pcre/ChangeLog

Change Log for PCRE2
--------------------


Version 10.30-DEV 09-March-2017
-------------------------------

1. The main interpreter, pcre2_match(), has been refactored into a new version 
that does not use recursive function calls (and therefore the stack) for 
remembering backtracking positions. This makes --disable-stack-for-recursion a 
NOOP. The new implementation allows backtracking into recursive group calls in
patterns, making it more compatible with Perl, and also fixes some other 
hard-to-do issues such as #1887 in Bugzilla. The code is also cleaner because 
the old code had a number of fudges to try to reduce stack usage. It seems to 
run no slower than the old code.

A number of bugs in the refactored code were subsequently fixed during testing
before release, but after the code was made available in the repository. These
bugs were never in fully released code, but are noted here for the record.

  (a) If a pattern had fewer capturing parentheses than the ovector supplied in 
      the match data block, a memory error (detectable by ASAN) occurred after
      a match, because the external block was being set from non-existent
      internal ovector fields. Fixes oss-fuzz issue 781.

  (b) A pattern with very many capturing parentheses (when the internal frame 
      size was greater than the initial frame vector on the stack) caused a 
      crash. A vector on the heap is now set up at the start of matching if the 
      vector on the stack is not big enough to handle at least 10 frames. 
      Fixes oss-fuzz issue 783. 
      
  (c) Handling of (*VERB)s in recursions was wrong in some cases. 
      
2. Now that pcre2_match() no longer uses recursive function calls (see above),
the "match limit recursion" value seems misnamed. It still exists, and limits 
the depth of tree that is searched. To avoid future confusion, it has been 
renamed as "depth limit" in all relevant places (--with-depth-limit, 
(*LIMIT_DEPTH), pcre2_set_depth_limit(), etc) but the old names are still 
available for backwards compatibility. 
       
3. Hardened pcre2test so as to reduce the number of bugs reported by fuzzers:

  (a) Check for malloc failures when getting memory for the ovector (POSIX) or 
      the match data block (non-POSIX). 
      
4. In the 32-bit library in non-UTF mode, an attempt to find a Unicode property
for a character with a code point greater than 0x10ffff (the Unicode maximum)
caused a crash.

5. If a lookbehind assertion that contained a back reference to a group 
appearing later in the pattern was compiled with the PCRE2_ANCHORED option,
undefined actions (often a segmentation fault) could occur, depending on what 
other options were set. An example assertion is (?<!\1(abc)) where the 
reference \1 precedes the group (abc). This fixes oss-fuzz issue 865.

6. Added the PCRE2_INFO_FRAMESIZE item to pcre2_pattern_info() and arranged for 
pcre2test to use it to output the frame size when the "framesize" modifier is
given.

7. Reworked the recursive pattern matching in the JIT compiler to follow the
interpreter changes.

8. When the zero_terminate modifier was specified on a pcre2test subject line 
for global matching, unpredictable things could happen. For example, in UTF-8 
mode, the pattern //g,zero_terminate read random memory when matched against an 
empty string with zero_terminate. This was a bug in pcre2test, not the library.

9. Moved some Windows-specific code in pcre2grep (introduced in 10.23/13) out
of the section that is compiled when Unix-style directory scanning is
available, and into a new section that is always compiled for Windows.

10. In pcre2test, explicitly close the file after an error during serialization 
or deserialization (the "load" or "save" commands).

11. Fix memory leak in pcre2_serialize_decode() when the input is invalid.

12. Fix potential NULL dereference in pcre2_callout_enumerate() if called with 
a NULL pattern pointer when Unicode support is available.

13. When the 32-bit library was being tested by pcre2test, error messages that 
were longer than 64 code units could cause a buffer overflow. This was a bug in 
pcre2test.

14. The alternative matching function, pcre2_dfa_match() misbehaved if it 
encountered a character class with a possessive repeat, for example [a-f]{3}+.

15. The depth (formerly recursion) limit now applies to DFA matching (as 
of 10.23/36); pcre2test has been upgraded so that \=find_limits works with DFA 
matching to find the minimum value for this limit.

16. Since 10.21, if pcre2_match() was called with a null context, default 
memory allocation functions were used instead of whatever was used when the 
pattern was compiled.

17. Changes to the pcre2test "memory" modifier on a subject line. These apply
only to pcre2_match():

  (a) Warn if null_context is set on both pattern and subject, because the
      memory details cannot then be shown.
      
  (b) Remember (up to a certain number of) memory allocations and their
      lengths, and list only the lengths, so as to be system-independent.
      (In practice, the new interpreter never has more than 2 blocks allocated 
      simultaneously.) 
      
18. Make pcre2test detect an error return from pcre2_get_error_message(), give
a message, and abandon the run (this would have detected #13 above). 


Version 10.23 14-February-2017
------------------------------

1. Extended pcre2test with the utf8_input modifier so that it is able to
generate all possible 16-bit and 32-bit code unit values in non-UTF modes.

2. In any wide-character mode (8-bit UTF or any 16-bit or 32-bit mode), without
PCRE2_UCP set, a negative character type such as \D in a positive class should
cause all characters greater than 255 to match, whatever else is in the class.
There was a bug that caused this not to happen if a Unicode property item was
added to such a class, for example [\D\P{Nd}] or [\W\pL].

3. There has been a major re-factoring of the pcre2_compile.c file. Most syntax
checking is now done in the pre-pass that identifies capturing groups. This has
reduced the amount of duplication and made the code tidier. While doing this,
some minor bugs and Perl incompatibilities were fixed, including:

  (a) \Q\E in the middle of a quantifier such as A+\Q\E+ is now ignored instead
      of giving an invalid quantifier error.

  (b) {0} can now be used after a group in a lookbehind assertion; previously
      this caused an "assertion is not fixed length" error.

  (c) Perl always treats (?(DEFINE) as a "define" group, even if a group with
      the name "DEFINE" exists. PCRE2 now does likewise.

  (d) A recursion condition test such as (?(R2)...) must now refer to an
      existing subpattern.

  (e) A conditional recursion test such as (?(R)...) misbehaved if there was a
      group whose name began with "R".

  (f) When testing zero-terminated patterns under valgrind, the terminating
      zero is now marked "no access". This catches bugs that would otherwise
      show up only with non-zero-terminated patterns.

  (g) A hyphen appearing immediately after a POSIX character class (for example
      /[[:ascii:]-z]/) now generates an error. Perl does accept this as a
      literal, but gives a warning, so it seems best to fail it in PCRE.

  (h) An empty \Q\E sequence may appear after a callout that precedes an
      assertion condition (it is, of course, ignored).

One effect of the refactoring is that some error numbers and messages have
changed, and the pattern offset given for compiling errors is not always the
right-most character that has been read. In particular, for a variable-length
lookbehind assertion it now points to the start of the assertion. Another
change is that when a callout appears before a group, the "length of next
pattern item" that is passed now just gives the length of the opening
parenthesis item, not the length of the whole group. A length of zero is now
given only for a callout at the end of the pattern. Automatic callouts are no
longer inserted before and after explicit callouts in the pattern.

A number of bugs in the refactored code were subsequently fixed during testing
before release, but after the code was made available in the repository. Many
of the bugs were discovered by fuzzing testing. Several of them were related to
the change from assuming a zero-terminated pattern (which previously had
required non-zero terminated strings to be copied). These bugs were never in
fully released code, but are noted here for the record.

