_dfa_exec(), so the
    study optimizations had never been tested with that matching function.
    Oops. What is worse, even when it was passed study data, there was a bug in
    pcre_dfa_exec() that meant it never actually used it. Double oops. There
    were also very few tests of studied patterns with pcre_dfa_exec().

27. If (?| is used to create subpatterns with duplicate numbers, they are now
    allowed to have the same name, even if PCRE_DUPNAMES is not set. However,
    on the other side of the coin, they are no longer allowed to have different
    names, because these cannot be distinguished in PCRE, and this has caused
    confusion. (This is a difference from Perl.)

28. When duplicate subpattern names are present (necessarily with different
    numbers, as required by 27 above), and a test is made by name in a
    conditional pattern, either for a subpattern having been matched, or for
    recursion in such a pattern, all the associated numbered subpatterns are
    tested, and the overall condition is true if the condition is true for any
    one of them. This is the way Perl works, and is also more like the way
    testing by number works.


Version 7.9 11-Apr-09
---------------------

1.  When building with support for bzlib/zlib (pcregrep) and/or readline
    (pcretest), all targets were linked against these libraries. This included
    libpcre, libpcreposix, and libpcrecpp, even though they do not use these
    libraries. This caused unwanted dependencies to be created. This problem
    has been fixed, and now only pcregrep is linked with bzlib/zlib and only
    pcretest is linked with readline.

2.  The "typedef int BOOL" in pcre_internal.h that was included inside the
    "#ifndef FALSE" condition by an earlier change (probably 7.8/18) has been
    moved outside it again, because FALSE and TRUE are already defined in AIX,
    but BOOL is not.

3.  The pcre_config() function was treating the PCRE_MATCH_LIMIT and
    PCRE_MATCH_LIMIT_RECURSION values as ints, when they should be long ints.

4.  The pcregrep documentation said spaces were inserted as well as colons (or
    hyphens) following file names and line numbers when outputting matching
    lines. This is not true; no spaces are inserted. I have also clarified the
    wording for the --colour (or --color) option.

5.  In pcregrep, when --colour was used with -o, the list of matching strings
    was not coloured; this is different to GNU grep, so I have changed it to be
    the same.

6.  When --colo(u)r was used in pcregrep, only the first matching substring in
    each matching line was coloured. Now it goes on to look for further matches
    of any of the test patterns, which is the same behaviour as GNU grep.

7.  A pattern that could match an empty string could cause pcregrep to loop; it
    doesn't make sense to accept an empty string match in pcregrep, so I have
    locked it out (using PCRE's PCRE_NOTEMPTY option). By experiment, this
    seems to be how GNU grep behaves.

8.  The pattern (?(?=.*b)b|^) was incorrectly compiled as "match must be at
    start or after a newline", because the conditional assertion was not being
    correctly handled. The rule now is that both the assertion and what follows
    in the first alternative must satisfy the test.

9.  If auto-callout was enabled in a pattern with a conditional group whose
    condition was an assertion, PCRE could crash during matching, both with
    pcre_exec() and pcre_dfa_exec().

10. The PCRE_DOLLAR_ENDONLY option was not working when pcre_dfa_exec() was
    used for matching.

11. Unicode property support in character classes was not working for
    characters (bytes) greater than 127 when not in UTF-8 mode.

12. Added the -M command line option to pcretest.

14. Added the non-standard REG_NOTEMPTY option to the POSIX interface.

15. Added the PCRE_NO_START_OPTIMIZE match-time option.

16. Added comments and documentation about mis-use of no_arg in the C++
    wrapper.

17. Implemented support for UTF-8 encoding in EBCDIC environments, a patch
    from Martin Jerabek that uses macro names for all relevant character and
    string constants.

18. Added to pcre_internal.h two configuration checks: (a) If both EBCDIC and
    SUPPORT_UTF8 are set, give an error; (b) If SUPPORT_UCP is set without
    SUPPORT_UTF8, define SUPPORT_UTF8. The "configure" script handles both of
    these, but not everybody uses configure.

19. A conditional group that had only one branch was not being correctly
    recognized as an item that could match an empty string. This meant that an
    enclosing group might also not be so recognized, causing infinite looping
    (and probably a segfault) for patterns such as ^"((?(?=[a])[^"])|b)*"$
    with the subject "ab", where knowledge that the repeated group can match
    nothing is needed in order to break the loop.

20. If a pattern that was compiled with callouts was matched using pcre_dfa_
    exec(), but without supplying a callout function, matching went wrong.

21. If PCRE_ERROR_MATCHLIMIT occurred during a recursion, there was a memory
    leak if the size of the offset vector was greater than 30. When the vector
    is smaller, the saved offsets during recursion go onto a local stack
    vector, but for larger vectors malloc() is used. It was failing to free
    when the recursion yielded PCRE_ERROR_MATCH_LIMIT (or any other "abnormal"
    error, in fact).

22. There was a missing #ifdef SUPPORT_UTF8 round one of the variables in the
    heapframe that is used only when UTF-8 support is enabled. This caused no
    problem, but was untidy.

23. Steven Van Ingelgem's patch to CMakeLists.txt to change the name
    CMAKE_BINARY_DIR to PROJECT_BINARY_DIR so that it works when PCRE is
    included within another project.

24. Steven Van Ingelgem's patches to add more options to the CMake support,
    slightly modified by me:

      (a) PCRE_BUILD_TESTS can be set OFF not to build the tests, including
          not building pcregrep.

      (b) PCRE_BUILD_PCREGREP can be see OFF not to build pcregrep, but only
          if PCRE_BUILD_TESTS is also set OFF, because the tests use pcregrep.

25. Forward references, both numeric and by name, in patterns that made use of
    duplicate group numbers, could behave incorrectly or give incorrect errors,
    because when scanning forward to find the reference group, PCRE was not
    taking into account the duplicate group numbers. A pattern such as
    ^X(?3)(a)(?|(b)|(q))(Y) is an example.

26. Changed a few more instances of "const unsigned char *" to USPTR, making
    the feature of a custom pointer more persuasive (as requested by a user).

27. Wrapped the definitions of fileno and isatty for Windows, which appear in
    pcretest.c, inside #ifndefs, because it seems they are sometimes already
    pre-defined.

28. Added support for (*UTF8) at the start of a pattern.

29. Arrange for flags added by the "release type" setting in CMake to be shown
    in the configuration summary.


Version 7.8 05-Sep-08
---------------------

1.  Replaced UCP searching code with optimized version as implemented for Ad
    Muncher (http://www.admuncher.com/) by Peter Kankowski. This uses a two-
    stage table and inline lookup instead of a function, giving speed ups of 2
    to 5 times on some simple patterns that I tested. Permission was given to
    distribute the MultiStage2.py script that generates the tables (it's not in
    the tarball, but is in the Subversion repository).

2.  Updated the Unicode datatables to Unicode 5.1.0. This adds yet more
    scripts.

3.  Change 12 for 7.7 introduced a bug in pcre_study() when a pattern contained
    a group with a zero qualifier. The result of the study could be incorrect,
    or the function might crash, depending on the pattern.

4.  Caseless matching was not working for non-ASCII characters in back
    references. For example, /(\x{de})\1/8i was not matching \x{de}\x{fe}.
    It now works when Unicode Property Support is available.

5.  In pcretest, an escape such as \x{de} in the data was always generating
    a UTF-8 string, even in non-UTF-8 mode. Now it generates a single byte in
    non-UTF-8 mode. If the value is greater than 255, it gives a warning about
    truncation.

6.  Minor bugfix in pcrecpp.cc (change "" == ... to NULL == ...).

7.  Added two (int) casts to pcregrep when printing the difference of two
    pointers, in case they are 64-bit values.

8.  Added comments about Mac OS X stack usage to the pcrestack man page and to
    test 2 if it fails.

9.  Added PCRE_CALL_CONVENTION just before the names of all exported functions,
    and a #define of that name to empty if it is not externally set. This is to
    allow users of MSVC to set it if necessary.

10. The PCRE_EXP_DEFN macro which precedes exported functions was missing from
    the convenience functions in the pcre_get.c source file.

11. An option change at the start of a pattern that had top-level alternatives
    could cause overwriting and/or a crash. This command provoked a crash in
    some environments:

      printf "/(?i)[\xc3\xa9\xc3\xbd]|[\xc3\xa9\xc3\xbdA]/8\n" | pcretest

    This potential security problem was recorded as CVE-2008-2371.

12. For a pattern where the match had to start at the beginning or immediately
    after a newline (e.g /.*anything/ without the DOTALL flag), pcre_exec() and
    pcre_dfa_exec() could read past the end of the passed subject if there was
    no match. To help with detecting such bugs (e.g. with valgrind), I modified
    pcretest so that it places the subject at the end of its malloc-ed buffer.

13. The change to pcretest in 12 above threw up a couple more cases when pcre_
    exec() might read past the end of the data buffer in UTF-8 mode.

14. A similar bug to 7.3/2 existed when the PCRE_FIRSTLINE option was set and
    the data contained the byte 0x85 as part of a UTF-8 character within its
    first line. This applied both to normal and DFA matching.

15. Lazy qualifiers were not working in some cases in UTF-8 mode. For example,
    /^[^d]*?$/8 failed to match "abc".

16. Added a missing copyright notice to pcrecpp_internal.h.

17. Make it more clear in the documentation that values returned from
    pcre_exec() in ovector are byte offsets, not character counts.

18. Tidied a few places to stop certain compilers from issuing warnings.

19. Updated the Virtual Pascal + BCC files to compile the latest v7.7, as
    supplied by Stefan Weber. I made a further small update for 7.8 because
    there is a change of source arrangements: the pcre_searchfuncs.c module is
    replaced by pcre_ucd.c.


Version 7.7 07-May-08
---------------------

1.  Applied Craig's patch to sort out a long long problem: "If we can't convert
    a string to a long long, pretend we don't even have a long long." This is
    done by checking for the strtoq, strtoll, and _strtoi64 functions.

2.  Applied Craig's patch to pcrecpp.cc to restore ABI compatibility with
    pre-7.6 versions, which defined a global no_arg variable instead of putting
    it in the RE class. (See also #8 below.)

3.  Remove a line of dead code, identified by coverity and reported by Nuno
    Lopes.

4.  Fixed two related pcregrep bugs involving -r with --include or --exclude:

    (1) The include/exclude patterns were being applied to the whole pathnames
        of files, instead of just to the final components.

    (2) If there was more than one level of directory, the subdirectories were
        skipped unless they satisfied the include/exclude conditions. This is
        inconsistent with GNU grep (and could even be seen as contrary to the
        pcregrep specification - which I improved to make it absolutely clear).
        The action now is always to scan all levels of directory, and just
        apply the include/exclude patterns to regular files.

5.  Added the --include_dir and --exclude_dir patterns to pcregrep, and used
    --exclude_dir in the tests to avoid scanning .svn directories.

6.  Applied Craig's patch to the QuoteMeta function so that it escapes the
    NUL character as backslash + 0 rather than backslash + NUL, because PCRE
    doesn't support NULs in patterns.

7.  Added some missing "const"s to declarations of static tables in
    pcre_compile.c and pcre_dfa_exec.c.

8.  Applied Craig's patch to pcrecpp.cc to fix a problem in OS X that was
    caused by fix #2  above. (Subsequently also a second patch to fix the
    first patch. And a third patch - this was a messy problem.)

9.  Applied Craig's patch to remove the use of push_back().

10. Applied Alan Lehotsky's patch to add REG_STARTEND support to the POSIX
    matching function regexec().

11. Added support for the Oniguruma syntax \g<name>, \g<n>, \g'name', \g'n',
    which, however, unlike Perl's \g{...}, are subroutine calls, not back
    references. PCRE supports relative numbers with this syntax (I don't think
    Oniguruma does).

12. Previously, a group with a zero repeat such as (...){0} was completely
    omitted from the compiled regex. However, this means that if the group
    was called as a subroutine from elsewhere in the pattern, things went wrong
    (an internal error was given). Such groups are now left in the compiled
    pattern, with a new opcode that causes them to be skipped at execution
    time.

13. Added the PCRE_JAVASCRIPT_COMPAT option. This makes the following changes
    to the way PCRE behaves:

    (a) A lone ] character is dis-allowed (Perl treats it as data).

    (b) A back reference to an unmatched subpattern matches an empty string
        (Perl fails the current match path).

    (c) A data ] in a character class must be notated as \] because if the
        first data character in a class is ], it defines an empty class. (In
        Perl it is not possible to have an empty class.) The empty class []
        never matches; it forces failure and is equivalent to (*FAIL) or (?!).
        The negative empty class [^] matches any one character, independently
        of the DOTALL setting.

14. A pattern such as /(?2)[]a()b](abc)/ which had a forward reference to a
    non-existent subpattern following a character class starting with ']' and
    containing () gave an internal compiling error instead of "reference to
    non-existent subpattern". Fortunately, when the pattern did exist, the
    compiled code was correct. (When scanning forwards to check for the
    existencd of the subpattern, it was treating the data ']' as terminating
    the class, so got the count wrong. When actually compiling, the reference
    was subsequently set up correctly.)

15. The "always fail" assertion (?!) is optimzed to (*FAIL) by pcre_compile;
    it was being rejected as not supported by pcre_dfa_exec(), even though
    other assertions are supported. I have made pcre_dfa_exec() support
    (*FAIL).

16. The implementation of 13c above involved the invention of a new opcode,
    OP_ALLANY, which is like OP_ANY but doesn't check the /s flag. Since /s
    cannot be changed at match time, I realized I could make a small
    improvement to matching performance by compiling OP_ALLANY instead of
    OP_ANY for "." when DOTALL was set, and then removing the runtime tests
    on the OP_ANY path.

17. Compiling pcretest on Windows with readline support failed without the
    following two fixes: (1) Make the unistd.h include conditional on
    HAVE_UNISTD_H; (2) #define isatty and fileno as _isatty and _fileno.

18. Changed CMakeLists.txt and cmake/FindReadline.cmake to arrange for the
    ncurses library to be included for pcretest when ReadLine support is
    requested, but also to allow for it to be overridden. This patch came from
    Daniel Bergström.

19. There was a typo in the file ucpinternal.h where f0_rangeflag was defined
    as 0x00f00000 instead of 0x00800000. Luckily, this would not have caused
    any errors with the current Unicode tables. Thanks to Peter Kankowski for
    spotting this.


Version 7.6 28-Jan-08
---------------------

1.  A character class containing a very large number of characters with
    codepoints greater than 255 (in UTF-8 mode, of course) caused a buffer
    overflow.

2.  Patch to cut out the "long long" test in pcrecpp_unittest when
    HAVE_LONG_LONG is not defined.

3.  Applied Christian Ehrlicher's patch to update the CMake build files to
    bring them up to date and include new features. This patch includes:

    - Fixed PH's badly added libz and libbz2 support.
    - Fixed a problem with static linking.
    - Added pcredemo. [But later removed - see 7 below.]
    - Fixed dftables problem and added an option.
    - Added a number of HAVE_XXX tests, including HAVE_WINDOWS_H and
        HAVE_LONG_LONG.
    - Added readline support for pcretest.
    - Added an listing of the option settings after cmake has run.

4.  A user submitted a patch to Makefile that makes it easy to create
    "pcre.dll" under mingw when using Configure/Make. I added stuff to
    Makefile.am that cause it to include this special target, without
    affecting anything else. Note that the same mingw target plus all
    the other distribution libraries and programs are now supported
    when configuring with CMake (see 6 below) instead of with
    Configure/Make.

5.  Applied Craig's patch that moves no_arg into the RE class in the C++ code.
    This is an attempt to solve the reported problem "pcrecpp::no_arg is not
    exported in the Windows port". It has not yet been confirmed that the patch
    solves the problem, but it does no harm.

6.  Applied Sheri's patch to CMakeLists.txt to add NON_STANDARD_LIB_PREFIX and
    NON_STANDARD_LIB_SUFFIX for dll names built with mingw when configured
    with CMake, and also correct the comment about stack recursion.

7.  Remove the automatic building of pcredemo from the ./configure system and
    from CMakeLists.txt. The whole idea of pcredemo.c is that it is an example
    of a program that users should build themselves after PCRE is installed, so
    building it automatically is not really right. What is more, it gave
    trouble in some build environments.

8.  Further tidies to CMakeLists.txt from Sheri and Christian.


Version 7.5 10-Jan-08
---------------------

1.  Applied a patch from Craig: "This patch makes it possible to 'ignore'
    values in parens when parsing an RE using the C++ wrapper."

2.  Negative specials like \S did not work in character classes in UTF-8 mode.
    Characters greater than 255 were excluded from the class instead of being
    included.

3.  The same bug as (2) above applied to negated POSIX classes such as
    [:^space:].

4.  PCRECPP_STATIC was referenced in pcrecpp_internal.h, but nowhere was it
    defined or documented. It seems to have been a typo for PCRE_STATIC, so
    I have changed it.

5.  The construct (?&) was not diagnosed as a syntax error (it referenced the
    first named subpattern) and a construct such as (?&a) would reference the
    first named subpattern whose name started with "a" (in other words, the
    length check was missing). Both these problems are fixed. "Subpattern name
    expected" is now given for (?&) (a zero-length name), and this patch also
    makes it give the same error for \k'' (previously it complained that that
    was a reference to a non-existent subpattern).

6.  The erroneous patterns (?+-a) and (?-+a) give different error messages;
    this is right because (?- can be followed by option settings as well as by
    digits. I have, however, made the messages clearer.

7.  Patterns such as (?(1)a|b) (a pattern that contains fewer subpatterns
    than the number used in the conditional) now cause a compile-time error.
    This is actually not compatible with Perl, which accepts such patterns, but
    treats the conditional as always being FALSE (as PCRE used to), but it
    seems to me that giving a diagnostic is better.