  (a) An overall recursion such as (?0) inside a lookbehind assertion was not
      being diagnosed as an error.

  (b) In utf mode, the length of a *MARK (or other verb) name was being checked
      in characters instead of code units, which could lead to bad code being
      compiled, leading to unpredictable behaviour.

  (c) In extended /x mode, characters whose code was greater than 255 caused
      a lookup outside one of the global tables. A similar bug existed for wide
      characters in *VERB names.

  (d) The amount of memory needed for a compiled pattern was miscalculated if a
      lookbehind contained more than one toplevel branch and the first branch
      was of length zero.

  (e) In UTF-8 or UTF-16 modes with PCRE2_EXTENDED (/x) set and a non-zero-
      terminated pattern, if a # comment ran on to the end of the pattern, one
      or more code units past the end were being read.

  (f) An unterminated repeat at the end of a non-zero-terminated pattern (e.g.
      "{2,2") could cause reading beyond the pattern.

  (g) When reading a callout string, if the end delimiter was at the end of the
      pattern one further code unit was read.

  (h) An unterminated number after \g' could cause reading beyond the pattern.

  (i) An insufficient memory size was being computed for compiling with
      PCRE2_AUTO_CALLOUT.

  (j) A conditional group with an assertion condition used more memory than was
      allowed for it during parsing, so too many of them could therefore
      overrun a buffer.

  (k) If parsing a pattern exactly filled the buffer, the internal test for
      overrun did not check when the final META_END item was added.

  (l) If a lookbehind contained a subroutine call, and the called group
      contained an option setting such as (?s), and the PCRE2_ANCHORED option
      was set, unpredictable behaviour could occur. The underlying bug was
      incorrect code and insufficient checking while searching for the end of
      the called subroutine in the parsed pattern.

  (m) Quantifiers following (*VERB)s were not being diagnosed as errors.

  (n) The use of \Q...\E in a (*VERB) name when PCRE2_ALT_VERBNAMES and
      PCRE2_AUTO_CALLOUT were both specified caused undetermined behaviour.

  (o) If \Q was preceded by a quantified item, and the following \E was
      followed by '?' or '+', and there was at least one literal character
      between them, an internal error "unexpected repeat" occurred (example:
      /.+\QX\E+/).

  (p) A buffer overflow could occur while sorting the names in the group name
      list (depending on the order in which the names were seen).

  (q) A conditional group that started with a callout was not doing the right
      check for a following assertion, leading to compiling bad code. Example:
      /(?(C'XX))?!XX/

  (r) If a character whose code point was greater than 0xffff appeared within
      a lookbehind that was within another lookbehind, the calculation of the
      lookbehind length went wrong and could provoke an internal error.

  (t) The sequence \E- or \Q\E- after a POSIX class in a character class caused
      an internal error. Now the hyphen is treated as a literal.

4. Back references are now permitted in lookbehind assertions when there are
no duplicated group numbers (that is, (?| has not been used), and, if the
reference is by name, there is only one group of that name. The referenced
group must, of course be of fixed length.

5. pcre2test has been upgraded so that, when run under valgrind with valgrind
support enabled, reading past the end of the pattern is detected, both when
compiling and during callout processing.

6. \g{+<number>} (e.g. \g{+2} ) is now supported. It is a "forward back
reference" and can be useful in repetitions (compare \g{-<number>} ). Perl does
not recognize this syntax.

7. Automatic callouts are no longer generated before and after callouts in the
pattern.

8. When pcre2test was outputing information from a callout, the caret indicator
for the current position in the subject line was incorrect if it was after an
escape sequence for a character whose code point was greater than \x{ff}.

9. Change 19 for 10.22 had a typo (PCRE_STATIC_RUNTIME should be
PCRE2_STATIC_RUNTIME). Fix from David Gaussmann.

10. Added --max-buffer-size to pcre2grep, to allow for automatic buffer
expansion when long lines are encountered. Original patch by Dmitry
Cherniachenko.

11. If pcre2grep was compiled with JIT support, but the library was compiled
without it (something that neither ./configure nor CMake allow, but it can be
done by editing config.h), pcre2grep was giving a JIT error. Now it detects
this situation and does not try to use JIT.

12. Added some "const" qualifiers to variables in pcre2grep.

13. Added Dmitry Cherniachenko's patch for colouring output in Windows
(untested by me). Also, look for GREP_COLOUR or GREP_COLOR if the environment
variables PCRE2GREP_COLOUR and PCRE2GREP_COLOR are not found.

14. Add the -t (grand total) option to pcre2grep.

15. A number of bugs have been mended relating to match start-up optimizations
when the first thing in a pattern is a positive lookahead. These all applied
only when PCRE2_NO_START_OPTIMIZE was *not* set:

    (a) A pattern such as (?=.*X)X$ was incorrectly optimized as if it needed
        both an initial 'X' and a following 'X'.
    (b) Some patterns starting with an assertion that started with .* were
        incorrectly optimized as having to match at the start of the subject or
        after a newline. There are cases where this is not true, for example,
        (?=.*[A-Z])(?=.{8,16})(?!.*[\s]) matches after the start in lines that
        start with spaces. Starting .* in an assertion is no longer taken as an
        indication of matching at the start (or after a newline).

16. The "offset" modifier in pcre2test was not being ignored (as documented)
when the POSIX API was in use.

17. Added --enable-fuzz-support to "configure", causing an non-installed
library containing a test function that can be called by fuzzers to be
compiled. A non-installed  binary to run the test function locally, called
pcre2fuzzcheck is also compiled.

18. A pattern with PCRE2_DOTALL (/s) set but not PCRE2_NO_DOTSTAR_ANCHOR, and
which started with .* inside a positive lookahead was incorrectly being
compiled as implicitly anchored.

19. Removed all instances of "register" declarations, as they are considered
obsolete these days and in any case had become very haphazard.

20. Add strerror() to pcre2test for failed file opening.

21. Make pcre2test -C list valgrind support when it is enabled.

22. Add the use_length modifier to pcre2test.

23. Fix an off-by-one bug in pcre2test for the list of names for 'get' and
'copy' modifiers.

24. Add PCRE2_CALL_CONVENTION into the prototype declarations in pcre2.h as it
is apparently needed there as well as in the function definitions. (Why did
nobody ask for this in PCRE1?)

25. Change the _PCRE2_H and _PCRE2_UCP_H guard macros in the header files to
PCRE2_H_IDEMPOTENT_GUARD and PCRE2_UCP_H_IDEMPOTENT_GUARD to be more standard
compliant and unique.

26. pcre2-config --libs-posix was listing -lpcre2posix instead of
-lpcre2-posix. Also, the CMake build process was building the library with the
wrong name.

27. In pcre2test, give some offset information for errors in hex patterns.
This uses the C99 formatting sequence %td, except for MSVC which doesn't
support it - %lu is used instead.

28. Implemented pcre2_code_copy_with_tables(), and added pushtablescopy to
pcre2test for testing it.

29. Fix small memory leak in pcre2test.

30. Fix out-of-bounds read for partial matching of /./ against an empty string
when the newline type is CRLF.

31. Fix a bug in pcre2test that caused a crash when a locale was set either in
the current pattern or a previous one and a wide character was matched.

32. The appearance of \p, \P, or \X in a substitution string when
PCRE2_SUBSTITUTE_EXTENDED was set caused a segmentation fault (NULL
dereference).

33. If the starting offset was specified as greater than the subject length in
a call to pcre2_substitute() an out-of-bounds memory reference could occur.

34. When PCRE2 was compiled to use the heap instead of the stack for recursive
calls to match(), a repeated minimizing caseless back reference, or a
maximizing one where the two cases had different numbers of code units,
followed by a caseful back reference, could lose the caselessness of the first
repeated back reference (example: /(Z)(a)\2{1,2}?(?-i)\1X/i should match ZaAAZX
but didn't).

35. When a pattern is too complicated, PCRE2 gives up trying to find a minimum
matching length and just records zero. Typically this happens when there are
too many nested or recursive back references. If the limit was reached in
certain recursive cases it failed to be triggered and an internal error could
be the result.

36. The pcre2_dfa_match() function now takes note of the recursion limit for
the internal recursive calls that are used for lookrounds and recursions within
the pattern.

37. More refactoring has got rid of the internal could_be_empty_branch()
function (around 400 lines of code, including comments) by keeping track of
could-be-emptiness as the pattern is compiled instead of scanning compiled
groups. (This would have been much harder before the refactoring of #3 above.)
This lifts a restriction on the number of branches in a group (more than about
1100 would give "pattern is too complicated").

38. Add the "-ac" command line option to pcre2test as a synonym for "-pattern
auto_callout".

39. In a library with Unicode support, incorrect data was compiled for a
pattern with PCRE2_UCP set without PCRE2_UTF if a class required all wide
characters to match (for example, /[\s[:^ascii:]]/).

40. The callout_error modifier has been added to pcre2test to make it possible
to return PCRE2_ERROR_CALLOUT from a callout.

41. A minor change to pcre2grep: colour reset is now "<esc>[0m" instead of
"<esc>[00m".

42. The limit in the auto-possessification code that was intended to catch
overly-complicated patterns and not spend too much time auto-possessifying was
being reset too often, resulting in very long compile times for some patterns.
Now such patterns are no longer completely auto-possessified.

43. Applied Jason Hood's revised patch for RunTest.bat.

44. Added a new Windows script RunGrepTest.bat, courtesy of Jason Hood.

45. Minor cosmetic fix to pcre2test: move a variable that is not used under
Windows into the "not Windows" code.

46. Applied Jason Hood's patches to upgrade pcre2grep under Windows and tidy
some of the code:

  * normalised the Windows condition by ensuring WIN32 is defined;
  * enables the callout feature under Windows;
  * adds globbing (Microsoft's implementation expands quoted args),
    using a tweaked opendirectory;
  * implements the is_*_tty functions for Windows;
  * --color=always will write the ANSI sequences to file;
  * add sequences 4 (underline works on Win10) and 5 (blink as bright
    background, relatively standard on DOS/Win);
  * remove the (char *) casts for the now-const strings;
  * remove GREP_COLOUR (grep's command line allowed the 'u', but not
    the environment), parsing GREP_COLORS instead;
  * uses the current colour if not set, rather than black;
  * add print_match for the undefined case;
  * fixes a typo.

In addition, colour settings containing anything other than digits and
semicolon are ignored, and the colour controls are no longer output for empty
strings.

47. Detecting patterns that are too large inside the length-measuring loop
saves processing ridiculously long patterns to their end.

48. Ignore PCRE2_CASELESS when processing \h, \H, \v, and \V in classes as it
just wastes time. In the UTF case it can also produce redundant entries in
XCLASS lists caused by characters with multiple other cases and pairs of
characters in the same "not-x" sublists.

49. A pattern such as /(?=(a\K))/ can report the end of the match being before
its start; pcre2test was not handling this correctly when using the POSIX
interface (it was OK with the native interface).

50. In pcre2grep, ignore all JIT compile errors. This means that pcre2grep will
continue to work, falling back to interpretation if anything goes wrong with
JIT.

51. Applied patches from Christian Persch to configure.ac to make use of the
AC_USE_SYSTEM_EXTENSIONS macro and to test for functions used by the JIT
modules.

52. Minor fixes to pcre2grep from Jason Hood:
    * fixed some spacing;
    * Windows doesn't usually use single quotes, so I've added a define
      to use appropriate quotes [in an example];
    * LC_ALL was displayed as "LCC_ALL";
    * numbers 11, 12 & 13 should end in "th";
    * use double quotes in usage message.

53. When autopossessifying, skip empty branches without recursion, to reduce
stack usage for the benefit of clang with -fsanitize-address, which uses huge
stack frames. Example pattern: /X?(R||){3335}/. Fixes oss-fuzz issue 553.

54. A pattern with very many explicit back references to a group that is a long
way from the start of the pattern could take a long time to compile because
searching for the referenced group in order to find the minimum length was
being done repeatedly. Now up to 128 group minimum lengths are cached and the
attempt to find a minimum length is abandoned if there is a back reference to a
group whose number is greater than 128. (In that case, the pattern is so
complicated that this optimization probably isn't worth it.) This fixes
oss-fuzz issue 557.

55. Issue 32 for 10.22 below was not correctly fixed. If pcre2grep in multiline
mode with --only-matching matched several lines, it restarted scanning at the
next line instead of moving on to the end of the matched string, which can be
several lines after the start.

56. Applied Jason Hood's new patch for RunGrepTest.bat that updates it in line
with updates to the non-Windows version.



Version 10.22 29-July-2016
--------------------------

1. Applied Jason Hood's patches to RunTest.bat and testdata/wintestoutput3
to fix problems with running the tests under Windows.

2. Implemented a facility for quoting literal characters within hexadecimal
patterns in pcre2test, to make it easier to create patterns with just a few
non-printing characters.

3. Binary zeros are not supported in pcre2test input files. It now detects them
and gives an error.

4. Updated the valgrind parameters in RunTest: (a) changed smc-check=all to
smc-check=all-non-file; (b) changed obj:* in the suppression file to obj:??? so
that it matches only unknown objects.

5. Updated the maintenance script maint/ManyConfigTests to make it easier to
select individual groups of tests.

6. When the POSIX wrapper function regcomp() is called, the REG_NOSUB option
used to set PCRE2_NO_AUTO_CAPTURE when calling pcre2_compile(). However, this
disables the use of back references (and subroutine calls), which are supported
by other implementations of regcomp() with RE_NOSUB. Therefore, REG_NOSUB no
longer causes PCRE2_NO_AUTO_CAPTURE to be set, though it still ignores nmatch
and pmatch when regexec() is called.

7. Because of 6 above, pcre2test has been modified with a new modifier called
posix_nosub, to call regcomp() with REG_NOSUB. Previously the no_auto_capture
modifier had this effect. That option is now ignored when the POSIX API is in
use.

8. Minor tidies to the pcre2demo.c sample program, including more comments
about its 8-bit-ness.

9. Detect unmatched closing parentheses and give the error in the pre-scan
instead of later. Previously the pre-scan carried on and could give a
misleading incorrect error message. For example, /(?J)(?'a'))(?'a')/ gave a
message about invalid duplicate group names.

10. It has happened that pcre2test was accidentally linked with another POSIX
regex library instead of libpcre2-posix. In this situation, a call to regcomp()
(in the other library) may succeed, returning zero, but of course putting its
own data into the regex_t block. In one example the re_pcre2_code field was
left as NULL, which made pcre2test think it had not got a compiled POSIX regex,
so it treated the next line as another pattern line, resulting in a confusing
error message. A check has been added to pcre2test to see if the data returned
from a successful call of regcomp() are valid for PCRE2's regcomp(). If they
are not, an error message is output and the pcre2test run is abandoned. The
message points out the possibility of a mis-linking. Hopefully this will avoid
some head-scratching the next time this happens.