8.  Change "alphameric" to the more common word "alphanumeric" in comments
    and messages.

9.  Fix two occurrences of "backslash" in comments that should have been
    "backspace".

10. Remove two redundant lines of code that can never be obeyed (their function
    was moved elsewhere).

11. The program that makes PCRE's Unicode character property table had a bug
    which caused it to generate incorrect table entries for sequences of
    characters that have the same character type, but are in different scripts.
    It amalgamated them into a single range, with the script of the first of
    them. In other words, some characters were in the wrong script. There were
    thirteen such cases, affecting characters in the following ranges:

      U+002b0 - U+002c1
      U+0060c - U+0060d
      U+0061e - U+00612
      U+0064b - U+0065e
      U+0074d - U+0076d
      U+01800 - U+01805
      U+01d00 - U+01d77
      U+01d9b - U+01dbf
      U+0200b - U+0200f
      U+030fc - U+030fe
      U+03260 - U+0327f
      U+0fb46 - U+0fbb1
      U+10450 - U+1049d

12. The -o option (show only the matching part of a line) for pcregrep was not
    compatible with GNU grep in that, if there was more than one match in a
    line, it showed only the first of them. It now behaves in the same way as
    GNU grep.

13. If the -o and -v options were combined for pcregrep, it printed a blank
    line for every non-matching line. GNU grep prints nothing, and pcregrep now
    does the same. The return code can be used to tell if there were any
    non-matching lines.

14. Added --file-offsets and --line-offsets to pcregrep.

15. The pattern (?=something)(?R) was not being diagnosed as a potentially
    infinitely looping recursion. The bug was that positive lookaheads were not
    being skipped when checking for a possible empty match (negative lookaheads
    and both kinds of lookbehind were skipped).

16. Fixed two typos in the Windows-only code in pcregrep.c, and moved the
    inclusion of <windows.h> to before rather than after the definition of
    INVALID_FILE_ATTRIBUTES (patch from David Byron).

17. Specifying a possessive quantifier with a specific limit for a Unicode
    character property caused pcre_compile() to compile bad code, which led at
    runtime to PCRE_ERROR_INTERNAL (-14). Examples of patterns that caused this
    are: /\p{Zl}{2,3}+/8 and /\p{Cc}{2}+/8. It was the possessive "+" that
    caused the error; without that there was no problem.

18. Added --enable-pcregrep-libz and --enable-pcregrep-libbz2.

19. Added --enable-pcretest-libreadline.

20. In pcrecpp.cc, the variable 'count' was incremented twice in
    RE::GlobalReplace(). As a result, the number of replacements returned was
    double what it should be. I removed one of the increments, but Craig sent a
    later patch that removed the other one (the right fix) and added unit tests
    that check the return values (which was not done before).

21. Several CMake things:

    (1) Arranged that, when cmake is used on Unix, the libraries end up with
        the names libpcre and libpcreposix, not just pcre and pcreposix.

    (2) The above change means that pcretest and pcregrep are now correctly
        linked with the newly-built libraries, not previously installed ones.

    (3) Added PCRE_SUPPORT_LIBREADLINE, PCRE_SUPPORT_LIBZ, PCRE_SUPPORT_LIBBZ2.

22. In UTF-8 mode, with newline set to "any", a pattern such as .*a.*=.b.*
    crashed when matching a string such as a\x{2029}b (note that \x{2029} is a
    UTF-8 newline character). The key issue is that the pattern starts .*;
    this means that the match must be either at the beginning, or after a
    newline. The bug was in the code for advancing after a failed match and
    checking that the new position followed a newline. It was not taking
    account of UTF-8 characters correctly.

23. PCRE was behaving differently from Perl in the way it recognized POSIX
    character classes. PCRE was not treating the sequence [:...:] as a
    character class unless the ... were all letters. Perl, however, seems to
    allow any characters between [: and :], though of course it rejects as
    unknown any "names" that contain non-letters, because all the known class
    names consist only of letters. Thus, Perl gives an error for [[:1234:]],
    for example, whereas PCRE did not - it did not recognize a POSIX character
    class. This seemed a bit dangerous, so the code has been changed to be
    closer to Perl. The behaviour is not identical to Perl, because PCRE will
    diagnose an unknown class for, for example, [[:l\ower:]] where Perl will
    treat it as [[:lower:]]. However, PCRE does now give "unknown" errors where
    Perl does, and where it didn't before.

24. Rewrite so as to remove the single use of %n from pcregrep because in some
    Windows environments %n is disabled by default.


Version 7.4 21-Sep-07
---------------------

1.  Change 7.3/28 was implemented for classes by looking at the bitmap. This
    means that a class such as [\s] counted as "explicit reference to CR or
    LF". That isn't really right - the whole point of the change was to try to
    help when there was an actual mention of one of the two characters. So now
    the change happens only if \r or \n (or a literal CR or LF) character is
    encountered.

2.  The 32-bit options word was also used for 6 internal flags, but the numbers
    of both had grown to the point where there were only 3 bits left.
    Fortunately, there was spare space in the data structure, and so I have
    moved the internal flags into a new 16-bit field to free up more option
    bits.

3.  The appearance of (?J) at the start of a pattern set the DUPNAMES option,
    but did not set the internal JCHANGED flag - either of these is enough to
    control the way the "get" function works - but the PCRE_INFO_JCHANGED
    facility is supposed to tell if (?J) was ever used, so now (?J) at the
    start sets both bits.

4.  Added options (at build time, compile time, exec time) to change \R from
    matching any Unicode line ending sequence to just matching CR, LF, or CRLF.

5.  doc/pcresyntax.html was missing from the distribution.

6.  Put back the definition of PCRE_ERROR_NULLWSLIMIT, for backward
    compatibility, even though it is no longer used.

7.  Added macro for snprintf to pcrecpp_unittest.cc and also for strtoll and
    strtoull to pcrecpp.cc to select the available functions in WIN32 when the
    windows.h file is present (where different names are used). [This was
    reversed later after testing - see 16 below.]

8.  Changed all #include <config.h> to #include "config.h". There were also
    some further <pcre.h> cases that I changed to "pcre.h".

9.  When pcregrep was used with the --colour option, it missed the line ending
    sequence off the lines that it output.

10. It was pointed out to me that arrays of string pointers cause lots of
    relocations when a shared library is dynamically loaded. A technique of
    using a single long string with a table of offsets can drastically reduce
    these. I have refactored PCRE in four places to do this. The result is
    dramatic:

      Originally:                          290
      After changing UCP table:            187
      After changing error message table:   43
      After changing table of "verbs"       36
      After changing table of Posix names   22

    Thanks to the folks working on Gregex for glib for this insight.

11. --disable-stack-for-recursion caused compiling to fail unless -enable-
    unicode-properties was also set.

12. Updated the tests so that they work when \R is defaulted to ANYCRLF.

13. Added checks for ANY and ANYCRLF to pcrecpp.cc where it previously
    checked only for CRLF.

14. Added casts to pcretest.c to avoid compiler warnings.

15. Added Craig's patch to various pcrecpp modules to avoid compiler warnings.

16. Added Craig's patch to remove the WINDOWS_H tests, that were not working,
    and instead check for _strtoi64 explicitly, and avoid the use of snprintf()
    entirely. This removes changes made in 7 above.

17. The CMake files have been updated, and there is now more information about
    building with CMake in the NON-UNIX-USE document.


Version 7.3 28-Aug-07
---------------------

 1. In the rejigging of the build system that eventually resulted in 7.1, the
    line "#include <pcre.h>" was included in pcre_internal.h. The use of angle
    brackets there is not right, since it causes compilers to look for an
    installed pcre.h, not the version that is in the source that is being
    compiled (which of course may be different). I have changed it back to:

      #include "pcre.h"

    I have a vague recollection that the change was concerned with compiling in
    different directories, but in the new build system, that is taken care of
    by the VPATH setting the Makefile.

 2. The pattern .*$ when run in not-DOTALL UTF-8 mode with newline=any failed
    when the subject happened to end in the byte 0x85 (e.g. if the last
    character was \x{1ec5}). *Character* 0x85 is one of the "any" newline
    characters but of course it shouldn't be taken as a newline when it is part
    of another character. The bug was that, for an unlimited repeat of . in
    not-DOTALL UTF-8 mode, PCRE was advancing by bytes rather than by
    characters when looking for a newline.

 3. A small performance improvement in the DOTALL UTF-8 mode .* case.

 4. Debugging: adjusted the names of opcodes for different kinds of parentheses
    in debug output.

 5. Arrange to use "%I64d" instead of "%lld" and "%I64u" instead of "%llu" for
    long printing in the pcrecpp unittest when running under MinGW.

 6. ESC_K was left out of the EBCDIC table.

 7. Change 7.0/38 introduced a new limit on the number of nested non-capturing
    parentheses; I made it 1000, which seemed large enough. Unfortunately, the
    limit also applies to "virtual nesting" when a pattern is recursive, and in
    this case 1000 isn't so big. I have been able to remove this limit at the
    expense of backing off one optimization in certain circumstances. Normally,
    when pcre_exec() would call its internal match() function recursively and
    immediately return the result unconditionally, it uses a "tail recursion"
    feature to save stack. However, when a subpattern that can match an empty
    string has an unlimited repetition quantifier, it no longer makes this
    optimization. That gives it a stack frame in which to save the data for
    checking that an empty string has been matched. Previously this was taken
    from the 1000-entry workspace that had been reserved. So now there is no
    explicit limit, but more stack is used.

 8. Applied Daniel's patches to solve problems with the import/export magic
    syntax that is required for Windows, and which was going wrong for the
    pcreposix and pcrecpp parts of the library. These were overlooked when this
    problem was solved for the main library.

 9. There were some crude static tests to avoid integer overflow when computing
    the size of patterns that contain repeated groups with explicit upper
    limits. As the maximum quantifier is 65535, the maximum group length was
    set at 30,000 so that the product of these two numbers did not overflow a
    32-bit integer. However, it turns out that people want to use groups that
    are longer than 30,000 bytes (though not repeat them that many times).
    Change 7.0/17 (the refactoring of the way the pattern size is computed) has
    made it possible to implement the integer overflow checks in a much more
    dynamic way, which I have now done. The artificial limitation on group
    length has been removed - we now have only the limit on the total length of
    the compiled pattern, which depends on the LINK_SIZE setting.

10. Fixed a bug in the documentation for get/copy named substring when
    duplicate names are permitted. If none of the named substrings are set, the
    functions return PCRE_ERROR_NOSUBSTRING (7); the doc said they returned an
    empty string.

11. Because Perl interprets \Q...\E at a high level, and ignores orphan \E
    instances, patterns such as [\Q\E] or [\E] or even [^\E] cause an error,
    because the ] is interpreted as the first data character and the
    terminating ] is not found. PCRE has been made compatible with Perl in this
    regard. Previously, it interpreted [\Q\E] as an empty class, and [\E] could
    cause memory overwriting.

10. Like Perl, PCRE automatically breaks an unlimited repeat after an empty
    string has been matched (to stop an infinite loop). It was not recognizing
    a conditional subpattern that could match an empty string if that
    subpattern was within another subpattern. For example, it looped when
    trying to match  (((?(1)X|))*)  but it was OK with  ((?(1)X|)*)  where the
    condition was not nested. This bug has been fixed.

12. A pattern like \X?\d or \P{L}?\d in non-UTF-8 mode could cause a backtrack
    past the start of the subject in the presence of bytes with the top bit
    set, for example "\x8aBCD".

13. Added Perl 5.10 experimental backtracking controls (*FAIL), (*F), (*PRUNE),
    (*SKIP), (*THEN), (*COMMIT), and (*ACCEPT).

14. Optimized (?!) to (*FAIL).

15. Updated the test for a valid UTF-8 string to conform to the later RFC 3629.
    This restricts code points to be within the range 0 to 0x10FFFF, excluding
    the "low surrogate" sequence 0xD800 to 0xDFFF. Previously, PCRE allowed the
    full range 0 to 0x7FFFFFFF, as defined by RFC 2279. Internally, it still
    does: it's just the validity check that is more restrictive.

16. Inserted checks for integer overflows during escape sequence (backslash)
    processing, and also fixed erroneous offset values for syntax errors during
    backslash processing.

17. Fixed another case of looking too far back in non-UTF-8 mode (cf 12 above)
    for patterns like [\PPP\x8a]{1,}\x80 with the subject "A\x80".

18. An unterminated class in a pattern like (?1)\c[ with a "forward reference"
    caused an overrun.

19. A pattern like (?:[\PPa*]*){8,} which had an "extended class" (one with
    something other than just ASCII characters) inside a group that had an
    unlimited repeat caused a loop at compile time (while checking to see
    whether the group could match an empty string).

20. Debugging a pattern containing \p or \P could cause a crash. For example,
    [\P{Any}] did so. (Error in the code for printing property names.)

21. An orphan \E inside a character class could cause a crash.

22. A repeated capturing bracket such as (A)? could cause a wild memory
    reference during compilation.

23. There are several functions in pcre_compile() that scan along a compiled
    expression for various reasons (e.g. to see if it's fixed length for look
    behind). There were bugs in these functions when a repeated \p or \P was
    present in the pattern. These operators have additional parameters compared
    with \d, etc, and these were not being taken into account when moving along
    the compiled data. Specifically:

    (a) A item such as \p{Yi}{3} in a lookbehind was not treated as fixed
        length.

    (b) An item such as \pL+ within a repeated group could cause crashes or
        loops.

    (c) A pattern such as \p{Yi}+(\P{Yi}+)(?1) could give an incorrect
        "reference to non-existent subpattern" error.

    (d) A pattern like (\P{Yi}{2}\277)? could loop at compile time.

24. A repeated \S or \W in UTF-8 mode could give wrong answers when multibyte
    characters were involved (for example /\S{2}/8g with "A\x{a3}BC").

25. Using pcregrep in multiline, inverted mode (-Mv) caused it to loop.

26. Patterns such as [\P{Yi}A] which include \p or \P and just one other
    character were causing crashes (broken optimization).

27. Patterns such as (\P{Yi}*\277)* (group with possible zero repeat containing
    \p or \P) caused a compile-time loop.

28. More problems have arisen in unanchored patterns when CRLF is a valid line
    break. For example, the unstudied pattern [\r\n]A does not match the string
    "\r\nA" because change 7.0/46 below moves the current point on by two
    characters after failing to match at the start. However, the pattern \nA
    *does* match, because it doesn't start till \n, and if [\r\n]A is studied,
    the same is true. There doesn't seem any very clean way out of this, but
    what I have chosen to do makes the common cases work: PCRE now takes note
    of whether there can be an explicit match for \r or \n anywhere in the
    pattern, and if so, 7.0/46 no longer applies. As part of this change,
    there's a new PCRE_INFO_HASCRORLF option for finding out whether a compiled
    pattern has explicit CR or LF references.

29. Added (*CR) etc for changing newline setting at start of pattern.


Version 7.2 19-Jun-07
---------------------

 1. If the fr_FR locale cannot be found for test 3, try the "french" locale,
    which is apparently normally available under Windows.

 2. Re-jig the pcregrep tests with different newline settings in an attempt
    to make them independent of the local environment's newline setting.

 3. Add code to configure.ac to remove -g from the CFLAGS default settings.

 4. Some of the "internals" tests were previously cut out when the link size
    was not 2, because the output contained actual offsets. The recent new
    "Z" feature of pcretest means that these can be cut out, making the tests
    usable with all link sizes.

 5. Implemented Stan Switzer's goto replacement for longjmp() when not using
    stack recursion. This gives a massive performance boost under BSD, but just
    a small improvement under Linux. However, it saves one field in the frame
    in all cases.

 6. Added more features from the forthcoming Perl 5.10:

    (a) (?-n) (where n is a string of digits) is a relative subroutine or
        recursion call. It refers to the nth most recently opened parentheses.

    (b) (?+n) is also a relative subroutine call; it refers to the nth next
        to be opened parentheses.

    (c) Conditions that refer to capturing parentheses can be specified
        relatively, for example, (?(-2)... or (?(+3)...

    (d) \K resets the start of the current match so that everything before
        is not part of it.

    (e) \k{name} is synonymous with \k<name> and \k'name' (.NET compatible).

    (f) \g{name} is another synonym - part of Perl 5.10's unification of
        reference syntax.

    (g) (?| introduces a group in which the numbering of parentheses in each
        alternative starts with the same number.

    (h) \h, \H, \v, and \V match horizontal and vertical whitespace.

 7. Added two new calls to pcre_fullinfo(): PCRE_INFO_OKPARTIAL and
    PCRE_INFO_JCHANGED.

 8. A pattern such as  (.*(.)?)*  caused pcre_exec() to fail by either not
    terminating or by crashing. Diagnosed by Viktor Griph; it was in the code
    for detecting groups that can match an empty string.

 9. A pattern with a very large number of alternatives (more than several
    hundred) was running out of internal workspace during the pre-compile
    phase, where pcre_compile() figures out how much memory will be needed. A
    bit of new cunning has reduced the workspace needed for groups with
    alternatives. The 1000-alternative test pattern now uses 12 bytes of
    workspace instead of running out of the 4096 that are available.

10. Inserted some missing (unsigned int) casts to get rid of compiler warnings.

11. Applied patch from Google to remove an optimization that didn't quite work.
    The report of the bug said:

      pcrecpp::RE("a*").FullMatch("aaa") matches, while
      pcrecpp::RE("a*?").FullMatch("aaa") does not, and
      pcrecpp::RE("a*?\\z").FullMatch("aaa") does again.

12. If \p or \P was used in non-UTF-8 mode on a character greater than 127
    it matched the wrong number of bytes.


Version 7.1 24-Apr-07
---------------------

 1. Applied Bob Rossi and Daniel G's patches to convert the build system to one
    that is more "standard", making use of automake and other Autotools. There
    is some re-arrangement of the files and adjustment of comments consequent
    on this.

 2. Part of the patch fixed a problem with the pcregrep tests. The test of -r
    for recursive directory scanning broke on some systems because the files
    are not scanned in any specific order and on different systems the order
    was different. A call to "sort" has been inserted into RunGrepTest for the
    approprate test as a short-term fix. In the longer term there may be an
    alternative.