11. A pattern such as /(?<=((?C)0))/, which has a callout inside a lookbehind
assertion, caused pcre2test to output a very large number of spaces when the
callout was taken, making the program appearing to loop.

12. A pattern that included (*ACCEPT) in the middle of a sufficiently deeply
nested set of parentheses of sufficient size caused an overflow of the
compiling workspace (which was diagnosed, but of course is not desirable).

13. Detect missing closing parentheses during the pre-pass for group
identification.

14. Changed some integer variable types and put in a number of casts, following
a report of compiler warnings from Visual Studio 2013 and a few tests with
gcc's -Wconversion (which still throws up a lot).

15. Implemented pcre2_code_copy(), and added pushcopy and #popcopy to pcre2test
for testing it.

16. Change 66 for 10.21 introduced the use of snprintf() in PCRE2's version of
regerror(). When the error buffer is too small, my version of snprintf() puts a
binary zero in the final byte. Bug #1801 seems to show that other versions do
not do this, leading to bad output from pcre2test when it was checking for
buffer overflow. It no longer assumes a binary zero at the end of a too-small
regerror() buffer.

17. Fixed typo ("&&" for "&") in pcre2_study(). Fortunately, this could not
actually affect anything, by sheer luck.

18. Two minor fixes for MSVC compilation: (a) removal of apparently incorrect
"const" qualifiers in pcre2test and (b) defining snprintf as _snprintf for
older MSVC compilers. This has been done both in src/pcre2_internal.h for most
of the library, and also in src/pcre2posix.c, which no longer includes
pcre2_internal.h (see 24 below).

19. Applied Chris Wilson's patch (Bugzilla #1681) to CMakeLists.txt for MSVC
static compilation. Subsequently applied Chris Wilson's second patch, putting
the first patch under a new option instead of being unconditional when
PCRE_STATIC is set.

20. Updated pcre2grep to set stdout as binary when run under Windows, so as not
to convert \r\n at the ends of reflected lines into \r\r\n. This required
ensuring that other output that is written to stdout (e.g. file names) uses the
appropriate line terminator: \r\n for Windows, \n otherwise.

21. When a line is too long for pcre2grep's internal buffer, show the maximum
length in the error message.

22. Added support for string callouts to pcre2grep (Zoltan's patch with PH
additions).

23. RunTest.bat was missing a "set type" line for test 22.

24. The pcre2posix.c file was including pcre2_internal.h, and using some
"private" knowledge of the data structures. This is unnecessary; the code has
been re-factored and no longer includes pcre2_internal.h.

25. A racing condition is fixed in JIT reported by Mozilla.

26. Minor code refactor to avoid "array subscript is below array bounds"
compiler warning.

27. Minor code refactor to avoid "left shift of negative number" warning.

28. Add a bit more sanity checking to pcre2_serialize_decode() and document
that it expects trusted data.

29. Fix typo in pcre2_jit_test.c

30. Due to an oversight, pcre2grep was not making use of JIT when available.
This is now fixed.

31. The RunGrepTest script is updated to use the valgrind suppressions file
when testing with JIT under valgrind (compare 10.21/51 below). The suppressions
file is updated so that is now the same as for PCRE1: it suppresses the
Memcheck warnings Addr16 and Cond in unknown objects (that is, JIT-compiled
code). Also changed smc-check=all to smc-check=all-non-file as was done for
RunTest (see 4 above).

32. Implemented the PCRE2_NO_JIT option for pcre2_match().

33. Fix typo that gave a compiler error when JIT not supported.

34. Fix comment describing the returns from find_fixedlength().

35. Fix potential negative index in pcre2test.

36. Calls to pcre2_get_error_message() with error numbers that are never
returned by PCRE2 functions were returning empty strings. Now the error code
PCRE2_ERROR_BADDATA is returned. A facility has been added to pcre2test to
show the texts for given error numbers (i.e. to call pcre2_get_error_message()
and display what it returns) and a few representative error codes are now
checked in RunTest.

37. Added "&& !defined(__INTEL_COMPILER)" to the test for __GNUC__ in
pcre2_match.c, in anticipation that this is needed for the same reason it was
recently added to pcrecpp.cc in PCRE1.

38. Using -o with -M in pcre2grep could cause unnecessary repeated output when
the match extended over a line boundary, as it tried to find more matches "on
the same line" - but it was already over the end.

39. Allow \C in lookbehinds and DFA matching in UTF-32 mode (by converting it
to the same code as '.' when PCRE2_DOTALL is set).

40. Fix two clang compiler warnings in pcre2test when only one code unit width
is supported.

41. Upgrade RunTest to automatically re-run test 2 with a large (64M) stack if
it fails when running the interpreter with a 16M stack (and if changing the
stack size via pcre2test is possible). This avoids having to manually set a
large stack size when testing with clang.

42. Fix register overwite in JIT when SSE2 acceleration is enabled.

43. Detect integer overflow in pcre2test pattern and data repetition counts.

44. In pcre2test, ignore "allcaptures" after DFA matching.

45. Fix unaligned accesses on x86. Patch by Marc Mutz.

46. Fix some more clang compiler warnings.


Version 10.21 12-January-2016
-----------------------------

1. Improve matching speed of patterns starting with + or * in JIT.

2. Use memchr() to find the first character in an unanchored match in 8-bit
mode in the interpreter. This gives a significant speed improvement.

3. Removed a redundant copy of the opcode_possessify table in the
pcre2_auto_possessify.c source.

4. Fix typos in dftables.c for z/OS.

5. Change 36 for 10.20 broke the handling of [[:>:]] and [[:<:]] in that
processing them could involve a buffer overflow if the following character was
an opening parenthesis.

6. Change 36 for 10.20 also introduced a bug in processing this pattern:
/((?x)(*:0))#(?'/. Specifically: if a setting of (?x) was followed by a (*MARK)
setting (which (*:0) is), then (?x) did not get unset at the end of its group
during the scan for named groups, and hence the external # was incorrectly
treated as a comment and the invalid (?' at the end of the pattern was not
diagnosed. This caused a buffer overflow during the real compile. This bug was
discovered by Karl Skomski with the LLVM fuzzer.

7. Moved the pcre2_find_bracket() function from src/pcre2_compile.c into its
own source module to avoid a circular dependency between src/pcre2_compile.c
and src/pcre2_study.c

8. A callout with a string argument containing an opening square bracket, for
example /(?C$[$)(?<]/, was incorrectly processed and could provoke a buffer
overflow. This bug was discovered by Karl Skomski with the LLVM fuzzer.

9. The handling of callouts during the pre-pass for named group identification
has been tightened up.

10. The quantifier {1} can be ignored, whether greedy, non-greedy, or
possessive. This is a very minor optimization.

11. A possessively repeated conditional group that could match an empty string,
for example, /(?(R))*+/, was incorrectly compiled.

12. The Unicode tables have been updated to Unicode 8.0.0 (thanks to Christian
Persch).

13. An empty comment (?#) in a pattern was incorrectly processed and could
provoke a buffer overflow. This bug was discovered by Karl Skomski with the
LLVM fuzzer.

14. Fix infinite recursion in the JIT compiler when certain patterns such as
/(?:|a|){100}x/ are analysed.

15. Some patterns with character classes involving [: and \\ were incorrectly
compiled and could cause reading from uninitialized memory or an incorrect
error diagnosis. Examples are: /[[:\\](?<[::]/ and /[[:\\](?'abc')[a:]. The
first of these bugs was discovered by Karl Skomski with the LLVM fuzzer.

16. Pathological patterns containing many nested occurrences of [: caused
pcre2_compile() to run for a very long time. This bug was found by the LLVM
fuzzer.

17. A missing closing parenthesis for a callout with a string argument was not
being diagnosed, possibly leading to a buffer overflow. This bug was found by
the LLVM fuzzer.

18. A conditional group with only one branch has an implicit empty alternative
branch and must therefore be treated as potentially matching an empty string.

19. If (?R was followed by - or + incorrect behaviour happened instead of a
diagnostic. This bug was discovered by Karl Skomski with the LLVM fuzzer.

20. Another bug that was introduced by change 36 for 10.20: conditional groups
whose condition was an assertion preceded by an explicit callout with a string
argument might be incorrectly processed, especially if the string contained \Q.
This bug was discovered by Karl Skomski with the LLVM fuzzer.

21. Compiling PCRE2 with the sanitize options of clang showed up a number of
very pedantic coding infelicities and a buffer overflow while checking a UTF-8
string if the final multi-byte UTF-8 character was truncated.

22. For Perl compatibility in EBCDIC environments, ranges such as a-z in a
class, where both values are literal letters in the same case, omit the
non-letter EBCDIC code points within the range.

23. Finding the minimum matching length of complex patterns with back
references and/or recursions can take a long time. There is now a cut-off that
gives up trying to find a minimum length when things get too complex.

24. An optimization has been added that speeds up finding the minimum matching
length for patterns containing repeated capturing groups or recursions.

25. If a pattern contained a back reference to a group whose number was
duplicated as a result of appearing in a (?|...) group, the computation of the
minimum matching length gave a wrong result, which could cause incorrect "no
match" errors. For such patterns, a minimum matching length cannot at present
be computed.

26. Added a check for integer overflow in conditions (?(<digits>) and
(?(R<digits>). This omission was discovered by Karl Skomski with the LLVM
fuzzer.

27. Fixed an issue when \p{Any} inside an xclass did not read the current
character.

28. If pcre2grep was given the -q option with -c or -l, or when handling a
binary file, it incorrectly wrote output to stdout.

29. The JIT compiler did not restore the control verb head in case of *THEN
control verbs. This issue was found by Karl Skomski with a custom LLVM fuzzer.

30. The way recursive references such as (?3) are compiled has been re-written
because the old way was the cause of many issues. Now, conversion of the group
number into a pattern offset does not happen until the pattern has been
completely compiled. This does mean that detection of all infinitely looping
recursions is postponed till match time. In the past, some easy ones were
detected at compile time. This re-writing was done in response to yet another
bug found by the LLVM fuzzer.

31. A test for a back reference to a non-existent group was missing for items
such as \987. This caused incorrect code to be compiled. This issue was found
by Karl Skomski with a custom LLVM fuzzer.

32. Error messages for syntax errors following \g and \k were giving inaccurate
offsets in the pattern.

33. Improve the performance of starting single character repetitions in JIT.

34. (*LIMIT_MATCH=) now gives an error instead of setting the value to 0.

35. Error messages for syntax errors in *LIMIT_MATCH and *LIMIT_RECURSION now
give the right offset instead of zero.

36. The JIT compiler should not check repeats after a {0,1} repeat byte code.
This issue was found by Karl Skomski with a custom LLVM fuzzer.

37. The JIT compiler should restore the control chain for empty possessive
repeats. This issue was found by Karl Skomski with a custom LLVM fuzzer.

38. A bug which was introduced by the single character repetition optimization
was fixed.

39. Match limit check added to recursion. This issue was found by Karl Skomski
with a custom LLVM fuzzer.

40. Arrange for the UTF check in pcre2_match() and pcre2_dfa_match() to look
only at the part of the subject that is relevant when the starting offset is
non-zero.

41. Improve first character match in JIT with SSE2 on x86.

42. Fix two assertion fails in JIT. These issues were found by Karl Skomski
with a custom LLVM fuzzer.

43. Correct the setting of CMAKE_C_FLAGS in CMakeLists.txt (patch from Roy Ivy
III).

44. Fix bug in RunTest.bat for new test 14, and adjust the script for the added
test (there are now 20 in total).

45. Fixed a corner case of range optimization in JIT.

46. Add the ${*MARK} facility to pcre2_substitute().

47. Modifier lists in pcre2test were splitting at spaces without the required
commas.

48. Implemented PCRE2_ALT_VERBNAMES.

49. Fixed two issues in JIT. These were found by Karl Skomski with a custom
LLVM fuzzer.

50. The pcre2test program has been extended by adding the #newline_default
command. This has made it possible to run the standard tests when PCRE2 is
compiled with either CR or CRLF as the default newline convention. As part of
this work, the new command was added to several test files and the testing
scripts were modified. The pcre2grep tests can now also be run when there is no
LF in the default newline convention.

51. The RunTest script has been modified so that, when JIT is used and valgrind
is specified, a valgrind suppressions file is set up to ignore "Invalid read of
size 16" errors because these are false positives when the hardware supports
the SSE2 instruction set.

52. It is now possible to have comment lines amid the subject strings in
pcre2test (and perltest.sh) input.

53. Implemented PCRE2_USE_OFFSET_LIMIT and pcre2_set_offset_limit().

54. Add the null_context modifier to pcre2test so that calling pcre2_compile()
and the matching functions with NULL contexts can be tested.

55. Implemented PCRE2_SUBSTITUTE_EXTENDED.

56. In a character class such as [\W\p{Any}] where both a negative-type escape
("not a word character") and a property escape were present, the property
escape was being ignored.

57. Fixed integer overflow for patterns whose minimum matching length is very,
very large.

58. Implemented --never-backslash-C.

59. Change 55 above introduced a bug by which certain patterns provoked the
erroneous error "\ at end of pattern".

60. The special sequences [[:<:]] and [[:>:]] gave rise to incorrect compiling
errors or other strange effects if compiled in UCP mode. Found with libFuzzer
and AddressSanitizer.

61. Whitespace at the end of a pcre2test pattern line caused a spurious error
message if there were only single-character modifiers. It should be ignored.

62. The use of PCRE2_NO_AUTO_CAPTURE could cause incorrect compilation results
or segmentation errors for some patterns. Found with libFuzzer and
AddressSanitizer.

63. Very long names in (*MARK) or (*THEN) etc. items could provoke a buffer
overflow.

64. Improve error message for overly-complicated patterns.

65. Implemented an optional replication feature for patterns in pcre2test, to
make it easier to test long repetitive patterns. The tests for 63 above are
converted to use the new feature.

66. In the POSIX wrapper, if regerror() was given too small a buffer, it could
misbehave.

67. In pcre2_substitute() in UTF mode, the UTF validity check on the
replacement string was happening before the length setting when the replacement
string was zero-terminated.

68. In pcre2_substitute() in UTF mode, PCRE2_NO_UTF_CHECK can be set for the
second and subsequent calls to pcre2_match().

69. There was no check for integer overflow for a replacement group number in
pcre2_substitute(). An added check for a number greater than the largest group
number in the pattern means this is not now needed.

70. The PCRE2-specific VERSION condition didn't work correctly if only one
digit was given after the decimal point, or if more than two digits were given.
It now works with one or two digits, and gives a compile time error if more are
given.