 3. I had an email from Eric Raymond about problems translating some of PCRE's
    man pages to HTML (despite the fact that I distribute HTML pages, some
    people do their own conversions for various reasons). The problems
    concerned the use of low-level troff macros .br and .in. I have therefore
    removed all such uses from the man pages (some were redundant, some could
    be replaced by .nf/.fi pairs). The 132html script that I use to generate
    HTML has been updated to handle .nf/.fi and to complain if it encounters
    .br or .in.

 4. Updated comments in configure.ac that get placed in config.h.in and also
    arranged for config.h to be included in the distribution, with the name
    config.h.generic, for the benefit of those who have to compile without
    Autotools (compare pcre.h, which is now distributed as pcre.h.generic).

 5. Updated the support (such as it is) for Virtual Pascal, thanks to Stefan
    Weber: (1) pcre_internal.h was missing some function renames; (2) updated
    makevp.bat for the current PCRE, using the additional files
    makevp_c.txt, makevp_l.txt, and pcregexp.pas.

 6. A Windows user reported a minor discrepancy with test 2, which turned out
    to be caused by a trailing space on an input line that had got lost in his
    copy. The trailing space was an accident, so I've just removed it.

 7. Add -Wl,-R... flags in pcre-config.in for *BSD* systems, as I'm told
    that is needed.

 8. Mark ucp_table (in ucptable.h) and ucp_gentype (in pcre_ucp_searchfuncs.c)
    as "const" (a) because they are and (b) because it helps the PHP
    maintainers who have recently made a script to detect big data structures
    in the php code that should be moved to the .rodata section. I remembered
    to update Builducptable as well, so it won't revert if ucptable.h is ever
    re-created.

 9. Added some extra #ifdef SUPPORT_UTF8 conditionals into pcretest.c,
    pcre_printint.src, pcre_compile.c, pcre_study.c, and pcre_tables.c, in
    order to be able to cut out the UTF-8 tables in the latter when UTF-8
    support is not required. This saves 1.5-2K of code, which is important in
    some applications.

    Later: more #ifdefs are needed in pcre_ord2utf8.c and pcre_valid_utf8.c
    so as not to refer to the tables, even though these functions will never be
    called when UTF-8 support is disabled. Otherwise there are problems with a
    shared library.

10. Fixed two bugs in the emulated memmove() function in pcre_internal.h:

    (a) It was defining its arguments as char * instead of void *.

    (b) It was assuming that all moves were upwards in memory; this was true
        a long time ago when I wrote it, but is no longer the case.

    The emulated memove() is provided for those environments that have neither
    memmove() nor bcopy(). I didn't think anyone used it these days, but that
    is clearly not the case, as these two bugs were recently reported.

11. The script PrepareRelease is now distributed: it calls 132html, CleanTxt,
    and Detrail to create the HTML documentation, the .txt form of the man
    pages, and it removes trailing spaces from listed files. It also creates
    pcre.h.generic and config.h.generic from pcre.h and config.h. In the latter
    case, it wraps all the #defines with #ifndefs. This script should be run
    before "make dist".

12. Fixed two fairly obscure bugs concerned with quantified caseless matching
    with Unicode property support.

    (a) For a maximizing quantifier, if the two different cases of the
        character were of different lengths in their UTF-8 codings (there are
        some cases like this - I found 11), and the matching function had to
        back up over a mixture of the two cases, it incorrectly assumed they
        were both the same length.

    (b) When PCRE was configured to use the heap rather than the stack for
        recursion during matching, it was not correctly preserving the data for
        the other case of a UTF-8 character when checking ahead for a match
        while processing a minimizing repeat. If the check also involved
        matching a wide character, but failed, corruption could cause an
        erroneous result when trying to check for a repeat of the original
        character.

13. Some tidying changes to the testing mechanism:

    (a) The RunTest script now detects the internal link size and whether there
        is UTF-8 and UCP support by running ./pcretest -C instead of relying on
        values substituted by "configure". (The RunGrepTest script already did
        this for UTF-8.) The configure.ac script no longer substitutes the
        relevant variables.

    (b) The debugging options /B and /D in pcretest show the compiled bytecode
        with length and offset values. This means that the output is different
        for different internal link sizes. Test 2 is skipped for link sizes
        other than 2 because of this, bypassing the problem. Unfortunately,
        there was also a test in test 3 (the locale tests) that used /B and
        failed for link sizes other than 2. Rather than cut the whole test out,
        I have added a new /Z option to pcretest that replaces the length and
        offset values with spaces. This is now used to make test 3 independent
        of link size. (Test 2 will be tidied up later.)

14. If erroroffset was passed as NULL to pcre_compile, it provoked a
    segmentation fault instead of returning the appropriate error message.

15. In multiline mode when the newline sequence was set to "any", the pattern
    ^$ would give a match between the \r and \n of a subject such as "A\r\nB".
    This doesn't seem right; it now treats the CRLF combination as the line
    ending, and so does not match in that case. It's only a pattern such as ^$
    that would hit this one: something like ^ABC$ would have failed after \r
    and then tried again after \r\n.

16. Changed the comparison command for RunGrepTest from "diff -u" to "diff -ub"
    in an attempt to make files that differ only in their line terminators
    compare equal. This works on Linux.

17. Under certain error circumstances pcregrep might try to free random memory
    as it exited. This is now fixed, thanks to valgrind.

19. In pcretest, if the pattern /(?m)^$/g<any> was matched against the string
    "abc\r\n\r\n", it found an unwanted second match after the second \r. This
    was because its rules for how to advance for /g after matching an empty
    string at the end of a line did not allow for this case. They now check for
    it specially.

20. pcretest is supposed to handle patterns and data of any length, by
    extending its buffers when necessary. It was getting this wrong when the
    buffer for a data line had to be extended.

21. Added PCRE_NEWLINE_ANYCRLF which is like ANY, but matches only CR, LF, or
    CRLF as a newline sequence.

22. Code for handling Unicode properties in pcre_dfa_exec() wasn't being cut
    out by #ifdef SUPPORT_UCP. This did no harm, as it could never be used, but
    I have nevertheless tidied it up.

23. Added some casts to kill warnings from HP-UX ia64 compiler.

24. Added a man page for pcre-config.


Version 7.0 19-Dec-06
---------------------

 1. Fixed a signed/unsigned compiler warning in pcre_compile.c, shown up by
    moving to gcc 4.1.1.

 2. The -S option for pcretest uses setrlimit(); I had omitted to #include
    sys/time.h, which is documented as needed for this function. It doesn't
    seem to matter on Linux, but it showed up on some releases of OS X.

 3. It seems that there are systems where bytes whose values are greater than
    127 match isprint() in the "C" locale. The "C" locale should be the
    default when a C program starts up. In most systems, only ASCII printing
    characters match isprint(). This difference caused the output from pcretest
    to vary, making some of the tests fail. I have changed pcretest so that:

    (a) When it is outputting text in the compiled version of a pattern, bytes
        other than 32-126 are always shown as hex escapes.

    (b) When it is outputting text that is a matched part of a subject string,
        it does the same, unless a different locale has been set for the match
        (using the /L modifier). In this case, it uses isprint() to decide.

 4. Fixed a major bug that caused incorrect computation of the amount of memory
    required for a compiled pattern when options that changed within the
    pattern affected the logic of the preliminary scan that determines the
    length. The relevant options are -x, and -i in UTF-8 mode. The result was
    that the computed length was too small. The symptoms of this bug were
    either the PCRE error "internal error: code overflow" from pcre_compile(),
    or a glibc crash with a message such as "pcretest: free(): invalid next
    size (fast)". Examples of patterns that provoked this bug (shown in
    pcretest format) are:

      /(?-x: )/x
      /(?x)(?-x: \s*#\s*)/
      /((?i)[\x{c0}])/8
      /(?i:[\x{c0}])/8

    HOWEVER: Change 17 below makes this fix obsolete as the memory computation
    is now done differently.

 5. Applied patches from Google to: (a) add a QuoteMeta function to the C++
    wrapper classes; (b) implement a new function in the C++ scanner that is
    more efficient than the old way of doing things because it avoids levels of
    recursion in the regex matching; (c) add a paragraph to the documentation
    for the FullMatch() function.

 6. The escape sequence \n was being treated as whatever was defined as
    "newline". Not only was this contrary to the documentation, which states
    that \n is character 10 (hex 0A), but it also went horribly wrong when
    "newline" was defined as CRLF. This has been fixed.

 7. In pcre_dfa_exec.c the value of an unsigned integer (the variable called c)
    was being set to -1 for the "end of line" case (supposedly a value that no
    character can have). Though this value is never used (the check for end of
    line is "zero bytes in current character"), it caused compiler complaints.
    I've changed it to 0xffffffff.

 8. In pcre_version.c, the version string was being built by a sequence of
    C macros that, in the event of PCRE_PRERELEASE being defined as an empty
    string (as it is for production releases) called a macro with an empty
    argument. The C standard says the result of this is undefined. The gcc
    compiler treats it as an empty string (which was what was wanted) but it is
    reported that Visual C gives an error. The source has been hacked around to
    avoid this problem.

 9. On the advice of a Windows user, included <io.h> and <fcntl.h> in Windows
    builds of pcretest, and changed the call to _setmode() to use _O_BINARY
    instead of 0x8000. Made all the #ifdefs test both _WIN32 and WIN32 (not all
    of them did).

10. Originally, pcretest opened its input and output without "b"; then I was
    told that "b" was needed in some environments, so it was added for release
    5.0 to both the input and output. (It makes no difference on Unix-like
    systems.) Later I was told that it is wrong for the input on Windows. I've
    now abstracted the modes into two macros, to make it easier to fiddle with
    them, and removed "b" from the input mode under Windows.

11. Added pkgconfig support for the C++ wrapper library, libpcrecpp.

12. Added -help and --help to pcretest as an official way of being reminded
    of the options.

13. Removed some redundant semicolons after macro calls in pcrecpparg.h.in
    and pcrecpp.cc because they annoy compilers at high warning levels.

14. A bit of tidying/refactoring in pcre_exec.c in the main bumpalong loop.

15. Fixed an occurrence of == in configure.ac that should have been = (shell
    scripts are not C programs :-) and which was not noticed because it works
    on Linux.

16. pcretest is supposed to handle any length of pattern and data line (as one
    line or as a continued sequence of lines) by extending its input buffer if
    necessary. This feature was broken for very long pattern lines, leading to
    a string of junk being passed to pcre_compile() if the pattern was longer
    than about 50K.

17. I have done a major re-factoring of the way pcre_compile() computes the
    amount of memory needed for a compiled pattern. Previously, there was code
    that made a preliminary scan of the pattern in order to do this. That was
    OK when PCRE was new, but as the facilities have expanded, it has become
    harder and harder to keep it in step with the real compile phase, and there
    have been a number of bugs (see for example, 4 above). I have now found a
    cunning way of running the real compile function in a "fake" mode that
    enables it to compute how much memory it would need, while actually only
    ever using a few hundred bytes of working memory and without too many
    tests of the mode. This should make future maintenance and development
    easier. A side effect of this work is that the limit of 200 on the nesting
    depth of parentheses has been removed (though this was never a serious
    limitation, I suspect). However, there is a downside: pcre_compile() now
    runs more slowly than before (30% or more, depending on the pattern). I
    hope this isn't a big issue. There is no effect on runtime performance.

18. Fixed a minor bug in pcretest: if a pattern line was not terminated by a
    newline (only possible for the last line of a file) and it was a
    pattern that set a locale (followed by /Lsomething), pcretest crashed.

19. Added additional timing features to pcretest. (1) The -tm option now times
    matching only, not compiling. (2) Both -t and -tm can be followed, as a
    separate command line item, by a number that specifies the number of
    repeats to use when timing. The default is 50000; this gives better
    precision, but takes uncomfortably long for very large patterns.

20. Extended pcre_study() to be more clever in cases where a branch of a
    subpattern has no definite first character. For example, (a*|b*)[cd] would
    previously give no result from pcre_study(). Now it recognizes that the
    first character must be a, b, c, or d.

21. There was an incorrect error "recursive call could loop indefinitely" if
    a subpattern (or the entire pattern) that was being tested for matching an
    empty string contained only one non-empty item after a nested subpattern.
    For example, the pattern (?>\x{100}*)\d(?R) provoked this error
    incorrectly, because the \d was being skipped in the check.

22. The pcretest program now has a new pattern option /B and a command line
    option -b, which is equivalent to adding /B to every pattern. This causes
    it to show the compiled bytecode, without the additional information that
    -d shows. The effect of -d is now the same as -b with -i (and similarly, /D
    is the same as /B/I).

23. A new optimization is now able automatically to treat some sequences such
    as a*b as a*+b. More specifically, if something simple (such as a character
    or a simple class like \d) has an unlimited quantifier, and is followed by
    something that cannot possibly match the quantified thing, the quantifier
    is automatically "possessified".

24. A recursive reference to a subpattern whose number was greater than 39
    went wrong under certain circumstances in UTF-8 mode. This bug could also
    have affected the operation of pcre_study().

25. Realized that a little bit of performance could be had by replacing
    (c & 0xc0) == 0xc0 with c >= 0xc0 when processing UTF-8 characters.

26. Timing data from pcretest is now shown to 4 decimal places instead of 3.

27. Possessive quantifiers such as a++ were previously implemented by turning
    them into atomic groups such as ($>a+). Now they have their own opcodes,
    which improves performance. This includes the automatically created ones
    from 23 above.

28. A pattern such as (?=(\w+))\1: which simulates an atomic group using a
    lookahead was broken if it was not anchored. PCRE was mistakenly expecting
    the first matched character to be a colon. This applied both to named and
    numbered groups.

29. The ucpinternal.h header file was missing its idempotency #ifdef.

30. I was sent a "project" file called libpcre.a.dev which I understand makes
    building PCRE on Windows easier, so I have included it in the distribution.

31. There is now a check in pcretest against a ridiculously large number being
    returned by pcre_exec() or pcre_dfa_exec(). If this happens in a /g or /G
    loop, the loop is abandoned.

32. Forward references to subpatterns in conditions such as (?(2)...) where
    subpattern 2 is defined later cause pcre_compile() to search forwards in
    the pattern for the relevant set of parentheses. This search went wrong
    when there were unescaped parentheses in a character class, parentheses
    escaped with \Q...\E, or parentheses in a #-comment in /x mode.

33. "Subroutine" calls and backreferences were previously restricted to
    referencing subpatterns earlier in the regex. This restriction has now
    been removed.

34. Added a number of extra features that are going to be in Perl 5.10. On the
    whole, these are just syntactic alternatives for features that PCRE had
    previously implemented using the Python syntax or my own invention. The
    other formats are all retained for compatibility.

    (a) Named groups can now be defined as (?<name>...) or (?'name'...) as well
        as (?P<name>...). The new forms, as well as being in Perl 5.10, are
        also .NET compatible.

    (b) A recursion or subroutine call to a named group can now be defined as
        (?&name) as well as (?P>name).

    (c) A backreference to a named group can now be defined as \k<name> or
        \k'name' as well as (?P=name). The new forms, as well as being in Perl
        5.10, are also .NET compatible.

    (d) A conditional reference to a named group can now use the syntax
        (?(<name>) or (?('name') as well as (?(name).

    (e) A "conditional group" of the form (?(DEFINE)...) can be used to define
        groups (named and numbered) that are never evaluated inline, but can be
        called as "subroutines" from elsewhere. In effect, the DEFINE condition
        is always false. There may be only one alternative in such a group.

    (f) A test for recursion can be given as (?(R1).. or (?(R&name)... as well
        as the simple (?(R). The condition is true only if the most recent
        recursion is that of the given number or name. It does not search out
        through the entire recursion stack.

    (g) The escape \gN or \g{N} has been added, where N is a positive or
        negative number, specifying an absolute or relative reference.

35. Tidied to get rid of some further signed/unsigned compiler warnings and
    some "unreachable code" warnings.

36. Updated the Unicode property tables to Unicode version 5.0.0. Amongst other
    things, this adds five new scripts.

37. Perl ignores orphaned \E escapes completely. PCRE now does the same.
    There were also incompatibilities regarding the handling of \Q..\E inside
    character classes, for example with patterns like [\Qa\E-\Qz\E] where the
    hyphen was adjacent to \Q or \E. I hope I've cleared all this up now.

38. Like Perl, PCRE detects when an indefinitely repeated parenthesized group
    matches an empty string, and forcibly breaks the loop. There were bugs in
    this code in non-simple cases. For a pattern such as  ^(a()*)*  matched
    against  aaaa  the result was just "a" rather than "aaaa", for example. Two
    separate and independent bugs (that affected different cases) have been
    fixed.

39. Refactored the code to abolish the use of different opcodes for small
    capturing bracket numbers. This is a tidy that I avoided doing when I
    removed the limit on the number of capturing brackets for 3.5 back in 2001.
    The new approach is not only tidier, it makes it possible to reduce the
    memory needed to fix the previous bug (38).

40. Implemented PCRE_NEWLINE_ANY to recognize any of the Unicode newline
    sequences (http://unicode.org/unicode/reports/tr18/) as "newline" when
    processing dot, circumflex, or dollar metacharacters, or #-comments in /x
    mode.

41. Add \R to match any Unicode newline sequence, as suggested in the Unicode
    report.

42. Applied patch, originally from Ari Pollak, modified by Google, to allow
    copy construction and assignment in the C++ wrapper.

43. Updated pcregrep to support "--newline=any". In the process, I fixed a
    couple of bugs that could have given wrong results in the "--newline=crlf"
    case.

44. Added a number of casts and did some reorganization of signed/unsigned int
    variables following suggestions from Dair Grant. Also renamed the variable
    "this" as "item" because it is a C++ keyword.

45. Arranged for dftables to add

      #include "pcre_internal.h"

    to pcre_chartables.c because without it, gcc 4.x may remove the array
    definition from the final binary if PCRE is built into a static library and
    dead code stripping is activated.

46. For an unanchored pattern, if a match attempt fails at the start of a
    newline sequence, and the newline setting is CRLF or ANY, and the next two
    characters are CRLF, advance by two characters instead of one.


Version 6.7 04-Jul-06
---------------------

 1. In order to handle tests when input lines are enormously long, pcretest has
    been re-factored so that it automatically extends its buffers when
    necessary. The code is crude, but this _is_ just a test program. The
    default size has been increased from 32K to 50K.

 2. The code in pcre_study() was using the value of the re argument before
    testing it for NULL. (Of course, in any sensible call of the function, it
    won't be NULL.)

 3. The memmove() emulation function in pcre_internal.h, which is used on
    systems that lack both memmove() and bcopy() - that is, hardly ever -
    was missing a "static" storage class specifier.

 4. When UTF-8 mode was not set, PCRE looped when compiling certain patterns
    containing an extended class (one that cannot be represented by a bitmap
    because it contains high-valued characters or Unicode property items, e.g.
    [\pZ]). Almost always one would set UTF-8 mode when processing such a
    pattern, but PCRE should not loop if you do not (it no longer does).
    [Detail: two cases were found: (a) a repeated subpattern containing an
    extended class; (b) a recursive reference to a subpattern that followed a
    previous extended class. It wasn't skipping over the extended class
    correctly when UTF-8 mode was not set.]

 5. A negated single-character class was not being recognized as fixed-length
    in lookbehind assertions such as (?<=[^f]), leading to an incorrect
    compile error "lookbehind assertion is not fixed length".

 6. The RunPerlTest auxiliary script was showing an unexpected difference
    between PCRE and Perl for UTF-8 tests. It turns out that it is hard to
    write a Perl script that can interpret lines of an input file either as
    byte characters or as UTF-8, which is what "perltest" was being required to
    do for the non-UTF-8 and UTF-8 tests, respectively. Essentially what you
    can't do is switch easily at run time between having the "use utf8;" pragma
    or not. In the end, I fudged it by using the RunPerlTest script to insert
    "use utf8;" explicitly for the UTF-8 tests.

 7. In multiline (/m) mode, PCRE was matching ^ after a terminating newline at
    the end of the subject string, contrary to the documentation and to what
    Perl does. This was true of both matching functions. Now it matches only at
    the start of the subject and immediately after *internal* newlines.

 8. A call of pcre_fullinfo() from pcretest to get the option bits was passing
    a pointer to an int instead of a pointer to an unsigned long int. This
    caused problems on 64-bit systems.

 9. Applied a patch from the folks at Google to pcrecpp.cc, to fix "another
    instance of the 'standard' template library not being so standard".

10. There was no check on the number of named subpatterns nor the maximum
    length of a subpattern name. The product of these values is used to compute
    the size of the memory block for a compiled pattern. By supplying a very
    long subpattern name and a large number of named subpatterns, the size
    computation could be caused to overflow. This is now prevented by limiting
    the length of names to 32 characters, and the number of named subpatterns
    to 10,000.

11. Subpatterns that are repeated with specific counts have to be replicated in
    the compiled pattern. The size of memory for this was computed from the
    length of the subpattern and the repeat count. The latter is limited to
    65535, but there was no limit on the former, meaning that integer overflow
    could in principle occur. The compiled length of a repeated subpattern is
    now limited to 30,000 bytes in order to prevent this.

12. Added the optional facility to have named substrings with the same name.

13. Added the ability to use a named substring as a condition, using the
    Python syntax: (?(name)yes|no). This overloads (?(R)... and names that
    are numbers (not recommended). Forward references are permitted.

14. Added forward references in named backreferences (if you see what I mean).

15. In UTF-8 mode, with the PCRE_DOTALL option set, a quantified dot in the
    pattern could run off the end of the subject. For example, the pattern
    "(?s)(.{1,5})"8 did this with the subject "ab".

16. If PCRE_DOTALL or PCRE_MULTILINE were set, pcre_dfa_exec() behaved as if
    PCRE_CASELESS was set when matching characters that were quantified with ?
    or *.

17. A character class other than a single negated character that had a minimum
    but no maximum quantifier - for example [ab]{6,} - was not handled
    correctly by pce_dfa_exec(). It would match only one character.

18. A valid (though odd) pattern that looked like a POSIX character
    class but used an invalid character after [ (for example [[,abc,]]) caused
    pcre_compile() to give the error "Failed: internal error: code overflow" or
    in some cases to crash with a glibc free() error. This could even happen if
    the pattern terminated after [[ but there just happened to be a sequence of
    letters, a binary zero, and a closing ] in the memory that followed.

19. Perl's treatment of octal escapes in the range \400 to \777 has changed
    over the years. Originally (before any Unicode support), just the bottom 8
    bits were taken. Thus, for example, \500 really meant \100. Nowadays the
    output from "man perlunicode" includes this:

      The regular expression compiler produces polymorphic opcodes.  That
      is, the pattern adapts to the data and automatically switches to
      the Unicode character scheme when presented with Unicode data--or
      instead uses a traditional byte scheme when presented with byte
      data.

    Sadly, a wide octal escape does not cause a switch, and in a string with
    no other multibyte characters, these octal escapes are treated as before.
    Thus, in Perl, the pattern  /\500/ actually matches \100 but the pattern
    /\500|\x{1ff}/ matches \500 or \777 because the whole thing is treated as a
    Unicode string.

    I have not perpetrated such confusion in PCRE. Up till now, it took just
    the bottom 8 bits, as in old Perl. I have now made octal escapes with
    values greater than \377 illegal in non-UTF-8 mode. In UTF-8 mode they
    translate to the appropriate multibyte character.

29. Applied some refactoring to reduce the number of warnings from Microsoft
    and Borland compilers. This has included removing the fudge introduced
    seven years ago for the OS/2 compiler (see 2.02/2 below) because it caused
    a warning about an unused variable.

21. PCRE has not included VT (character 0x0b) in the set of whitespace
    characters since release 4.0, because Perl (from release 5.004) does not.
    [Or at least, is documented not to: some releases seem to be in conflict
    with the documentation.] However, when a pattern was studied with
    pcre_study() and all its branches started with \s, PCRE still included VT
    as a possible starting character. Of course, this did no harm; it just
    caused an unnecessary match attempt.

22. Removed a now-redundant internal flag bit that recorded the fact that case
    dependency changed within the pattern. This was once needed for "required
    byte" processing, but is no longer used. This recovers a now-scarce options
    bit. Also moved the least significant internal flag bit to the most-
    significant bit of the word, which was not previously used (hangover from
    the days when it was an int rather than a uint) to free up another bit for
    the future.

23. Added support for CRLF line endings as well as CR and LF. As well as the
    default being selectable at build time, it can now be changed at runtime
    via the PCRE_NEWLINE_xxx flags. There are now options for pcregrep to
    specify that it is scanning data with non-default line endings.

24. Changed the definition of CXXLINK to make it agree with the definition of
    LINK in the Makefile, by replacing LDFLAGS to CXXFLAGS.

25. Applied Ian Taylor's patches to avoid using another stack frame for tail
    recursions. This makes a big different to stack usage for some patterns.

26. If a subpattern containing a named recursion or subroutine reference such
    as (?P>B) was quantified, for example (xxx(?P>B)){3}, the calculation of
    the space required for the compiled pattern went wrong and gave too small a
    value. Depending on the environment, this could lead to "Failed: internal
    error: code overflow at offset 49" or "glibc detected double free or
    corruption" errors.

27. Applied patches from Google (a) to support the new newline modes and (b) to
    advance over multibyte UTF-8 characters in GlobalReplace.

28. Change free() to pcre_free() in pcredemo.c. Apparently this makes a
    difference for some implementation of PCRE in some Windows version.

29. Added some extra testing facilities to pcretest:

    \q<number>   in a data line sets the "match limit" value
    \Q<number>   in a data line sets the "match recursion limt" value
    -S <number>  sets the stack size, where <number> is in megabytes

    The -S option isn't available for Windows.


Version 6.6 06-Feb-06
---------------------

 1. Change 16(a) for 6.5 broke things, because PCRE_DATA_SCOPE was not defined
    in pcreposix.h. I have copied the definition from pcre.h.

 2. Change 25 for 6.5 broke compilation in a build directory out-of-tree
    because pcre.h is no longer a built file.

 3. Added Jeff Friedl's additional debugging patches to pcregrep. These are
    not normally included in the compiled code.


Version 6.5 01-Feb-06
---------------------

 1. When using the partial match feature with pcre_dfa_exec(), it was not
    anchoring the second and subsequent partial matches at the new starting
    point. This could lead to incorrect results. For example, with the pattern
    /1234/, partially matching against "123" and then "a4" gave a match.

 2. Changes to pcregrep:

    (a) All non-match returns from pcre_exec() were being treated as failures
        to match the line. Now, unless the error is PCRE_ERROR_NOMATCH, an
        error message is output. Some extra information is given for the
        PCRE_ERROR_MATCHLIMIT and PCRE_ERROR_RECURSIONLIMIT errors, which are
        probably the only errors that are likely to be caused by users (by
        specifying a regex that has nested indefinite repeats, for instance).
        If there are more than 20 of these errors, pcregrep is abandoned.

    (b) A binary zero was treated as data while matching, but terminated the
        output line if it was written out. This has been fixed: binary zeroes
        are now no different to any other data bytes.

    (c) Whichever of the LC_ALL or LC_CTYPE environment variables is set is
        used to set a locale for matching. The --locale=xxxx long option has
        been added (no short equivalent) to specify a locale explicitly on the
        pcregrep command, overriding the environment variables.

    (d) When -B was used with -n, some line numbers in the output were one less
        than they should have been.

    (e) Added the -o (--only-matching) option.

    (f) If -A or -C was used with -c (count only), some lines of context were
        accidentally printed for the final match.

    (g) Added the -H (--with-filename) option.

    (h) The combination of options -rh failed to suppress file names for files
        that were found from directory arguments.

    (i) Added the -D (--devices) and -d (--directories) options.

    (j) Added the -F (--fixed-strings) option.

    (k) Allow "-" to be used as a file name for -f as well as for a data file.

    (l) Added the --colo(u)r option.

    (m) Added Jeffrey Friedl's -S testing option, but within #ifdefs so that it
        is not present by default.

 3. A nasty bug was discovered in the handling of recursive patterns, that is,
    items such as (?R) or (?1), when the recursion could match a number of
    alternatives. If it matched one of the alternatives, but subsequently,
    outside the recursion, there was a failure, the code tried to back up into
    the recursion. However, because of the way PCRE is implemented, this is not
    possible, and the result was an incorrect result from the match.

    In order to prevent this happening, the specification of recursion has
    been changed so that all such subpatterns are automatically treated as
    atomic groups. Thus, for example, (?R) is treated as if it were (?>(?R)).

 4. I had overlooked the fact that, in some locales, there are characters for
    which isalpha() is true but neither isupper() nor islower() are true. In
    the fr_FR locale, for instance, the \xAA and \xBA characters (ordmasculine
    and ordfeminine) are like this. This affected the treatment of \w and \W
    when they appeared in character classes, but not when they appeared outside
    a character class. The bit map for "word" characters is now created
    separately from the results of isalnum() instead of just taking it from the
    upper, lower, and digit maps. (Plus the underscore character, of course.)

 5. The above bug also affected the handling of POSIX character classes such as
    [[:alpha:]] and [[:alnum:]]. These do not have their own bit maps in PCRE's
    permanent tables. Instead, the bit maps for such a class were previously
    created as the appropriate unions of the upper, lower, and digit bitmaps.
    Now they are created by subtraction from the [[:word:]] class, which has
    its own bitmap.

 6. The [[:blank:]] character class matches horizontal, but not vertical space.
    It is created by subtracting the vertical space characters (\x09, \x0a,
    \x0b, \x0c) from the [[:space:]] bitmap. Previously, however, the
    subtraction was done in the overall bitmap for a character class, meaning
    that a class such as [\x0c[:blank:]] was incorrect because \x0c would not
    be recognized. This bug has been fixed.

 7. Patches from the folks at Google:

      (a) pcrecpp.cc: "to handle a corner case that may or may not happen in
      real life, but is still worth protecting against".

      (b) pcrecpp.cc: "corrects a bug when negative radixes are used with
      regular expressions".

      (c) pcre_scanner.cc: avoid use of std::count() because not all systems
      have it.

      (d) Split off pcrecpparg.h from pcrecpp.h and had the former built by
      "configure" and the latter not, in order to fix a problem somebody had
      with compiling the Arg class on HP-UX.

      (e) Improve the error-handling of the C++ wrapper a little bit.

      (f) New tests for checking recursion limiting.

 8. The pcre_memmove() function, which is used only if the environment does not
    have a standard memmove() function (and is therefore rarely compiled),
    contained two bugs: (a) use of int instead of size_t, and (b) it was not
    returning a result (though PCRE never actually uses the result).

 9. In the POSIX regexec() interface, if nmatch is specified as a ridiculously
    large number - greater than INT_MAX/(3*sizeof(int)) - REG_ESPACE is
    returned instead of calling malloc() with an overflowing number that would
    most likely cause subsequent chaos.

10. The debugging option of pcretest was not showing the NO_AUTO_CAPTURE flag.

11. The POSIX flag REG_NOSUB is now supported. When a pattern that was compiled
    with this option is matched, the nmatch and pmatch options of regexec() are
    ignored.

12. Added REG_UTF8 to the POSIX interface. This is not defined by POSIX, but is
    provided in case anyone wants to the the POSIX interface with UTF-8
    strings.

13. Added CXXLDFLAGS to the Makefile parameters to provide settings only on the
    C++ linking (needed for some HP-UX environments).

14. Avoid compiler warnings in get_ucpname() when compiled without UCP support
    (unused parameter) and in the pcre_printint() function (omitted "default"
    switch label when the default is to do nothing).

15. Added some code to make it possible, when PCRE is compiled as a C++
    library, to replace subject pointers for pcre_exec() with a smart pointer
    class, thus making it possible to process discontinuous strings.

16. The two macros PCRE_EXPORT and PCRE_DATA_SCOPE are confusing, and perform
    much the same function. They were added by different people who were trying
    to make PCRE easy to compile on non-Unix systems. It has been suggested
    that PCRE_EXPORT be abolished now that there is more automatic apparatus
    for compiling on Windows systems. I have therefore replaced it with
    PCRE_DATA_SCOPE. This is set automatically for Windows; if not set it
    defaults to "extern" for C or "extern C" for C++, which works fine on
    Unix-like systems. It is now possible to override the value of PCRE_DATA_
    SCOPE with something explicit in config.h. In addition:

    (a) pcreposix.h still had just "extern" instead of either of these macros;
        I have replaced it with PCRE_DATA_SCOPE.

    (b) Functions such as _pcre_xclass(), which are internal to the library,
        but external in the C sense, all had PCRE_EXPORT in their definitions.
        This is apparently wrong for the Windows case, so I have removed it.
        (It makes no difference on Unix-like systems.)

17. Added a new limit, MATCH_LIMIT_RECURSION, which limits the depth of nesting
    of recursive calls to match(). This is different to MATCH_LIMIT because
    that limits the total number of calls to match(), not all of which increase
    the depth of recursion. Limiting the recursion depth limits the amount of
    stack (or heap if NO_RECURSE is set) that is used. The default can be set
    when PCRE is compiled, and changed at run time. A patch from Google adds
    this functionality to the C++ interface.

18. Changes to the handling of Unicode character properties:

    (a) Updated the table to Unicode 4.1.0.

    (b) Recognize characters that are not in the table as "Cn" (undefined).

    (c) I revised the way the table is implemented to a much improved format
        which includes recognition of ranges. It now supports the ranges that
        are defined in UnicodeData.txt, and it also amalgamates other
        characters into ranges. This has reduced the number of entries in the
        table from around 16,000 to around 3,000, thus reducing its size
        considerably. I realized I did not need to use a tree structure after
        all - a binary chop search is just as efficient. Having reduced the
        number of entries, I extended their size from 6 bytes to 8 bytes to
        allow for more data.

    (d) Added support for Unicode script names via properties such as \p{Han}.

19. In UTF-8 mode, a backslash followed by a non-Ascii character was not
    matching that character.

20. When matching a repeated Unicode property with a minimum greater than zero,
    (for example \pL{2,}), PCRE could look past the end of the subject if it
    reached it while seeking the minimum number of characters. This could
    happen only if some of the characters were more than one byte long, because
    there is a check for at least the minimum number of bytes.

21. Refactored the implementation of \p and \P so as to be more general, to
    allow for more different types of property in future. This has changed the
    compiled form incompatibly. Anybody with saved compiled patterns that use
    \p or \P will have to recompile them.

22. Added "Any" and "L&" to the supported property types.

23. Recognize \x{...} as a code point specifier, even when not in UTF-8 mode,
    but give a compile time error if the value is greater than 0xff.

24. The man pages for pcrepartial, pcreprecompile, and pcre_compile2 were
    accidentally not being installed or uninstalled.

25. The pcre.h file was built from pcre.h.in, but the only changes that were
    made were to insert the current release number. This seemed silly, because
    it made things harder for people building PCRE on systems that don't run
    "configure". I have turned pcre.h into a distributed file, no longer built
    by "configure", with the version identification directly included. There is
    no longer a pcre.h.in file.

    However, this change necessitated a change to the pcre-config script as
    well. It is built from pcre-config.in, and one of the substitutions was the
    release number. I have updated configure.ac so that ./configure now finds
    the release number by grepping pcre.h.

26. Added the ability to run the tests under valgrind.


Version 6.4 05-Sep-05
---------------------

 1. Change 6.0/10/(l) to pcregrep introduced a bug that caused separator lines
    "--" to be printed when multiple files were scanned, even when none of the
    -A, -B, or -C options were used. This is not compatible with Gnu grep, so I
    consider it to be a bug, and have restored the previous behaviour.