71. In pcre2_substitute() there was the possibility of reading one code unit
beyond the end of the replacement string.

72. The code for checking a subject's UTF-32 validity for a pattern with a
lookbehind involved an out-of-bounds pointer, which could potentially cause
trouble in some environments.

73. The maximum lookbehind length was incorrectly calculated for patterns such
as /(?<=(a)(?-1))x/ which have a recursion within a backreference.

74. Give an error if a lookbehind assertion is longer than 65535 code units.

75. Give an error in pcre2_substitute() if a match ends before it starts (as a
result of the use of \K).

76. Check the length of subpattern names and the names in (*MARK:xx) etc.
dynamically to avoid the possibility of integer overflow.

77. Implement pcre2_set_max_pattern_length() so that programs can restrict the
size of patterns that they are prepared to handle.

78. (*NO_AUTO_POSSESS) was not working.

79. Adding group information caching improves the speed of compiling when
checking whether a group has a fixed length and/or could match an empty string,
especially when recursion or subroutine calls are involved. However, this
cannot be used when (?| is present in the pattern because the same number may
be used for groups of different sizes. To catch runaway patterns in this
situation, counts have been introduced to the functions that scan for empty
branches or compute fixed lengths.

80. Allow for the possibility of the size of the nest_save structure not being
a factor of the size of the compiling workspace (it currently is).

81. Check for integer overflow in minimum length calculation and cap it at
65535.

82. Small optimizations in code for finding the minimum matching length.

83. Lock out configuring for EBCDIC with non-8-bit libraries.

84. Test for error code <= 0 in regerror().

85. Check for too many replacements (more than INT_MAX) in pcre2_substitute().

86. Avoid the possibility of computing with an out-of-bounds pointer (though
not dereferencing it) while handling lookbehind assertions.

87. Failure to get memory for the match data in regcomp() is now given as a
regcomp() error instead of waiting for regexec() to pick it up.

88. In pcre2_substitute(), ensure that CRLF is not split when it is a valid
newline sequence.

89. Paranoid check in regcomp() for bad error code from pcre2_compile().

90. Run test 8 (internal offsets and code sizes) for link sizes 3 and 4 as well
as for link size 2.

91. Document that JIT has a limit on pattern size, and give more information
about JIT compile failures in pcre2test.

92. Implement PCRE2_INFO_HASBACKSLASHC.

93. Re-arrange valgrind support code in pcre2test to avoid spurious reports
with JIT (possibly caused by SSE2?).

94. Support offset_limit in JIT.

95. A sequence such as [[:punct:]b] that is, a POSIX character class followed
by a single ASCII character in a class item, was incorrectly compiled in UCP
mode. The POSIX class got lost, but only if the single character followed it.

96. [:punct:] in UCP mode was matching some characters in the range 128-255
that should not have been matched.

97. If [:^ascii:] or [:^xdigit:] are present in a non-negated class, all
characters with code points greater than 255 are in the class. When a Unicode
property was also in the class (if PCRE2_UCP is set, escapes such as \w are
turned into Unicode properties), wide characters were not correctly handled,
and could fail to match.

98. In pcre2test, make the "startoffset" modifier a synonym of "offset",
because it sets the "startoffset" parameter for pcre2_match().

99. If PCRE2_AUTO_CALLOUT was set on a pattern that had a (?# comment between
an item and its qualifier (for example, A(?#comment)?B) pcre2_compile()
misbehaved. This bug was found by the LLVM fuzzer.

100. The error for an invalid UTF pattern string always gave the code unit
offset as zero instead of where the invalidity was found.

101. Further to 97 above, negated classes such as [^[:^ascii:]\d] were also not
working correctly in UCP mode.

102. Similar to 99 above, if an isolated \E was present between an item and its
qualifier when PCRE2_AUTO_CALLOUT was set, pcre2_compile() misbehaved. This bug
was found by the LLVM fuzzer.

103. The POSIX wrapper function regexec() crashed if the option REG_STARTEND
was set when the pmatch argument was NULL. It now returns REG_INVARG.

104. Allow for up to 32-bit numbers in the ordin() function in pcre2grep.

105. An empty \Q\E sequence between an item and its qualifier caused
pcre2_compile() to misbehave when auto callouts were enabled. This bug
was found by the LLVM fuzzer.

106. If both PCRE2_ALT_VERBNAMES and PCRE2_EXTENDED were set, and a (*MARK) or
other verb "name" ended with whitespace immediately before the closing
parenthesis, pcre2_compile() misbehaved. Example: /(*:abc )/, but only when
both those options were set.

107. In a number of places pcre2_compile() was not handling NULL characters
correctly, and pcre2test with the "bincode" modifier was not always correctly
displaying fields containing NULLS:

   (a) Within /x extended #-comments
   (b) Within the "name" part of (*MARK) and other *verbs
   (c) Within the text argument of a callout

108. If a pattern that was compiled with PCRE2_EXTENDED started with white
space or a #-type comment that was followed by (?-x), which turns off
PCRE2_EXTENDED, and there was no subsequent (?x) to turn it on again,
pcre2_compile() assumed that (?-x) applied to the whole pattern and
consequently mis-compiled it. This bug was found by the LLVM fuzzer. The fix
for this bug means that a setting of any of the (?imsxJU) options at the start
of a pattern is no longer transferred to the options that are returned by
PCRE2_INFO_ALLOPTIONS. In fact, this was an anachronism that should have
changed when the effects of those options were all moved to compile time.

109. An escaped closing parenthesis in the "name" part of a (*verb) when
PCRE2_ALT_VERBNAMES was set caused pcre2_compile() to malfunction. This bug
was found by the LLVM fuzzer.

110. Implemented PCRE2_SUBSTITUTE_UNSET_EMPTY, and updated pcre2test to make it
possible to test it.

111. "Harden" pcre2test against ridiculously large values in modifiers and
command line arguments.

112. Implemented PCRE2_SUBSTITUTE_UNKNOWN_UNSET and PCRE2_SUBSTITUTE_OVERFLOW_
LENGTH.

113. Fix printing of *MARK names that contain binary zeroes in pcre2test.


Version 10.20 30-June-2015
--------------------------

1. Callouts with string arguments have been added.

2. Assertion code generator in JIT has been optimized.

3. The invalid pattern (?(?C) has a missing assertion condition at the end. The
pcre2_compile() function read past the end of the input before diagnosing an
error. This bug was discovered by the LLVM fuzzer.

4. Implemented pcre2_callout_enumerate().

5. Fix JIT compilation of conditional blocks whose assertion is converted to
(*FAIL). E.g: /(?(?!))/.

6. The pattern /(?(?!)^)/ caused references to random memory. This bug was
discovered by the LLVM fuzzer.

7. The assertion (?!) is optimized to (*FAIL). This was not handled correctly
when this assertion was used as a condition, for example (?(?!)a|b). In
pcre2_match() it worked by luck; in pcre2_dfa_match() it gave an incorrect
error about an unsupported item.

8. For some types of pattern, for example /Z*(|d*){216}/, the auto-
possessification code could take exponential time to complete. A recursion
depth limit of 1000 has been imposed to limit the resources used by this
optimization. This infelicity was discovered by the LLVM fuzzer.