 2. A couple of code tidies to get rid of compiler warnings.

 3. The pcretest program used to cheat by referring to symbols in the library
    whose names begin with _pcre_. These are internal symbols that are not
    really supposed to be visible externally, and in some environments it is
    possible to suppress them. The cheating is now confined to including
    certain files from the library's source, which is a bit cleaner.

 4. Renamed pcre.in as pcre.h.in to go with pcrecpp.h.in; it also makes the
    file's purpose clearer.

 5. Reorganized pcre_ucp_findchar().


Version 6.3 15-Aug-05
---------------------

 1. The file libpcre.pc.in did not have general read permission in the tarball.

 2. There were some problems when building without C++ support:

    (a) If C++ support was not built, "make install" and "make test" still
        tried to test it.

    (b) There were problems when the value of CXX was explicitly set. Some
        changes have been made to try to fix these, and ...

    (c) --disable-cpp can now be used to explicitly disable C++ support.

    (d) The use of @CPP_OBJ@ directly caused a blank line preceded by a
        backslash in a target when C++ was disabled. This confuses some
        versions of "make", apparently. Using an intermediate variable solves
        this. (Same for CPP_LOBJ.)

 3. $(LINK_FOR_BUILD) now includes $(CFLAGS_FOR_BUILD) and $(LINK)
    (non-Windows) now includes $(CFLAGS) because these flags are sometimes
    necessary on certain architectures.

 4. Added a setting of -export-symbols-regex to the link command to remove
    those symbols that are exported in the C sense, but actually are local
    within the library, and not documented. Their names all begin with
    "_pcre_". This is not a perfect job, because (a) we have to except some
    symbols that pcretest ("illegally") uses, and (b) the facility isn't always
    available (and never for static libraries). I have made a note to try to
    find a way round (a) in the future.


Version 6.2 01-Aug-05
---------------------

 1. There was no test for integer overflow of quantifier values. A construction
    such as {1111111111111111} would give undefined results. What is worse, if
    a minimum quantifier for a parenthesized subpattern overflowed and became
    negative, the calculation of the memory size went wrong. This could have
    led to memory overwriting.

 2. Building PCRE using VPATH was broken. Hopefully it is now fixed.

 3. Added "b" to the 2nd argument of fopen() in dftables.c, for non-Unix-like
    operating environments where this matters.

 4. Applied Giuseppe Maxia's patch to add additional features for controlling
    PCRE options from within the C++ wrapper.

 5. Named capturing subpatterns were not being correctly counted when a pattern
    was compiled. This caused two problems: (a) If there were more than 100
    such subpatterns, the calculation of the memory needed for the whole
    compiled pattern went wrong, leading to an overflow error. (b) Numerical
    back references of the form \12, where the number was greater than 9, were
    not recognized as back references, even though there were sufficient
    previous subpatterns.

 6. Two minor patches to pcrecpp.cc in order to allow it to compile on older
    versions of gcc, e.g. 2.95.4.


Version 6.1 21-Jun-05
---------------------

 1. There was one reference to the variable "posix" in pcretest.c that was not
    surrounded by "#if !defined NOPOSIX".

 2. Make it possible to compile pcretest without DFA support, UTF8 support, or
    the cross-check on the old pcre_info() function, for the benefit of the
    cut-down version of PCRE that is currently imported into Exim.

 3. A (silly) pattern starting with (?i)(?-i) caused an internal space
    allocation error. I've done the easy fix, which wastes 2 bytes for sensible
    patterns that start (?i) but I don't think that matters. The use of (?i) is
    just an example; this all applies to the other options as well.

 4. Since libtool seems to echo the compile commands it is issuing, the output
    from "make" can be reduced a bit by putting "@" in front of each libtool
    compile command.

 5. Patch from the folks at Google for configure.in to be a bit more thorough
    in checking for a suitable C++ installation before trying to compile the
    C++ stuff. This should fix a reported problem when a compiler was present,
    but no suitable headers.

 6. The man pages all had just "PCRE" as their title. I have changed them to
    be the relevant file name. I have also arranged that these names are
    retained in the file doc/pcre.txt, which is a concatenation in text format
    of all the man pages except the little individual ones for each function.

 7. The NON-UNIX-USE file had not been updated for the different set of source
    files that come with release 6. I also added a few comments about the C++
    wrapper.


Version 6.0 07-Jun-05
---------------------

 1. Some minor internal re-organization to help with my DFA experiments.

 2. Some missing #ifdef SUPPORT_UCP conditionals in pcretest and printint that
    didn't matter for the library itself when fully configured, but did matter
    when compiling without UCP support, or within Exim, where the ucp files are
    not imported.

 3. Refactoring of the library code to split up the various functions into
    different source modules. The addition of the new DFA matching code (see
    below) to a single monolithic source would have made it really too
    unwieldy, quite apart from causing all the code to be include in a
    statically linked application, when only some functions are used. This is
    relevant even without the DFA addition now that patterns can be compiled in
    one application and matched in another.

    The downside of splitting up is that there have to be some external
    functions and data tables that are used internally in different modules of
    the library but which are not part of the API. These have all had their
    names changed to start with "_pcre_" so that they are unlikely to clash
    with other external names.

 4. Added an alternate matching function, pcre_dfa_exec(), which matches using
    a different (DFA) algorithm. Although it is slower than the original
    function, it does have some advantages for certain types of matching
    problem.

 5. Upgrades to pcretest in order to test the features of pcre_dfa_exec(),
    including restarting after a partial match.

 6. A patch for pcregrep that defines INVALID_FILE_ATTRIBUTES if it is not
    defined when compiling for Windows was sent to me. I have put it into the
    code, though I have no means of testing or verifying it.

 7. Added the pcre_refcount() auxiliary function.

 8. Added the PCRE_FIRSTLINE option. This constrains an unanchored pattern to
    match before or at the first newline in the subject string. In pcretest,
    the /f option on a pattern can be used to set this.

 9. A repeated \w when used in UTF-8 mode with characters greater than 256
    would behave wrongly. This has been present in PCRE since release 4.0.

10. A number of changes to the pcregrep command:

    (a) Refactored how -x works; insert ^(...)$ instead of setting
        PCRE_ANCHORED and checking the length, in preparation for adding
        something similar for -w.

    (b) Added the -w (match as a word) option.

    (c) Refactored the way lines are read and buffered so as to have more
        than one at a time available.

    (d) Implemented a pcregrep test script.

    (e) Added the -M (multiline match) option. This allows patterns to match
        over several lines of the subject. The buffering ensures that at least
        8K, or the rest of the document (whichever is the shorter) is available
        for matching (and similarly the previous 8K for lookbehind assertions).

    (f) Changed the --help output so that it now says

          -w, --word-regex(p)

        instead of two lines, one with "regex" and the other with "regexp"
        because that confused at least one person since the short forms are the
        same. (This required a bit of code, as the output is generated
        automatically from a table. It wasn't just a text change.)

    (g) -- can be used to terminate pcregrep options if the next thing isn't an
        option but starts with a hyphen. Could be a pattern or a path name
        starting with a hyphen, for instance.

    (h) "-" can be given as a file name to represent stdin.

    (i) When file names are being printed, "(standard input)" is used for
        the standard input, for compatibility with GNU grep. Previously
        "<stdin>" was used.

    (j) The option --label=xxx can be used to supply a name to be used for
        stdin when file names are being printed. There is no short form.

    (k) Re-factored the options decoding logic because we are going to add
        two more options that take data. Such options can now be given in four
        different ways, e.g. "-fname", "-f name", "--file=name", "--file name".

    (l) Added the -A, -B, and -C options for requesting that lines of context
        around matches be printed.

    (m) Added the -L option to print the names of files that do not contain
        any matching lines, that is, the complement of -l.

    (n) The return code is 2 if any file cannot be opened, but pcregrep does
        continue to scan other files.

    (o) The -s option was incorrectly implemented. For compatibility with other
        greps, it now suppresses the error message for a non-existent or non-
        accessible file (but not the return code). There is a new option called
        -q that suppresses the output of matching lines, which was what -s was
        previously doing.

    (p) Added --include and --exclude options to specify files for inclusion
        and exclusion when recursing.

11. The Makefile was not using the Autoconf-supported LDFLAGS macro properly.
    Hopefully, it now does.

12. Missing cast in pcre_study().

13. Added an "uninstall" target to the makefile.

14. Replaced "extern" in the function prototypes in Makefile.in with
    "PCRE_DATA_SCOPE", which defaults to 'extern' or 'extern "C"' in the Unix
    world, but is set differently for Windows.

15. Added a second compiling function called pcre_compile2(). The only
    difference is that it has an extra argument, which is a pointer to an
    integer error code. When there is a compile-time failure, this is set
    non-zero, in addition to the error test pointer being set to point to an
    error message. The new argument may be NULL if no error number is required
    (but then you may as well call pcre_compile(), which is now just a
    wrapper). This facility is provided because some applications need a
    numeric error indication, but it has also enabled me to tidy up the way
    compile-time errors are handled in the POSIX wrapper.

16. Added VPATH=.libs to the makefile; this should help when building with one
    prefix path and installing with another. (Or so I'm told by someone who
    knows more about this stuff than I do.)

17. Added a new option, REG_DOTALL, to the POSIX function regcomp(). This
    passes PCRE_DOTALL to the pcre_compile() function, making the "." character
    match everything, including newlines. This is not POSIX-compatible, but
    somebody wanted the feature. From pcretest it can be activated by using
    both the P and the s flags.

18. AC_PROG_LIBTOOL appeared twice in Makefile.in. Removed one.

19. libpcre.pc was being incorrectly installed as executable.

20. A couple of places in pcretest check for end-of-line by looking for '\n';
    it now also looks for '\r' so that it will work unmodified on Windows.

21. Added Google's contributed C++ wrapper to the distribution.

22. Added some untidy missing memory free() calls in pcretest, to keep
    Electric Fence happy when testing.



Version 5.0 13-Sep-04
---------------------

 1. Internal change: literal characters are no longer packed up into items
    containing multiple characters in a single byte-string. Each character
    is now matched using a separate opcode. However, there may be more than one
    byte in the character in UTF-8 mode.

 2. The pcre_callout_block structure has two new fields: pattern_position and
    next_item_length. These contain the offset in the pattern to the next match
    item, and its length, respectively.

 3. The PCRE_AUTO_CALLOUT option for pcre_compile() requests the automatic
    insertion of callouts before each pattern item. Added the /C option to
    pcretest to make use of this.

 4. On the advice of a Windows user, the lines

      #if defined(_WIN32) || defined(WIN32)
      _setmode( _fileno( stdout ), 0x8000 );
      #endif  /* defined(_WIN32) || defined(WIN32) */

    have been added to the source of pcretest. This apparently does useful
    magic in relation to line terminators.

 5. Changed "r" and "w" in the calls to fopen() in pcretest to "rb" and "wb"
    for the benefit of those environments where the "b" makes a difference.

 6. The icc compiler has the same options as gcc, but "configure" doesn't seem
    to know about it. I have put a hack into configure.in that adds in code
    to set GCC=yes if CC=icc. This seems to end up at a point in the
    generated configure script that is early enough to affect the setting of
    compiler options, which is what is needed, but I have no means of testing
    whether it really works. (The user who reported this had patched the
    generated configure script, which of course I cannot do.)

    LATER: After change 22 below (new libtool files), the configure script
    seems to know about icc (and also ecc). Therefore, I have commented out
    this hack in configure.in.

 7. Added support for pkg-config (2 patches were sent in).

 8. Negated POSIX character classes that used a combination of internal tables
    were completely broken. These were [[:^alpha:]], [[:^alnum:]], and
    [[:^ascii]]. Typically, they would match almost any characters. The other
    POSIX classes were not broken in this way.

 9. Matching the pattern "\b.*?" against "ab cd", starting at offset 1, failed
    to find the match, as PCRE was deluded into thinking that the match had to
    start at the start point or following a newline. The same bug applied to
    patterns with negative forward assertions or any backward assertions
    preceding ".*" at the start, unless the pattern required a fixed first
    character. This was a failing pattern: "(?!.bcd).*". The bug is now fixed.

10. In UTF-8 mode, when moving forwards in the subject after a failed match
    starting at the last subject character, bytes beyond the end of the subject
    string were read.

11. Renamed the variable "class" as "classbits" to make life easier for C++
    users. (Previously there was a macro definition, but it apparently wasn't
    enough.)

12. Added the new field "tables" to the extra data so that tables can be passed
    in at exec time, or the internal tables can be re-selected. This allows
    a compiled regex to be saved and re-used at a later time by a different
    program that might have everything at different addresses.

13. Modified the pcre-config script so that, when run on Solaris, it shows a
    -R library as well as a -L library.

14. The debugging options of pcretest (-d on the command line or D on a
    pattern) showed incorrect output for anything following an extended class
    that contained multibyte characters and which was followed by a quantifier.

15. Added optional support for general category Unicode character properties
    via the \p, \P, and \X escapes. Unicode property support implies UTF-8
    support. It adds about 90K to the size of the library. The meanings of the
    inbuilt class escapes such as \d and \s have NOT been changed.

16. Updated pcredemo.c to include calls to free() to release the memory for the
    compiled pattern.

17. The generated file chartables.c was being created in the source directory
    instead of in the building directory. This caused the build to fail if the
    source directory was different from the building directory, and was
    read-only.

18. Added some sample Win commands from Mark Tetrode into the NON-UNIX-USE
    file. No doubt somebody will tell me if they don't make sense... Also added
    Dan Mooney's comments about building on OpenVMS.

19. Added support for partial matching via the PCRE_PARTIAL option for
    pcre_exec() and the \P data escape in pcretest.

20. Extended pcretest with 3 new pattern features:

    (i)   A pattern option of the form ">rest-of-line" causes pcretest to
          write the compiled pattern to the file whose name is "rest-of-line".
          This is a straight binary dump of the data, with the saved pointer to
          the character tables forced to be NULL. The study data, if any, is
          written too. After writing, pcretest reads a new pattern.

    (ii)  If, instead of a pattern, "<rest-of-line" is given, pcretest reads a
          compiled pattern from the given file. There must not be any
          occurrences of "<" in the file name (pretty unlikely); if there are,
          pcretest will instead treat the initial "<" as a pattern delimiter.
          After reading in the pattern, pcretest goes on to read data lines as
          usual.

    (iii) The F pattern option causes pcretest to flip the bytes in the 32-bit
          and 16-bit fields in a compiled pattern, to simulate a pattern that
          was compiled on a host of opposite endianness.

21. The pcre-exec() function can now cope with patterns that were compiled on
    hosts of opposite endianness, with this restriction:

      As for any compiled expression that is saved and used later, the tables
      pointer field cannot be preserved; the extra_data field in the arguments
      to pcre_exec() should be used to pass in a tables address if a value
      other than the default internal tables were used at compile time.

22. Calling pcre_exec() with a negative value of the "ovecsize" parameter is
    now diagnosed as an error. Previously, most of the time, a negative number
    would have been treated as zero, but if in addition "ovector" was passed as
    NULL, a crash could occur.

23. Updated the files ltmain.sh, config.sub, config.guess, and aclocal.m4 with
    new versions from the libtool 1.5 distribution (the last one is a copy of
    a file called libtool.m4). This seems to have fixed the need to patch
    "configure" to support Darwin 1.3 (which I used to do). However, I still
    had to patch ltmain.sh to ensure that ${SED} is set (it isn't on my
    workstation).

24. Changed the PCRE licence to be the more standard "BSD" licence.


Version 4.5 01-Dec-03
---------------------

 1. There has been some re-arrangement of the code for the match() function so
    that it can be compiled in a version that does not call itself recursively.
    Instead, it keeps those local variables that need separate instances for
    each "recursion" in a frame on the heap, and gets/frees frames whenever it
    needs to "recurse". Keeping track of where control must go is done by means
    of setjmp/longjmp. The whole thing is implemented by a set of macros that
    hide most of the details from the main code, and operates only if
    NO_RECURSE is defined while compiling pcre.c. If PCRE is built using the
    "configure" mechanism, "--disable-stack-for-recursion" turns on this way of
    operating.

    To make it easier for callers to provide specially tailored get/free
    functions for this usage, two new functions, pcre_stack_malloc, and
    pcre_stack_free, are used. They are always called in strict stacking order,
    and the size of block requested is always the same.

    The PCRE_CONFIG_STACKRECURSE info parameter can be used to find out whether
    PCRE has been compiled to use the stack or the heap for recursion. The
    -C option of pcretest uses this to show which version is compiled.

    A new data escape \S, is added to pcretest; it causes the amounts of store
    obtained and freed by both kinds of malloc/free at match time to be added
    to the output.

 2. Changed the locale test to use "fr_FR" instead of "fr" because that's
    what's available on my current Linux desktop machine.

 3. When matching a UTF-8 string, the test for a valid string at the start has
    been extended. If start_offset is not zero, PCRE now checks that it points
    to a byte that is the start of a UTF-8 character. If not, it returns
    PCRE_ERROR_BADUTF8_OFFSET (-11). Note: the whole string is still checked;
    this is necessary because there may be backward assertions in the pattern.
    When matching the same subject several times, it may save resources to use
    PCRE_NO_UTF8_CHECK on all but the first call if the string is long.

 4. The code for checking the validity of UTF-8 strings has been tightened so
    that it rejects (a) strings containing 0xfe or 0xff bytes and (b) strings
    containing "overlong sequences".

 5. Fixed a bug (appearing twice) that I could not find any way of exploiting!
    I had written "if ((digitab[*p++] && chtab_digit) == 0)" where the "&&"
    should have been "&", but it just so happened that all the cases this let
    through by mistake were picked up later in the function.

 6. I had used a variable called "isblank" - this is a C99 function, causing
    some compilers to warn. To avoid this, I renamed it (as "blankclass").

 7. Cosmetic: (a) only output another newline at the end of pcretest if it is
    prompting; (b) run "./pcretest /dev/null" at the start of the test script
    so the version is shown; (c) stop "make test" echoing "./RunTest".

 8. Added patches from David Burgess to enable PCRE to run on EBCDIC systems.

 9. The prototype for memmove() for systems that don't have it was using
    size_t, but the inclusion of the header that defines size_t was later. I've
    moved the #includes for the C headers earlier to avoid this.

10. Added some adjustments to the code to make it easier to compiler on certain
    special systems:

      (a) Some "const" qualifiers were missing.
      (b) Added the macro EXPORT before all exported functions; by default this
          is defined to be empty.
      (c) Changed the dftables auxiliary program (that builds chartables.c) so
          that it reads its output file name as an argument instead of writing
          to the standard output and assuming this can be redirected.

11. In UTF-8 mode, if a recursive reference (e.g. (?1)) followed a character
    class containing characters with values greater than 255, PCRE compilation
    went into a loop.

12. A recursive reference to a subpattern that was within another subpattern
    that had a minimum quantifier of zero caused PCRE to crash. For example,
    (x(y(?2))z)? provoked this bug with a subject that got as far as the
    recursion. If the recursively-called subpattern itself had a zero repeat,
    that was OK.

13. In pcretest, the buffer for reading a data line was set at 30K, but the
    buffer into which it was copied (for escape processing) was still set at
    1024, so long lines caused crashes.

14. A pattern such as /[ab]{1,3}+/ failed to compile, giving the error
    "internal error: code overflow...". This applied to any character class
    that was followed by a possessive quantifier.

15. Modified the Makefile to add libpcre.la as a prerequisite for
    libpcreposix.la because I was told this is needed for a parallel build to
    work.

16. If a pattern that contained .* following optional items at the start was
    studied, the wrong optimizing data was generated, leading to matching
    errors. For example, studying /[ab]*.*c/ concluded, erroneously, that any
    matching string must start with a or b or c. The correct conclusion for
    this pattern is that a match can start with any character.


Version 4.4 13-Aug-03
---------------------

 1. In UTF-8 mode, a character class containing characters with values between
    127 and 255 was not handled correctly if the compiled pattern was studied.
    In fixing this, I have also improved the studying algorithm for such
    classes (slightly).

 2. Three internal functions had redundant arguments passed to them. Removal
    might give a very teeny performance improvement.

 3. Documentation bug: the value of the capture_top field in a callout is *one
    more than* the number of the hightest numbered captured substring.

 4. The Makefile linked pcretest and pcregrep with -lpcre, which could result
    in incorrectly linking with a previously installed version. They now link
    explicitly with libpcre.la.

 5. configure.in no longer needs to recognize Cygwin specially.

 6. A problem in pcre.in for Windows platforms is fixed.

 7. If a pattern was successfully studied, and the -d (or /D) flag was given to
    pcretest, it used to include the size of the study block as part of its
    output. Unfortunately, the structure contains a field that has a different
    size on different hardware architectures. This meant that the tests that
    showed this size failed. As the block is currently always of a fixed size,
    this information isn't actually particularly useful in pcretest output, so
    I have just removed it.

 8. Three pre-processor statements accidentally did not start in column 1.
    Sadly, there are *still* compilers around that complain, even though
    standard C has not required this for well over a decade. Sigh.

 9. In pcretest, the code for checking callouts passed small integers in the
    callout_data field, which is a void * field. However, some picky compilers
    complained about the casts involved for this on 64-bit systems. Now
    pcretest passes the address of the small integer instead, which should get
    rid of the warnings.

10. By default, when in UTF-8 mode, PCRE now checks for valid UTF-8 strings at
    both compile and run time, and gives an error if an invalid UTF-8 sequence
    is found. There is a option for disabling this check in cases where the
    string is known to be correct and/or the maximum performance is wanted.

11. In response to a bug report, I changed one line in Makefile.in from

        -Wl,--out-implib,.libs/lib@WIN_PREFIX@pcreposix.dll.a \
    to
        -Wl,--out-implib,.libs/@WIN_PREFIX@libpcreposix.dll.a \

    to look similar to other lines, but I have no way of telling whether this
    is the right thing to do, as I do not use Windows. No doubt I'll get told
    if it's wrong...


Version 4.3 21-May-03
---------------------

1. Two instances of @WIN_PREFIX@ omitted from the Windows targets in the
   Makefile.

2. Some refactoring to improve the quality of the code:

   (i)   The utf8_table... variables are now declared "const".

   (ii)  The code for \cx, which used the "case flipping" table to upper case
         lower case letters, now just substracts 32. This is ASCII-specific,
         but the whole concept of \cx is ASCII-specific, so it seems
         reasonable.

   (iii) PCRE was using its character types table to recognize decimal and
         hexadecimal digits in the pattern. This is silly, because it handles
         only 0-9, a-f, and A-F, but the character types table is locale-
         specific, which means strange things might happen. A private
         table is now used for this - though it costs 256 bytes, a table is
         much faster than multiple explicit tests. Of course, the standard
         character types table is still used for matching digits in subject
         strings against \d.

   (iv)  Strictly, the identifier ESC_t is reserved by POSIX (all identifiers
         ending in _t are). So I've renamed it as ESC_tee.

3. The first argument for regexec() in the POSIX wrapper should have been
   defined as "const".

4. Changed pcretest to use malloc() for its buffers so that they can be
   Electric Fenced for debugging.

5. There were several places in the code where, in UTF-8 mode, PCRE would try
   to read one or more bytes before the start of the subject string. Often this
   had no effect on PCRE's behaviour, but in some circumstances it could
   provoke a segmentation fault.

6. A lookbehind at the start of a pattern in UTF-8 mode could also cause PCRE
   to try to read one or more bytes before the start of the subject string.

7. A lookbehind in a pattern matched in non-UTF-8 mode on a PCRE compiled with
   UTF-8 support could misbehave in various ways if the subject string
   contained bytes with the 0x80 bit set and the 0x40 bit unset in a lookbehind
   area. (PCRE was not checking for the UTF-8 mode flag, and trying to move
   back over UTF-8 characters.)


Version 4.2 14-Apr-03
---------------------

1. Typo "#if SUPPORT_UTF8" instead of "#ifdef SUPPORT_UTF8" fixed.

2. Changes to the building process, supplied by Ronald Landheer-Cieslak
     [ON_WINDOWS]: new variable, "#" on non-Windows platforms
     [NOT_ON_WINDOWS]: new variable, "#" on Windows platforms
     [WIN_PREFIX]: new variable, "cyg" for Cygwin
     * Makefile.in: use autoconf substitution for OBJEXT, EXEEXT, BUILD_OBJEXT
       and BUILD_EXEEXT
     Note: automatic setting of the BUILD variables is not yet working
     set CPPFLAGS and BUILD_CPPFLAGS (but don't use yet) - should be used at
       compile-time but not at link-time
     [LINK]: use for linking executables only
     make different versions for Windows and non-Windows
     [LINKLIB]: new variable, copy of UNIX-style LINK, used for linking
       libraries
     [LINK_FOR_BUILD]: new variable
     [OBJEXT]: use throughout
     [EXEEXT]: use throughout
     <winshared>: new target
     <wininstall>: new target
     <dftables.o>: use native compiler
     <dftables>: use native linker
     <install>: handle Windows platform correctly
     <clean>: ditto
     <check>: ditto
     copy DLL to top builddir before testing

   As part of these changes, -no-undefined was removed again. This was reported
   to give trouble on HP-UX 11.0, so getting rid of it seems like a good idea
   in any case.

3. Some tidies to get rid of compiler warnings:

   . In the match_data structure, match_limit was an unsigned long int, whereas
     match_call_count was an int. I've made them both unsigned long ints.

   . In pcretest the fact that a const uschar * doesn't automatically cast to
     a void * provoked a warning.

   . Turning on some more compiler warnings threw up some "shadow" variables
     and a few more missing casts.

4. If PCRE was complied with UTF-8 support, but called without the PCRE_UTF8
   option, a class that contained a single character with a value between 128
   and 255 (e.g. /[\xFF]/) caused PCRE to crash.

5. If PCRE was compiled with UTF-8 support, but called without the PCRE_UTF8
   option, a class that contained several characters, but with at least one
   whose value was between 128 and 255 caused PCRE to crash.


Version 4.1 12-Mar-03
---------------------

1. Compiling with gcc -pedantic found a couple of places where casts were
needed, and a string in dftables.c that was longer than standard compilers are
required to support.

2. Compiling with Sun's compiler found a few more places where the code could
be tidied up in order to avoid warnings.

3. The variables for cross-compiling were called HOST_CC and HOST_CFLAGS; the
first of these names is deprecated in the latest Autoconf in favour of the name
CC_FOR_BUILD, because "host" is typically used to mean the system on which the
compiled code will be run. I can't find a reference for HOST_CFLAGS, but by
analogy I have changed it to CFLAGS_FOR_BUILD.

4. Added -no-undefined to the linking command in the Makefile, because this is
apparently helpful for Windows. To make it work, also added "-L. -lpcre" to the
linking step for the pcreposix library.

5. PCRE was failing to diagnose the case of two named groups with the same
name.

6. A problem with one of PCRE's optimizations was discovered. PCRE remembers a
literal character that is needed in the subject for a match, and scans along to
ensure that it is present before embarking on the full matching process. This
saves time in cases of nested unlimited repeats that are never going to match.
Problem: the scan can take a lot of time if the subject is very long (e.g.
megabytes), thus penalizing straightforward matches. It is now done only if the
amount of subject to be scanned is less than 1000 bytes.

7. A lesser problem with the same optimization is that it was recording the
first character of an anchored pattern as "needed", thus provoking a search
right along the subject, even when the first match of the pattern was going to
fail. The "needed" character is now not set for anchored patterns, unless it
follows something in the pattern that is of non-fixed length. Thus, it still
fulfils its original purpose of finding quick non-matches in cases of nested
unlimited repeats, but isn't used for simple anchored patterns such as /^abc/.


Version 4.0 17-Feb-03
---------------------

1. If a comment in an extended regex that started immediately after a meta-item
extended to the end of string, PCRE compiled incorrect data. This could lead to
all kinds of weird effects. Example: /#/ was bad; /()#/ was bad; /a#/ was not.

2. Moved to autoconf 2.53 and libtool 1.4.2.

3. Perl 5.8 no longer needs "use utf8" for doing UTF-8 things. Consequently,
the special perltest8 script is no longer needed - all the tests can be run
from a single perltest script.

4. From 5.004, Perl has not included the VT character (0x0b) in the set defined
by \s. It has now been removed in PCRE. This means it isn't recognized as
whitespace in /x regexes too, which is the same as Perl. Note that the POSIX
class [:space:] *does* include VT, thereby creating a mess.

5. Added the class [:blank:] (a GNU extension from Perl 5.8) to match only
space and tab.

6. Perl 5.005 was a long time ago. It's time to amalgamate the tests that use
its new features into the main test script, reducing the number of scripts.

7. Perl 5.8 has changed the meaning of patterns like /a(?i)b/. Earlier versions
were backward compatible, and made the (?i) apply to the whole pattern, as if
/i were given. Now it behaves more logically, and applies the option setting
only to what follows. PCRE has been changed to follow suit. However, if it
finds options settings right at the start of the pattern, it extracts them into
the global options, as before. Thus, they show up in the info data.

8. Added support for the \Q...\E escape sequence. Characters in between are
treated as literals. This is slightly different from Perl in that $ and @ are
also handled as literals inside the quotes. In Perl, they will cause variable
interpolation. Note the following examples:

    Pattern            PCRE matches      Perl matches

    \Qabc$xyz\E        abc$xyz           abc followed by the contents of $xyz
    \Qabc\$xyz\E       abc\$xyz          abc\$xyz
    \Qabc\E\$\Qxyz\E   abc$xyz           abc$xyz

For compatibility with Perl, \Q...\E sequences are recognized inside character
classes as well as outside them.

9. Re-organized 3 code statements in pcretest to avoid "overflow in
floating-point constant arithmetic" warnings from a Microsoft compiler. Added a
(size_t) cast to one statement in pcretest and one in pcreposix to avoid
signed/unsigned warnings.

10. SunOS4 doesn't have strtoul(). This was used only for unpicking the -o
option for pcretest, so I've replaced it by a simple function that does just
that job.

11. pcregrep was ending with code 0 instead of 2 for the commands "pcregrep" or
"pcregrep -".

12. Added "possessive quantifiers" ?+, *+, ++, and {,}+ which come from Sun's
Java package. This provides some syntactic sugar for simple cases of what my
documentation calls "once-only subpatterns". A pattern such as x*+ is the same
as (?>x*). In other words, if what is inside (?>...) is just a single repeated
item, you can use this simplified notation. Note that only makes sense with
greedy quantifiers. Consequently, the use of the possessive quantifier forces
greediness, whatever the setting of the PCRE_UNGREEDY option.

13. A change of greediness default within a pattern was not taking effect at
the current level for patterns like /(b+(?U)a+)/. It did apply to parenthesized
subpatterns that followed. Patterns like /b+(?U)a+/ worked because the option
was abstracted outside.

14. PCRE now supports the \G assertion. It is true when the current matching
position is at the start point of the match. This differs from \A when the
starting offset is non-zero. Used with the /g option of pcretest (or similar
code), it works in the same way as it does for Perl's /g option. If all
alternatives of a regex begin with \G, the expression is anchored to the start
match position, and the "anchored" flag is set in the compiled expression.

15. Some bugs concerning the handling of certain option changes within patterns
have been fixed. These applied to options other than (?ims). For example,
"a(?x: b c )d" did not match "XabcdY" but did match "Xa b c dY". It should have
been the other way round. Some of this was related to change 7 above.

16. PCRE now gives errors for /[.x.]/ and /[=x=]/ as unsupported POSIX
features, as Perl does. Previously, PCRE gave the warnings only for /[[.x.]]/
and /[[=x=]]/. PCRE now also gives an error for /[:name:]/ because it supports
POSIX classes only within a class (e.g. /[[:alpha:]]/).

17. Added support for Perl's \C escape. This matches one byte, even in UTF8
mode. Unlike ".", it always matches newline, whatever the setting of
PCRE_DOTALL. However, PCRE does not permit \C to appear in lookbehind
assertions. Perl allows it, but it doesn't (in general) work because it can't
calculate the length of the lookbehind. At least, that's the case for Perl
5.8.0 - I've been told they are going to document that it doesn't work in
future.

18. Added an error diagnosis for escapes that PCRE does not support: these are
\L, \l, \N, \P, \p, \U, \u, and \X.

19. Although correctly diagnosing a missing ']' in a character class, PCRE was
reading past the end of the pattern in cases such as /[abcd/.

20. PCRE was getting more memory than necessary for patterns with classes that
contained both POSIX named classes and other characters, e.g. /[[:space:]abc/.

21. Added some code, conditional on #ifdef VPCOMPAT, to make life easier for
compiling PCRE for use with Virtual Pascal.

22. Small fix to the Makefile to make it work properly if the build is done
outside the source tree.

23. Added a new extension: a condition to go with recursion. If a conditional
subpattern starts with (?(R) the "true" branch is used if recursion has
happened, whereas the "false" branch is used only at the top level.

24. When there was a very long string of literal characters (over 255 bytes
without UTF support, over 250 bytes with UTF support), the computation of how
much memory was required could be incorrect, leading to segfaults or other
strange effects.

25. PCRE was incorrectly assuming anchoring (either to start of subject or to
start of line for a non-DOTALL pattern) when a pattern started with (.*) and
there was a subsequent back reference to those brackets. This meant that, for
example, /(.*)\d+\1/ failed to match "abc123bc". Unfortunately, it isn't
possible to check for precisely this case. All we can do is abandon the
optimization if .* occurs inside capturing brackets when there are any back
references whatsoever. (See below for a better fix that came later.)

26. The handling of the optimization for finding the first character of a
non-anchored pattern, and for finding a character that is required later in the
match were failing in some cases. This didn't break the matching; it just
failed to optimize when it could. The way this is done has been re-implemented.

27. Fixed typo in error message for invalid (?R item (it said "(?p").

28. Added a new feature that provides some of the functionality that Perl
provides with (?{...}). The facility is termed a "callout". The way it is done
in PCRE is for the caller to provide an optional function, by setting
pcre_callout to its entry point. Like pcre_malloc and pcre_free, this is a
global variable. By default it is unset, which disables all calling out. To get
the function called, the regex must include (?C) at appropriate points. This
is, in fact, equivalent to (?C0), and any number <= 255 may be given with (?C).
This provides a means of identifying different callout points. When PCRE
reaches such a point in the regex, if pcre_callout has been set, the external
function is called. It is provided with data in a structure called
pcre_callout_block, which is defined in pcre.h. If the function returns 0,
matching continues; if it returns a non-zero value, the match at the current
point fails. However, backtracking will occur if possible. [This was changed
later and other features added - see item 49 below.]

29. pcretest is upgraded to test the callout functionality. It provides a
callout function that displays information. By default, it shows the start of
the match and the current position in the text. There are some new data escapes
to vary what happens:

    \C+         in addition, show current contents of captured substrings
    \C-         do not supply a callout function
    \C!n        return 1 when callout number n is reached
    \C!n!m      return 1 when callout number n is reached for the mth time

30. If pcregrep was called with the -l option and just a single file name, it
output "<stdin>" if a match was found, instead of the file name.

31. Improve the efficiency of the POSIX API to PCRE. If the number of capturing
slots is less than POSIX_MALLOC_THRESHOLD, use a block on the stack to pass to
pcre_exec(). This saves a malloc/free per call. The default value of
POSIX_MALLOC_THRESHOLD is 10; it can be changed by --with-posix-malloc-threshold
when configuring.

32. The default maximum size of a compiled pattern is 64K. There have been a
few cases of people hitting this limit. The code now uses macros to handle the
storing of links as offsets within the compiled pattern. It defaults to 2-byte
links, but this can be changed to 3 or 4 bytes by --with-link-size when
configuring. Tests 2 and 5 work only with 2-byte links because they output
debugging information about compiled patterns.