9. A pattern such as /(*UTF)[\S\V\H]/, which contains a negated special class
such as \S in non-UCP mode, explicit wide characters (> 255) can be ignored
because \S ensures they are all in the class. The code for doing this was
interacting badly with the code for computing the amount of space needed to
compile the pattern, leading to a buffer overflow. This bug was discovered by
the LLVM fuzzer.

10. A pattern such as /((?2)+)((?1))/ which has mutual recursion nested inside
other kinds of group caused stack overflow at compile time. This bug was
discovered by the LLVM fuzzer.

11. A pattern such as /(?1)(?#?'){8}(a)/ which had a parenthesized comment
between a subroutine call and its quantifier was incorrectly compiled, leading
to buffer overflow or other errors. This bug was discovered by the LLVM fuzzer.

12. The illegal pattern /(?(?<E>.*!.*)?)/ was not being diagnosed as missing an
assertion after (?(. The code was failing to check the character after (?(?<
for the ! or = that would indicate a lookbehind assertion. This bug was
discovered by the LLVM fuzzer.

13. A pattern such as /X((?2)()*+){2}+/ which has a possessive quantifier with
a fixed maximum following a group that contains a subroutine reference was
incorrectly compiled and could trigger buffer overflow. This bug was discovered
by the LLVM fuzzer.

14. Negative relative recursive references such as (?-7) to non-existent
subpatterns were not being diagnosed and could lead to unpredictable behaviour.
This bug was discovered by the LLVM fuzzer.

15. The bug fixed in 14 was due to an integer variable that was unsigned when
it should have been signed. Some other "int" variables, having been checked,
have either been changed to uint32_t or commented as "must be signed".

16. A mutual recursion within a lookbehind assertion such as (?<=((?2))((?1)))
caused a stack overflow instead of the diagnosis of a non-fixed length
lookbehind assertion. This bug was discovered by the LLVM fuzzer.

17. The use of \K in a positive lookbehind assertion in a non-anchored pattern
(e.g. /(?<=\Ka)/) could make pcre2grep loop.

18. There was a similar problem to 17 in pcre2test for global matches, though
the code there did catch the loop.

19. If a greedy quantified \X was preceded by \C in UTF mode (e.g. \C\X*),
and a subsequent item in the pattern caused a non-match, backtracking over the
repeated \X did not stop, but carried on past the start of the subject, causing
reference to random memory and/or a segfault. There were also some other cases
where backtracking after \C could crash. This set of bugs was discovered by the
LLVM fuzzer.

20. The function for finding the minimum length of a matching string could take
a very long time if mutual recursion was present many times in a pattern, for
example, /((?2){73}(?2))((?1))/. A better mutual recursion detection method has
been implemented. This infelicity was discovered by the LLVM fuzzer.

21. Implemented PCRE2_NEVER_BACKSLASH_C.

22. The feature for string replication in pcre2test could read from freed
memory if the replication required a buffer to be extended, and it was not
working properly in 16-bit and 32-bit modes. This issue was discovered by a
fuzzer: see http://lcamtuf.coredump.cx/afl/.

23. Added the PCRE2_ALT_CIRCUMFLEX option.

24. Adjust the treatment of \8 and \9 to be the same as the current Perl
behaviour.

25. Static linking against the PCRE2 library using the pkg-config module was
failing on missing pthread symbols.

26. If a group that contained a recursive back reference also contained a
forward reference subroutine call followed by a non-forward-reference
subroutine call, for example /.((?2)(?R)\1)()/, pcre2_compile() failed to
compile correct code, leading to undefined behaviour or an internally detected
error. This bug was discovered by the LLVM fuzzer.

27. Quantification of certain items (e.g. atomic back references) could cause
incorrect code to be compiled when recursive forward references were involved.
For example, in this pattern: /(?1)()((((((\1++))\x85)+)|))/. This bug was
discovered by the LLVM fuzzer.

28. A repeated conditional group whose condition was a reference by name caused
a buffer overflow if there was more than one group with the given name. This
bug was discovered by the LLVM fuzzer.

29. A recursive back reference by name within a group that had the same name as
another group caused a buffer overflow. For example: /(?J)(?'d'(?'d'\g{d}))/.
This bug was discovered by the LLVM fuzzer.

30. A forward reference by name to a group whose number is the same as the
current group, for example in this pattern: /(?|(\k'Pm')|(?'Pm'))/, caused a
buffer overflow at compile time. This bug was discovered by the LLVM fuzzer.

31. Fix -fsanitize=undefined warnings for left shifts of 1 by 31 (it treats 1
as an int; fixed by writing it as 1u).

32. Fix pcre2grep compile when -std=c99 is used with gcc, though it still gives
a warning for "fileno" unless -std=gnu99 us used.

33. A lookbehind assertion within a set of mutually recursive subpatterns could
provoke a buffer overflow. This bug was discovered by the LLVM fuzzer.

34. Give an error for an empty subpattern name such as (?'').

35. Make pcre2test give an error if a pattern that follows #forbud_utf contains
\P, \p, or \X.

36. The way named subpatterns are handled has been refactored. There is now a
pre-pass over the regex which does nothing other than identify named
subpatterns and count the total captures. This means that information about
named patterns is known before the rest of the compile. In particular, it means
that forward references can be checked as they are encountered. Previously, the
code for handling forward references was contorted and led to several errors in
computing the memory requirements for some patterns, leading to buffer
overflows.

37. There was no check for integer overflow in subroutine calls such as (?123).

38. The table entry for \l in EBCDIC environments was incorrect, leading to its
being treated as a literal 'l' instead of causing an error.

39. If a non-capturing group containing a conditional group that could match
an empty string was repeated, it was not identified as matching an empty string
itself. For example: /^(?:(?(1)x|)+)+$()/.

40. In an EBCDIC environment, pcretest was mishandling the escape sequences
\a and \e in test subject lines.

41. In an EBCDIC environment, \a in a pattern was converted to the ASCII
instead of the EBCDIC value.

42. The handling of \c in an EBCDIC environment has been revised so that it is
now compatible with the specification in Perl's perlebcdic page.

43. Single character repetition in JIT has been improved. 20-30% speedup
was achieved on certain patterns.

44. The EBCDIC character 0x41 is a non-breaking space, equivalent to 0xa0 in
ASCII/Unicode. This has now been added to the list of characters that are
recognized as white space in EBCDIC.

45. When PCRE2 was compiled without Unicode support, the use of \p and \P gave
an error (correctly) when used outside a class, but did not give an error
within a class.

46. \h within a class was incorrectly compiled in EBCDIC environments.

47. JIT should return with error when the compiled pattern requires
more stack space than the maximum.

48. Fixed a memory leak in pcre2grep when a locale is set.


Version 10.10 06-March-2015
---------------------------

1. When a pattern is compiled, it remembers the highest back reference so that
when matching, if the ovector is too small, extra memory can be obtained to
use instead. A conditional subpattern whose condition is a check on a capture
having happened, such as, for example in the pattern /^(?:(a)|b)(?(1)A|B)/, is
another kind of back reference, but it was not setting the highest
backreference number. This mattered only if pcre2_match() was called with an
ovector that was too small to hold the capture, and there was no other kind of
back reference (a situation which is probably quite rare). The effect of the
bug was that the condition was always treated as FALSE when the capture could
not be consulted, leading to a incorrect behaviour by pcre2_match(). This bug
has been fixed.