33. Internal code re-arrangements:

(a) Moved the debugging function for printing out a compiled regex into
    its own source file (printint.c) and used #include to pull it into
    pcretest.c and, when DEBUG is defined, into pcre.c, instead of having two
    separate copies.

(b) Defined the list of op-code names for debugging as a macro in
    internal.h so that it is next to the definition of the opcodes.

(c) Defined a table of op-code lengths for simpler skipping along compiled
    code. This is again a macro in internal.h so that it is next to the
    definition of the opcodes.

34. Added support for recursive calls to individual subpatterns, along the
lines of Robin Houston's patch (but implemented somewhat differently).

35. Further mods to the Makefile to help Win32. Also, added code to pcregrep to
allow it to read and process whole directories in Win32. This code was
contributed by Lionel Fourquaux; it has not been tested by me.

36. Added support for named subpatterns. The Python syntax (?P<name>...) is
used to name a group. Names consist of alphanumerics and underscores, and must
be unique. Back references use the syntax (?P=name) and recursive calls use
(?P>name) which is a PCRE extension to the Python extension. Groups still have
numbers. The function pcre_fullinfo() can be used after compilation to extract
a name/number map. There are three relevant calls:

  PCRE_INFO_NAMEENTRYSIZE        yields the size of each entry in the map
  PCRE_INFO_NAMECOUNT            yields the number of entries
  PCRE_INFO_NAMETABLE            yields a pointer to the map.

The map is a vector of fixed-size entries. The size of each entry depends on
the length of the longest name used. The first two bytes of each entry are the
group number, most significant byte first. There follows the corresponding
name, zero terminated. The names are in alphabetical order.

37. Make the maximum literal string in the compiled code 250 for the non-UTF-8
case instead of 255. Making it the same both with and without UTF-8 support
means that the same test output works with both.

38. There was a case of malloc(0) in the POSIX testing code in pcretest. Avoid
calling malloc() with a zero argument.

39. Change 25 above had to resort to a heavy-handed test for the .* anchoring
optimization. I've improved things by keeping a bitmap of backreferences with
numbers 1-31 so that if .* occurs inside capturing brackets that are not in
fact referenced, the optimization can be applied. It is unlikely that a
relevant occurrence of .* (i.e. one which might indicate anchoring or forcing
the match to follow \n) will appear inside brackets with a number greater than
31, but if it does, any back reference > 31 suppresses the optimization.

40. Added a new compile-time option PCRE_NO_AUTO_CAPTURE. This has the effect
of disabling numbered capturing parentheses. Any opening parenthesis that is
not followed by ? behaves as if it were followed by ?: but named parentheses
can still be used for capturing (and they will acquire numbers in the usual
way).

41. Redesigned the return codes from the match() function into yes/no/error so
that errors can be passed back from deep inside the nested calls. A malloc
failure while inside a recursive subpattern call now causes the
PCRE_ERROR_NOMEMORY return instead of quietly going wrong.

42. It is now possible to set a limit on the number of times the match()
function is called in a call to pcre_exec(). This facility makes it possible to
limit the amount of recursion and backtracking, though not in a directly
obvious way, because the match() function is used in a number of different
circumstances. The count starts from zero for each position in the subject
string (for non-anchored patterns). The default limit is, for compatibility, a
large number, namely 10 000 000. You can change this in two ways:

(a) When configuring PCRE before making, you can use --with-match-limit=n
    to set a default value for the compiled library.

(b) For each call to pcre_exec(), you can pass a pcre_extra block in which
    a different value is set. See 45 below.

If the limit is exceeded, pcre_exec() returns PCRE_ERROR_MATCHLIMIT.

43. Added a new function pcre_config(int, void *) to enable run-time extraction
of things that can be changed at compile time. The first argument specifies
what is wanted and the second points to where the information is to be placed.
The current list of available information is:

  PCRE_CONFIG_UTF8

The output is an integer that is set to one if UTF-8 support is available;
otherwise it is set to zero.

  PCRE_CONFIG_NEWLINE

The output is an integer that it set to the value of the code that is used for
newline. It is either LF (10) or CR (13).

  PCRE_CONFIG_LINK_SIZE

The output is an integer that contains the number of bytes used for internal
linkage in compiled expressions. The value is 2, 3, or 4. See item 32 above.

  PCRE_CONFIG_POSIX_MALLOC_THRESHOLD

The output is an integer that contains the threshold above which the POSIX
interface uses malloc() for output vectors. See item 31 above.

  PCRE_CONFIG_MATCH_LIMIT

The output is an unsigned integer that contains the default limit of the number
of match() calls in a pcre_exec() execution. See 42 above.

44. pcretest has been upgraded by the addition of the -C option. This causes it
to extract all the available output from the new pcre_config() function, and to
output it. The program then exits immediately.

45. A need has arisen to pass over additional data with calls to pcre_exec() in
order to support additional features. One way would have been to define
pcre_exec2() (for example) with extra arguments, but this would not have been
extensible, and would also have required all calls to the original function to
be mapped to the new one. Instead, I have chosen to extend the mechanism that
is used for passing in "extra" data from pcre_study().

The pcre_extra structure is now exposed and defined in pcre.h. It currently
contains the following fields:

  flags         a bitmap indicating which of the following fields are set
  study_data    opaque data from pcre_study()
  match_limit   a way of specifying a limit on match() calls for a specific
                  call to pcre_exec()
  callout_data  data for callouts (see 49 below)

The flag bits are also defined in pcre.h, and are

  PCRE_EXTRA_STUDY_DATA
  PCRE_EXTRA_MATCH_LIMIT
  PCRE_EXTRA_CALLOUT_DATA

The pcre_study() function now returns one of these new pcre_extra blocks, with
the actual study data pointed to by the study_data field, and the
PCRE_EXTRA_STUDY_DATA flag set. This can be passed directly to pcre_exec() as
before. That is, this change is entirely upwards-compatible and requires no
change to existing code.

If you want to pass in additional data to pcre_exec(), you can either place it
in a pcre_extra block provided by pcre_study(), or create your own pcre_extra
block.

46. pcretest has been extended to test the PCRE_EXTRA_MATCH_LIMIT feature. If a
data string contains the escape sequence \M, pcretest calls pcre_exec() several
times with different match limits, until it finds the minimum value needed for
pcre_exec() to complete. The value is then output. This can be instructive; for
most simple matches the number is quite small, but for pathological cases it
gets very large very quickly.

47. There's a new option for pcre_fullinfo() called PCRE_INFO_STUDYSIZE. It
returns the size of the data block pointed to by the study_data field in a
pcre_extra block, that is, the value that was passed as the argument to
pcre_malloc() when PCRE was getting memory in which to place the information
created by pcre_study(). The fourth argument should point to a size_t variable.
pcretest has been extended so that this information is shown after a successful
pcre_study() call when information about the compiled regex is being displayed.

48. Cosmetic change to Makefile: there's no need to have / after $(DESTDIR)
because what follows is always an absolute path. (Later: it turns out that this
is more than cosmetic for MinGW, because it doesn't like empty path
components.)

49. Some changes have been made to the callout feature (see 28 above):

(i)  A callout function now has three choices for what it returns:

       0  =>  success, carry on matching
     > 0  =>  failure at this point, but backtrack if possible
     < 0  =>  serious error, return this value from pcre_exec()

     Negative values should normally be chosen from the set of PCRE_ERROR_xxx
     values. In particular, returning PCRE_ERROR_NOMATCH forces a standard
     "match failed" error. The error number PCRE_ERROR_CALLOUT is reserved for
     use by callout functions. It will never be used by PCRE itself.

(ii) The pcre_extra structure (see 45 above) has a void * field called
     callout_data, with corresponding flag bit PCRE_EXTRA_CALLOUT_DATA. The
     pcre_callout_block structure has a field of the same name. The contents of
     the field passed in the pcre_extra structure are passed to the callout
     function in the corresponding field in the callout block. This makes it
     easier to use the same callout-containing regex from multiple threads. For
     testing, the pcretest program has a new data escape

       \C*n        pass the number n (may be negative) as callout_data

     If the callout function in pcretest receives a non-zero value as
     callout_data, it returns that value.

50. Makefile wasn't handling CFLAGS properly when compiling dftables. Also,
there were some redundant $(CFLAGS) in commands that are now specified as
$(LINK), which already includes $(CFLAGS).

51. Extensions to UTF-8 support are listed below. These all apply when (a) PCRE
has been compiled with UTF-8 support *and* pcre_compile() has been compiled
with the PCRE_UTF8 flag. Patterns that are compiled without that flag assume
one-byte characters throughout. Note that case-insensitive matching applies
only to characters whose values are less than 256. PCRE doesn't support the
notion of cases for higher-valued characters.

(i)   A character class whose characters are all within 0-255 is handled as
      a bit map, and the map is inverted for negative classes. Previously, a
      character > 255 always failed to match such a class; however it should
      match if the class was a negative one (e.g. [^ab]). This has been fixed.

(ii)  A negated character class with a single character < 255 is coded as
      "not this character" (OP_NOT). This wasn't working properly when the test
      character was multibyte, either singly or repeated.

(iii) Repeats of multibyte characters are now handled correctly in UTF-8
      mode, for example: \x{100}{2,3}.

(iv)  The character escapes \b, \B, \d, \D, \s, \S, \w, and \W (either
      singly or repeated) now correctly test multibyte characters. However,
      PCRE doesn't recognize any characters with values greater than 255 as
      digits, spaces, or word characters. Such characters always match \D, \S,
      and \W, and never match \d, \s, or \w.

(v)   Classes may now contain characters and character ranges with values
      greater than 255. For example: [ab\x{100}-\x{400}].

(vi)  pcregrep now has a --utf-8 option (synonym -u) which makes it call
      PCRE in UTF-8 mode.

52. The info request value PCRE_INFO_FIRSTCHAR has been renamed
PCRE_INFO_FIRSTBYTE because it is a byte value. However, the old name is
retained for backwards compatibility. (Note that LASTLITERAL is also a byte
value.)

53. The single man page has become too large. I have therefore split it up into
a number of separate man pages. These also give rise to individual HTML pages;
these are now put in a separate directory, and there is an index.html page that
lists them all. Some hyperlinking between the pages has been installed.

54. Added convenience functions for handling named capturing parentheses.

55. Unknown escapes inside character classes (e.g. [\M]) and escapes that
aren't interpreted therein (e.g. [\C]) are literals in Perl. This is now also
true in PCRE, except when the PCRE_EXTENDED option is set, in which case they
are faulted.

56. Introduced HOST_CC and HOST_CFLAGS which can be set in the environment when
calling configure. These values are used when compiling the dftables.c program
which is run to generate the source of the default character tables. They
default to the values of CC and CFLAGS. If you are cross-compiling PCRE,
you will need to set these values.

57. Updated the building process for Windows DLL, as provided by Fred Cox.


Version 3.9 02-Jan-02
---------------------

1. A bit of extraneous text had somehow crept into the pcregrep documentation.

2. If --disable-static was given, the building process failed when trying to
build pcretest and pcregrep. (For some reason it was using libtool to compile
them, which is not right, as they aren't part of the library.)


Version 3.8 18-Dec-01
---------------------

1. The experimental UTF-8 code was completely screwed up. It was packing the
bytes in the wrong order. How dumb can you get?


Version 3.7 29-Oct-01
---------------------

1. In updating pcretest to check change 1 of version 3.6, I screwed up.
This caused pcretest, when used on the test data, to segfault. Unfortunately,
this didn't happen under Solaris 8, where I normally test things.

2. The Makefile had to be changed to make it work on BSD systems, where 'make'
doesn't seem to recognize that ./xxx and xxx are the same file. (This entry
isn't in ChangeLog distributed with 3.7 because I forgot when I hastily made
this fix an hour or so after the initial 3.7 release.)


Version 3.6 23-Oct-01
---------------------

1. Crashed with /(sens|respons)e and \1ibility/ and "sense and sensibility" if
offsets passed as NULL with zero offset count.

2. The config.guess and config.sub files had not been updated when I moved to
the latest autoconf.


Version 3.5 15-Aug-01
---------------------

1. Added some missing #if !defined NOPOSIX conditionals in pcretest.c that
had been forgotten.

2. By using declared but undefined structures, we can avoid using "void"
definitions in pcre.h while keeping the internal definitions of the structures
private.

3. The distribution is now built using autoconf 2.50 and libtool 1.4. From a
user point of view, this means that both static and shared libraries are built
by default, but this can be individually controlled. More of the work of
handling this static/shared cases is now inside libtool instead of PCRE's make
file.

4. The pcretest utility is now installed along with pcregrep because it is
useful for users (to test regexs) and by doing this, it automatically gets
relinked by libtool. The documentation has been turned into a man page, so
there are now .1, .txt, and .html versions in /doc.

5. Upgrades to pcregrep:
   (i)   Added long-form option names like gnu grep.
   (ii)  Added --help to list all options with an explanatory phrase.
   (iii) Added -r, --recursive to recurse into sub-directories.
   (iv)  Added -f, --file to read patterns from a file.

6. pcre_exec() was referring to its "code" argument before testing that
argument for NULL (and giving an error if it was NULL).

7. Upgraded Makefile.in to allow for compiling in a different directory from
the source directory.

8. Tiny buglet in pcretest: when pcre_fullinfo() was called to retrieve the
options bits, the pointer it was passed was to an int instead of to an unsigned
long int. This mattered only on 64-bit systems.

9. Fixed typo (3.4/1) in pcre.h again. Sigh. I had changed pcre.h (which is
generated) instead of pcre.in, which it its source. Also made the same change
in several of the .c files.

10. A new release of gcc defines printf() as a macro, which broke pcretest
because it had an ifdef in the middle of a string argument for printf(). Fixed
by using separate calls to printf().

11. Added --enable-newline-is-cr and --enable-newline-is-lf to the configure
script, to force use of CR or LF instead of \n in the source. On non-Unix
systems, the value can be set in config.h.

12. The limit of 200 on non-capturing parentheses is a _nesting_ limit, not an
absolute limit. Changed the text of the error message to make this clear, and
likewise updated the man page.

13. The limit of 99 on the number of capturing subpatterns has been removed.
The new limit is 65535, which I hope will not be a "real" limit.


Version 3.4 22-Aug-00
---------------------

1. Fixed typo in pcre.h: unsigned const char * changed to const unsigned char *.

2. Diagnose condition (?(0) as an error instead of crashing on matching.


Version 3.3 01-Aug-00
---------------------

1. If an octal character was given, but the value was greater than \377, it
was not getting masked to the least significant bits, as documented. This could
lead to crashes in some systems.

2. Perl 5.6 (if not earlier versions) accepts classes like [a-\d] and treats
the hyphen as a literal. PCRE used to give an error; it now behaves like Perl.

3. Added the functions pcre_free_substring() and pcre_free_substring_list().
These just pass their arguments on to (pcre_free)(), but they are provided
because some uses of PCRE bind it to non-C systems that can call its functions,
but cannot call free() or pcre_free() directly.

4. Add "make test" as a synonym for "make check". Corrected some comments in
the Makefile.

5. Add $(DESTDIR)/ in front of all the paths in the "install" target in the
Makefile.

6. Changed the name of pgrep to pcregrep, because Solaris has introduced a
command called pgrep for grepping around the active processes.

7. Added the beginnings of support for UTF-8 character strings.

8. Arranged for the Makefile to pass over the settings of CC, CFLAGS, and
RANLIB to ./ltconfig so that they are used by libtool. I think these are all
the relevant ones. (AR is not passed because ./ltconfig does its own figuring
out for the ar command.)


Version 3.2 12-May-00
---------------------

This is purely a bug fixing release.

1. If the pattern /((Z)+|A)*/ was matched agained ZABCDEFG it matched Z instead
of ZA. This was just one example of several cases that could provoke this bug,
which was introduced by change 9 of version 2.00. The code for breaking
infinite loops after an iteration that matches an empty string was't working
correctly.

2. The pcretest program was not imitating Perl correctly for the pattern /a*/g
when matched against abbab (for example). After matching an empty string, it
wasn't forcing anchoring when setting PCRE_NOTEMPTY for the next attempt; this
caused it to match further down the string than it should.

3. The code contained an inclusion of sys/types.h. It isn't clear why this
was there because it doesn't seem to be needed, and it causes trouble on some
systems, as it is not a Standard C header. It has been removed.

4. Made 4 silly changes to the source to avoid stupid compiler warnings that
were reported on the Macintosh. The changes were from

  while ((c = *(++ptr)) != 0 && c != '\n');
to
  while ((c = *(++ptr)) != 0 && c != '\n') ;

Totally extraordinary, but if that's what it takes...

5. PCRE is being used in one environment where neither memmove() nor bcopy() is
available. Added HAVE_BCOPY and an autoconf test for it; if neither
HAVE_MEMMOVE nor HAVE_BCOPY is set, use a built-in emulation function which
assumes the way PCRE uses memmove() (always moving upwards).

6. PCRE is being used in one environment where strchr() is not available. There
was only one use in pcre.c, and writing it out to avoid strchr() probably gives
faster code anyway.


Version 3.1 09-Feb-00
---------------------

The only change in this release is the fixing of some bugs in Makefile.in for
the "install" target:

(1) It was failing to install pcreposix.h.

(2) It was overwriting the pcre.3 man page with the pcreposix.3 man page.


Version 3.0 01-Feb-00
---------------------

1. Add support for the /+ modifier to perltest (to output $` like it does in
pcretest).

2. Add support for the /g modifier to perltest.

3. Fix pcretest so that it behaves even more like Perl for /g when the pattern
matches null strings.

4. Fix perltest so that it doesn't do unwanted things when fed an empty
pattern. Perl treats empty patterns specially - it reuses the most recent
pattern, which is not what we want. Replace // by /(?#)/ in order to avoid this
effect.

5. The POSIX interface was broken in that it was just handing over the POSIX
captured string vector to pcre_exec(), but (since release 2.00) PCRE has
required a bigger vector, with some working space on the end. This means that
the POSIX wrapper now has to get and free some memory, and copy the results.

6. Added some simple autoconf support, placing the test data and the
documentation in separate directories, re-organizing some of the
information files, and making it build pcre-config (a GNU standard). Also added
libtool support for building PCRE as a shared library, which is now the
default.

7. Got rid of the leading zero in the definition of PCRE_MINOR because 08 and
09 are not valid octal constants. Single digits will be used for minor values
less than 10.

8. Defined REG_EXTENDED and REG_NOSUB as zero in the POSIX header, so that
existing programs that set these in the POSIX interface can use PCRE without
modification.

9. Added a new function, pcre_fullinfo() with an extensible interface. It can
return all that pcre_info() returns, plus additional data. The pcre_info()
function is retained for compatibility, but is considered to be obsolete.

10. Added experimental recursion feature (?R) to handle one common case that
Perl 5.6 will be able to do with (?p{...}).

11. Added support for POSIX character classes like [:alpha:], which Perl is
adopting.


Version 2.08 31-Aug-99
----------------------

1. When startoffset was not zero and the pattern began with ".*", PCRE was not
trying to match at the startoffset position, but instead was moving forward to
the next newline as if a previous match had failed.

2. pcretest was not making use of PCRE_NOTEMPTY when repeating for /g and /G,
and could get into a loop if a null string was matched other than at the start
of the subject.

3. Added definitions of PCRE_MAJOR and PCRE_MINOR to pcre.h so the version can
be distinguished at compile time, and for completeness also added PCRE_DATE.

5. Added Paul Sokolovsky's minor changes to make it easy to compile a Win32 DLL
in GnuWin32 environments.


Version 2.07 29-Jul-99
----------------------

1. The documentation is now supplied in plain text form and HTML as well as in
the form of man page sources.

2. C++ compilers don't like assigning (void *) values to other pointer types.
In particular this affects malloc(). Although there is no problem in Standard
C, I've put in casts to keep C++ compilers happy.

3. Typo on pcretest.c; a cast of (unsigned char *) in the POSIX regexec() call
should be (const char *).

4. If NOPOSIX is defined, pcretest.c compiles without POSIX support. This may
be useful for non-Unix systems who don't want to bother with the POSIX stuff.
However, I haven't made this a standard facility. The documentation doesn't
mention it, and the Makefile doesn't support it.

5. The Makefile now contains an "install" target, with editable destinations at
the top of the file. The pcretest program is not installed.

6. pgrep -V now gives the PCRE version number and date.

7. Fixed bug: a zero repetition after a literal string (e.g. /abcde{0}/) was
causing the entire string to be ignored, instead of just the last character.

8. If a pattern like /"([^\\"]+|\\.)*"/ is applied in the normal way to a
non-matching string, it can take a very, very long time, even for strings of
quite modest length, because of the nested recursion. PCRE now does better in
some of these cases. It does this by remembering the last required literal
character in the pattern, and pre-searching the subject to ensure it is present
before running the real match. In other words, it applies a heuristic to detect
some types of certain failure quickly, and in the above example, if presented
with a string that has no trailing " it gives "no match" very quickly.

9. A new runtime option PCRE_NOTEMPTY causes null string matches to be ignored;
other alternatives are tried instead.


Version 2.06 09-Jun-99
----------------------

1. Change pcretest's output for amount of store used to show just the code
space, because the remainder (the data block) varies in size between 32-bit and
64-bit systems.

2. Added an extra argument to pcre_exec() to supply an offset in the subject to
start matching at. This allows lookbehinds to work when searching for multiple
occurrences in a string.

3. Added additional options to pcretest for testing multiple occurrences:

   /+   outputs the rest of the string that follows a match
   /g   loops for multiple occurrences, using the new startoffset argument
   /G   loops for multiple occurrences by passing an incremented pointer

4. PCRE wasn't doing the "first character" optimization for patterns starting
with \b or \B, though it was doing it for other lookbehind assertions. That is,
it wasn't noticing that a match for a pattern such as /\bxyz/ has to start with
the letter 'x'. On long subject strings, this gives a significant speed-up.


Version 2.05 21-Apr-99
----------------------

1. Changed the type of magic_number from int to long int so that it works
properly on 16-bit systems.

2. Fixed a bug which caused patterns starting with .* not to work correctly
when the subject string contained newline characters. PCRE was assuming
anchoring for such patterns in all cases, which is not correct because .* will
not pass a newline unless PCRE_DOTALL is set. It now assumes anchoring only if
DOTALL is set at top level; otherwise it knows that patterns starting with .*
must be retried after every newline in the subject.


Version 2.04 18-Feb-99
----------------------

1. For parenthesized subpatterns with repeats whose minimum was zero, the
computation of the store needed to hold the pattern was incorrect (too large).
If such patterns were nested a few deep, this could multiply and become a real
problem.

2. Added /M option to pcretest to show the memory requirement of a specific
pattern. Made -m a synonym of -s (which does this globally) for compatibility.

3. Subpatterns of the form (regex){n,m} (i.e. limited maximum) were being
compiled in such a way that the backtracking after subsequent failure was
pessimal. Something like (a){0,3} was compiled as (a)?(a)?(a)? instead of
((a)((a)(a)?)?)? with disastrous performance if the maximum was of any size.


Version 2.03 02-Feb-99
----------------------

1. Fixed typo and small mistake in man page.

2. Added 4th condition (GPL supersedes if conflict) and created separate
LICENCE file containing the conditions.

3. Updated pcretest so that patterns such as /abc\/def/ work like they do in
Perl, that is the internal \ allows the delimiter to be included in the
pattern. Locked out the use of \ as a delimiter. If \ immediately follows
the final delimiter, add \ to the end of the pattern (to test the error).

4. Added the convenience functions for extracting substrings after a successful
match. Updated pcretest to make it able to test these functions.


Version 2.02 14-Jan-99
----------------------

1. Initialized the working variables associated with each extraction so that
their saving and restoring doesn't refer to uninitialized store.

2. Put dummy code into study.c in order to trick the optimizer of the IBM C
compiler for OS/2 into generating correct code. Apparently IBM isn't going to
fix the problem.

3. Pcretest: the timing code wasn't using LOOPREPEAT for timing execution
calls, and wasn't printing the correct value for compiling calls. Increased the
default value of LOOPREPEAT, and the number of significant figures in the
times.

4. Changed "/bin/rm" in the Makefile to "-rm" so it works on Windows NT.

5. Renamed "deftables" as "dftables" to get it down to 8 characters, to avoid
a building problem on Windows NT with a FAT file system.


Version 2.01 21-Oct-98
----------------------

1. Changed the API for pcre_compile() to allow for the provision of a pointer
to character tables built by pcre_maketables() in the current locale. If NULL
is passed, the default tables are used.


Version 2.00 24-Sep-98
----------------------

1. Since the (>?) facility is in Perl 5.005, don't require PCRE_EXTRA to enable
it any more.

2. Allow quantification of (?>) groups, and make it work correctly.

3. The first character computation wasn't working for (?>) groups.

4. Correct the implementation of \Z (it is permitted to match on the \n at the
end of the subject) and add 5.005's \z, which really does match only at the
very end of the subject.

5. Remove the \X "cut" facility; Perl doesn't have it, and (?> is neater.

6. Remove the ability to specify CASELESS, MULTILINE, DOTALL, and
DOLLAR_END_ONLY at runtime, to make it possible to implement the Perl 5.005
localized options. All options to pcre_study() were also removed.

7. Add other new features from 5.005:

   $(?<=           positive lookbehind
   $(?<!           negative lookbehind
   (?imsx-imsx)    added the unsetting capability
                   such a setting is global if at outer level; local otherwise
   (?imsx-imsx:)   non-capturing groups with option setting
   (?(cond)re|re)  conditional pattern matching

   A backreference to itself in a repeated group matches the previous
   captured string.

8. General tidying up of studying (both automatic and via "study")
consequential on the addition of new assertions.

9. As in 5.005, unlimited repeated groups that could match an empty substring
are no longer faulted at compile time. Instead, the loop is forcibly broken at
runtime if any iteration does actually match an empty substring.

10. Include the RunTest script in the distribution.

11. Added tests from the Perl 5.005_02 distribution. This showed up a few
discrepancies, some of which were old and were also with respect to 5.004. They
have now been fixed.


Version 1.09 28-Apr-98
----------------------

1. A negated single character class followed by a quantifier with a minimum
value of one (e.g.  [^x]{1,6}  ) was not compiled correctly. This could lead to
program crashes, or just wrong answers. This did not apply to negated classes
containing more than one character, or to minima other than one.


Version 1.08 27-Mar-98
----------------------

1. Add PCRE_UNGREEDY to invert the greediness of quantifiers.

2. Add (?U) and (?X) to set PCRE_UNGREEDY and PCRE_EXTRA respectively. The
latter must appear before anything that relies on it in the pattern.


Version 1.07 16-Feb-98
----------------------

1. A pattern such as /((a)*)*/ was not being diagnosed as in error (unlimited
repeat of a potentially empty string).


Version 1.06 23-Jan-98
----------------------

1. Added Markus Oberhumer's little patches for C++.

2. Literal strings longer than 255 characters were broken.


Version 1.05 23-Dec-97
----------------------

1. Negated character classes containing more than one character were failing if
PCRE_CASELESS was set at run time.


Version 1.04 19-Dec-97
----------------------

1. Corrected the man page, where some "const" qualifiers had been omitted.

2. Made debugging output print "{0,xxx}" instead of just "{,xxx}" to agree with
input syntax.

3. Fixed memory leak which occurred when a regex with back references was
matched with an offsets vector that wasn't big enough. The temporary memory
that is used in this case wasn't being freed if the match failed.

4. Tidied pcretest to ensure it frees memory that it gets.

5. Temporary memory was being obtained in the case where the passed offsets
vector was exactly big enough.

6. Corrected definition of offsetof() from change 5 below.

7. I had screwed up change 6 below and broken the rules for the use of
setjmp(). Now fixed.


Version 1.03 18-Dec-97
----------------------

1. A erroneous regex with a missing opening parenthesis was correctly
diagnosed, but PCRE attempted to access brastack[-1], which could cause crashes
on some systems.

2. Replaced offsetof(real_pcre, code) by offsetof(real_pcre, code[0]) because
it was reported that one broken compiler failed on the former because "code" is
also an independent variable.

3. The erroneous regex a[]b caused an array overrun reference.

4. A regex ending with a one-character negative class (e.g. /[^k]$/) did not
fail on data ending with that character. (It was going on too far, and checking
the next character, typically a binary zero.) This was specific to the
optimized code for single-character negative classes.

5. Added a contributed patch from the TIN world which does the following:

  + Add an undef for memmove, in case the the system defines a macro for it.

  + Add a definition of offsetof(), in case there isn't one. (I don't know
    the reason behind this - offsetof() is part of the ANSI standard - but
    it does no harm).

  + Reduce the ifdef's in pcre.c using macro DPRINTF, thereby eliminating
    most of the places where whitespace preceded '#'. I have given up and
    allowed the remaining 2 cases to be at the margin.

  + Rename some variables in pcre to eliminate shadowing. This seems very
    pedantic, but does no harm, of course.

6. Moved the call to setjmp() into its own function, to get rid of warnings
from gcc -Wall, and avoided calling it at all unless PCRE_EXTRA is used.

7. Constructs such as \d{8,} were compiling into the equivalent of
\d{8}\d{0,65527} instead of \d{8}\d* which didn't make much difference to the
outcome, but in this particular case used more store than had been allocated,
which caused the bug to be discovered because it threw up an internal error.

8. The debugging code in both pcre and pcretest for outputting the compiled
form of a regex was going wrong in the case of back references followed by
curly-bracketed repeats.


Version 1.02 12-Dec-97
----------------------

1. Typos in pcre.3 and comments in the source fixed.

2. Applied a contributed patch to get rid of places where it used to remove
'const' from variables, and fixed some signed/unsigned and uninitialized
variable warnings.

3. Added the "runtest" target to Makefile.

4. Set default compiler flag to -O2 rather than just -O.


Version 1.01 19-Nov-97
----------------------

1. PCRE was failing to diagnose unlimited repeat of empty string for patterns
like /([ab]*)*/, that is, for classes with more than one character in them.

2. Likewise, it wasn't diagnosing patterns with "once-only" subpatterns, such
as /((?>a*))*/ (a PCRE_EXTRA facility).


Version 1.00 18-Nov-97
----------------------

1. Added compile-time macros to support systems such as SunOS4 which don't have
memmove() or strerror() but have other things that can be used instead.

2. Arranged that "make clean" removes the executables.


Version 0.99 27-Oct-97
----------------------

1. Fixed bug in code for optimizing classes with only one character. It was
initializing a 32-byte map regardless, which could cause it to run off the end
of the memory it had got.

2. Added, conditional on PCRE_EXTRA, the proposed (?>REGEX) construction.


Version 0.98 22-Oct-97
----------------------

1. Fixed bug in code for handling temporary memory usage when there are more
back references than supplied space in the ovector. This could cause segfaults.


Version 0.97 21-Oct-97
----------------------

1. Added the \X "cut" facility, conditional on PCRE_EXTRA.

2. Optimized negated single characters not to use a bit map.

3. Brought error texts together as macro definitions; clarified some of them;
fixed one that was wrong - it said "range out of order" when it meant "invalid
escape sequence".

4. Changed some char * arguments to const char *.

5. Added PCRE_NOTBOL and PCRE_NOTEOL (from POSIX).

6. Added the POSIX-style API wrapper in pcreposix.a and testing facilities in
pcretest.


Version 0.96 16-Oct-97
----------------------

1. Added a simple "pgrep" utility to the distribution.

2. Fixed an incompatibility with Perl: "{" is now treated as a normal character
unless it appears in one of the precise forms "{ddd}", "{ddd,}", or "{ddd,ddd}"
where "ddd" means "one or more decimal digits".

3. Fixed serious bug. If a pattern had a back reference, but the call to
pcre_exec() didn't supply a large enough ovector to record the related
identifying subpattern, the match always failed. PCRE now remembers the number
of the largest back reference, and gets some temporary memory in which to save
the offsets during matching if necessary, in order to ensure that
backreferences always work.

4. Increased the compatibility with Perl in a number of ways:

  (a) . no longer matches \n by default; an option PCRE_DOTALL is provided
      to request this handling. The option can be set at compile or exec time.

  (b) $ matches before a terminating newline by default; an option
      PCRE_DOLLAR_ENDONLY is provided to override this (but not in multiline
      mode). The option can be set at compile or exec time.

  (c) The handling of \ followed by a digit other than 0 is now supposed to be
      the same as Perl's. If the decimal number it represents is less than 10
      or there aren't that many previous left capturing parentheses, an octal
      escape is read. Inside a character class, it's always an octal escape,
      even if it is a single digit.

  (d) An escaped but undefined alphabetic character is taken as a literal,
      unless PCRE_EXTRA is set. Currently this just reserves the remaining
      escapes.

  (e) {0} is now permitted. (The previous item is removed from the compiled
      pattern).

5. Changed all the names of code files so that the basic parts are no longer
than 10 characters, and abolished the teeny "globals.c" file.

6. Changed the handling of character classes; they are now done with a 32-byte
bit map always.

7. Added the -d and /D options to pcretest to make it possible to look at the
internals of compilation without having to recompile pcre.


Version 0.95 23-Sep-97
----------------------

1. Fixed bug in pre-pass concerning escaped "normal" characters such as \x5c or
\x20 at the start of a run of normal characters. These were being treated as
real characters, instead of the source characters being re-checked.


Version 0.94 18-Sep-97
----------------------

1. The functions are now thread-safe, with the caveat that the global variables
containing pointers to malloc() and free() or alternative functions are the
same for all threads.

2. Get pcre_study() to generate a bitmap of initial characters for non-
anchored patterns when this is possible, and use it if passed to pcre_exec().


Version 0.93 15-Sep-97
----------------------

1. /(b)|(:+)/ was computing an incorrect first character.

2. Add pcre_study() to the API and the passing of pcre_extra to pcre_exec(),
but not actually doing anything yet.

3. Treat "-" characters in classes that cannot be part of ranges as literals,
as Perl does (e.g. [-az] or [az-]).

4. Set the anchored flag if a branch starts with .* or .*? because that tests
all possible positions.

5. Split up into different modules to avoid including unneeded functions in a
compiled binary. However, compile and exec are still in one module. The "study"
function is split off.

6. The character tables are now in a separate module whose source is generated
by an auxiliary program - but can then be edited by hand if required. There are
now no calls to isalnum(), isspace(), isdigit(), isxdigit(), tolower() or
toupper() in the code.

7. Turn the malloc/free funtions variables into pcre_malloc and pcre_free and
make them global. Abolish the function for setting them, as the caller can now
set them directly.


Version 0.92 11-Sep-97
----------------------

1. A repeat with a fixed maximum and a minimum of 1 for an ordinary character
(e.g. /a{1,3}/) was broken (I mis-optimized it).

2. Caseless matching was not working in character classes if the characters in
the pattern were in upper case.

3. Make ranges like [W-c] work in the same way as Perl for caseless matching.

4. Make PCRE_ANCHORED public and accept as a compile option.

5. Add an options word to pcre_exec() and accept PCRE_ANCHORED and
PCRE_CASELESS at run time. Add escapes \A and \I to pcretest to cause it to
pass them.

6. Give an error if bad option bits passed at compile or run time.

7. Add PCRE_MULTILINE at compile and exec time, and (?m) as well. Add \M to
pcretest to cause it to pass that flag.

8. Add pcre_info(), to get the number of identifying subpatterns, the stored
options, and the first character, if set.

9. Recognize C+ or C{n,m} where n >= 1 as providing a fixed starting character.


Version 0.91 10-Sep-97
----------------------

1. PCRE was failing to diagnose unlimited repeats of subpatterns that could
match the empty string as in /(a*)*/. It was looping and ultimately crashing.

2. PCRE was looping on encountering an indefinitely repeated back reference to
a subpattern that had matched an empty string, e.g. /(a|)\1*/. It now does what
Perl does - treats the match as successful.

****