2. Functions for serialization and deserialization of sets of compiled patterns
have been added.

3. The value that is returned by PCRE2_INFO_SIZE has been corrected to remove
excess code units at the end of the data block that may occasionally occur if
the code for calculating the size over-estimates. This change stops the
serialization code copying uninitialized data, to which valgrind objects. The
documentation of PCRE2_INFO_SIZE was incorrect in stating that the size did not
include the general overhead. This has been corrected.

4. All code units in every slot in the table of group names are now set, again
in order to avoid accessing uninitialized data when serializing.

5. The (*NO_JIT) feature is implemented.

6. If a bug that caused pcre2_compile() to use more memory than allocated was
triggered when using valgrind, the code in (3) above passed a stupidly large
value to valgrind. This caused a crash instead of an "internal error" return.

7. A reference to a duplicated named group (either a back reference or a test
for being set in a conditional) that occurred in a part of the pattern where
PCRE2_DUPNAMES was not set caused the amount of memory needed for the pattern
to be incorrectly calculated, leading to overwriting.

8. A mutually recursive set of back references such as (\2)(\1) caused a
segfault at compile time (while trying to find the minimum matching length).
The infinite loop is now broken (with the minimum length unset, that is, zero).

9. If an assertion that was used as a condition was quantified with a minimum
of zero, matching went wrong. In particular, if the whole group had unlimited
repetition and could match an empty string, a segfault was likely. The pattern
(?(?=0)?)+ is an example that caused this. Perl allows assertions to be
quantified, but not if they are being used as conditions, so the above pattern
is faulted by Perl. PCRE2 has now been changed so that it also rejects such
patterns.

10. The error message for an invalid quantifier has been changed from "nothing
to repeat" to "quantifier does not follow a repeatable item".

11. If a bad UTF string is compiled with NO_UTF_CHECK, it may succeed, but
scanning the compiled pattern in subsequent auto-possessification can get out
of step and lead to an unknown opcode. Previously this could have caused an
infinite loop. Now it generates an "internal error" error. This is a tidyup,
not a bug fix; passing bad UTF with NO_UTF_CHECK is documented as having an
undefined outcome.

12. A UTF pattern containing a "not" match of a non-ASCII character and a
subroutine reference could loop at compile time. Example: /[^\xff]((?1))/.

13. The locale test (RunTest 3) has been upgraded. It now checks that a locale
that is found in the output of "locale -a" can actually be set by pcre2test
before it is accepted. Previously, in an environment where a locale was listed
but would not set (an example does exist), the test would "pass" without
actually doing anything. Also the fr_CA locale has been added to the list of
locales that can be used.

14. Fixed a bug in pcre2_substitute(). If a replacement string ended in a
capturing group number without parentheses, the last character was incorrectly
literally included at the end of the replacement string.

15. A possessive capturing group such as (a)*+ with a minimum repeat of zero
failed to allow the zero-repeat case if pcre2_match() was called with an
ovector too small to capture the group.

16. Improved error message in pcre2test when setting the stack size (-S) fails.

17. Fixed two bugs in CMakeLists.txt: (1) Some lines had got lost in the
transfer from PCRE1, meaning that CMake configuration failed if "build tests"
was selected. (2) The file src/pcre2_serialize.c had not been added to the list
of PCRE2 sources, which caused a failure to build pcre2test.

18. Fixed typo in pcre2_serialize.c (DECL instead of DEFN) that causes problems
only on Windows.

19. Use binary input when reading back saved serialized patterns in pcre2test.

20. Added RunTest.bat for running the tests under Windows.

21. "make distclean" was not removing config.h, a file that may be created for
use with CMake.

22. A pattern such as "((?2){0,1999}())?", which has a group containing a
forward reference repeated a large (but limited) number of times within a
repeated outer group that has a zero minimum quantifier, caused incorrect code
to be compiled, leading to the error "internal error: previously-checked
referenced subpattern not found" when an incorrect memory address was read.
This bug was reported as "heap overflow", discovered by Kai Lu of Fortinet's
FortiGuard Labs. (Added 24-March-2015: CVE-2015-2325 was given to this.)

23. A pattern such as "((?+1)(\1))/" containing a forward reference subroutine
call within a group that also contained a recursive back reference caused
incorrect code to be compiled. This bug was reported as "heap overflow",
discovered by Kai Lu of Fortinet's FortiGuard Labs. (Added 24-March-2015:
CVE-2015-2326 was given to this.)

24. Computing the size of the JIT read-only data in advance has been a source
of various issues, and new ones are still appear unfortunately. To fix
existing and future issues, size computation is eliminated from the code,
and replaced by on-demand memory allocation.

25. A pattern such as /(?i)[A-`]/, where characters in the other case are
adjacent to the end of the range, and the range contained characters with more
than one other case, caused incorrect behaviour when compiled in UTF mode. In
that example, the range a-j was left out of the class.


Version 10.00 05-January-2015
-----------------------------

Version 10.00 is the first release of PCRE2, a revised API for the PCRE
library. Changes prior to 10.00 are logged in the ChangeLog file for the old
API, up to item 20 for release 8.36.

The code of the library was heavily revised as part of the new API
implementation. Details of each and every modification were not individually
logged. In addition to the API changes, the following changes were made. They
are either new functionality, or bug fixes and other noticeable changes of
behaviour that were implemented after the code had been forked.

1. Including Unicode support at build time is now enabled by default, but it
can optionally be disabled. It is not enabled by default at run time (no
change).

2. The test program, now called pcre2test, was re-specified and almost
completely re-written. Its input is not compatible with input for pcretest.

3. Patterns may start with (*NOTEMPTY) or (*NOTEMPTY_ATSTART) to set the
PCRE2_NOTEMPTY or PCRE2_NOTEMPTY_ATSTART options for every subject line that is
matched by that pattern.

4. For the benefit of those who use PCRE2 via some other application, that is,
not writing the function calls themselves, it is possible to check the PCRE2
version by matching a pattern such as /(?(VERSION>=10)yes|no)/ against a
string such as "yesno".

5. There are case-equivalent Unicode characters whose encodings use different
numbers of code units in UTF-8. U+023A and U+2C65 are one example. (It is
theoretically possible for this to happen in UTF-16 too.) If a backreference to
a group containing one of these characters was greedily repeated, and during
the match a backtrack occurred, the subject might be backtracked by the wrong
number of code units. For example, if /^(\x{23a})\1*(.)/ is matched caselessly
(and in UTF-8 mode) against "\x{23a}\x{2c65}\x{2c65}\x{2c65}", group 2 should
capture the final character, which is the three bytes E2, B1, and A5 in UTF-8.
Incorrect backtracking meant that group 2 captured only the last two bytes.
This bug has been fixed; the new code is slower, but it is used only when the
strings matched by the repetition are not all the same length.

6. A pattern such as /()a/ was not setting the "first character must be 'a'"
information. This applied to any pattern with a group that matched no
characters, for example: /(?:(?=.)|(?<!x))a/.

7. When an (*ACCEPT) is triggered inside capturing parentheses, it arranges for
those parentheses to be closed with whatever has been captured so far. However,
it was failing to mark any other groups between the highest capture so far and
the currrent group as "unset". Thus, the ovector for those groups contained
whatever was previously there. An example is the pattern /(x)|((*ACCEPT))/ when
matched against "abcd".

8. The pcre2_substitute() function has been implemented.

9. If an assertion used as a condition was quantified with a minimum of zero
(an odd thing to do, but it happened), SIGSEGV or other misbehaviour could
occur.

10. The PCRE2_NO_DOTSTAR_ANCHOR option has been implemented.

****