23 months agoimproved optree generation improved_charclass_generator
Yves Orton [Wed, 3 Oct 2012 17:05:03 +0000]
improved optree generation

Karl noticed that we dont always deal with common suffixes in the most
efficient way. This change reworks how we convert a trie to an optree so
that common suffixes are always grouped together.

Needs furter review by Karl.

23 months agoadd comments and some minor code cleanup
Yves Orton [Wed, 3 Oct 2012 01:40:50 +0000]
add comments and some minor code cleanup

23 months agoFix [perl #115050] Double empty sub-regexp makes "panic!"
Yves Orton [Sun, 30 Sep 2012 23:01:26 +0000]
Fix [perl #115050] Double empty sub-regexp makes "panic!"

Code failed to check if we had a non-zero (END) trietype, and
thus under certain circumstances would panic. We now check that
the trietype is non-zero.

23 months agoSuggest cause of error requiring .pm file.
Paul Johnson [Wed, 26 Sep 2012 00:44:45 +0000]
Suggest cause of error requiring .pm file.

Following on from a recent thread I've put together a patch to expand
the error message when a module can't be loaded. With this patch,
instead of:

Can't locate Stuff/Of/Dreams.pm in @INC (@INC contains: ...)

You get:

Can't locate Stuff/Of/Dreams.pm in @INC (you may need to install the Stuff::Of::Dreams module) (@INC contains: ...)

[The committer tweaked the error message,
 based on a suggestion by Tony Cook.  See

23 months agoIncrease $warnings::VERSION to 1.15
Father Chrysostomos [Sun, 30 Sep 2012 05:43:38 +0000]
Increase $warnings::VERSION to 1.15

23 months agoIncrease $feature::VERSION to 1.31
Father Chrysostomos [Sun, 30 Sep 2012 05:43:09 +0000]
Increase $feature::VERSION to 1.31

23 months agoUse two colons for lexsub warning
Father Chrysostomos [Sun, 30 Sep 2012 05:39:37 +0000]
Use two colons for lexsub warning

23 months agoMerge branch 'rt_115078' into blead
Yves Orton [Sat, 29 Sep 2012 22:00:25 +0000]
Merge branch 'rt_115078' into blead

This is a no-op merge which fixes the problem reported in perl #115078
and at the same time includes some modest improvements in the code
generated in regcharclass.h.

23 months agoremove test define from regen/regcharclass.pl
Yves Orton [Sat, 29 Sep 2012 21:58:03 +0000]
remove test define from regen/regcharclass.pl

23 months agoimprove conditional folding logic in regen/regcharclass.pl
Yves Orton [Sat, 29 Sep 2012 21:57:10 +0000]
improve conditional folding logic in regen/regcharclass.pl

23 months agofix perl #115078, ternary folding logic failure
Yves Orton [Sat, 29 Sep 2012 21:01:25 +0000]
fix perl #115078, ternary folding logic failure

23 months agoadd a new define for testing perl #115078
Yves Orton [Sat, 29 Sep 2012 20:55:13 +0000]
add a new define for testing perl #115078

We dont have any easy way to test regen/regcharclass.pl currently.
Perl #115078 is related to a bug in the _cleanup() routine which is
fixed with next patch.

23 months agoUpdate CPANPLUS to CPAN version 0.9133
Chris 'BinGOs' Williams [Sat, 29 Sep 2012 11:04:06 +0000]
Update CPANPLUS to CPAN version 0.9133


  This a version bump only

  Changes for 0.9133      Sat Sep 29 11:49:02 2012
  * Fix MANIFEST for last release

  Changes for 0.9132      Sat Sep 29 11:40:53 2012
  * Resolve RT #79925, Module::Metadata was missing
    from inc/bundle. Reported by Steve Sanbeg

23 months agoAdd one test for ord('').
James E Keenan [Sat, 29 Sep 2012 00:37:00 +0000]
Add one test for ord('').

"perldoc -f ord" states that if ord()'s argument is an empty string, the
function returns 0.  Demonstrating that this is so.

23 months agoRefactor t/op/chars.t to use test.pl instead of making TAP by hand.
Colin Kuskie [Fri, 28 Sep 2012 04:43:34 +0000]
Refactor t/op/chars.t to use test.pl instead of making TAP by hand.

23 months ago[perl #20636] Make h2xs skip #define macros with empty rhs
Aaron Crane [Fri, 28 Sep 2012 13:14:47 +0000]
[perl #20636] Make h2xs skip #define macros with empty rhs

Otherwise the generated C code uses such macros in expressions, which causes
compilation errors because the macro is expanded to an empty token list.

23 months agoDocument :shared and :unique in attributes.pm
Jerry D. Hedden [Fri, 28 Sep 2012 15:51:56 +0000]
Document :shared and :unique in attributes.pm

Add documentation to attributes.pm for :shared and :unique, and bump version.

23 months agoBump $Carp::Heavy::VERSION too to avoid test failure from 858e29b93a
Steve Hay [Fri, 28 Sep 2012 08:12:07 +0000]
Bump $Carp::Heavy::VERSION too to avoid test failure from 858e29b93a

23 months agoRemove option to build without USE_SOCKETS_AS_HANDLES on Windows
Steve Hay [Fri, 28 Sep 2012 08:02:31 +0000]
Remove option to build without USE_SOCKETS_AS_HANDLES on Windows

The option is always defined by default and can't be disabled from the
makefiles. Manually disabling it causes several tests to fail, which
nobody has reported, so we presume nobody does this. The non-default
configuration is believed to be historical cruft with no value now, and
has clearly bitrotted in recent years (hence the test failures), so
remove it to simplify the codebase slightly.

23 months agoperldelta for ed5044532 and 858e29b9 (Carp)
Tony Cook [Fri, 28 Sep 2012 06:25:05 +0000]
perldelta for ed5044532 and 858e29b9 (Carp)

23 months agobump Carp version to 1.27
Tony Cook [Fri, 28 Sep 2012 06:22:44 +0000]
bump Carp version to 1.27

23 months agofix -DPERL_GLOBAL_STRUCT build failure introduced in 97b03d64 and e10681aa
Tony Cook [Fri, 28 Sep 2012 05:51:40 +0000]
fix -DPERL_GLOBAL_STRUCT build failure introduced in 97b03d64 and e10681aa

23 months agotodo: strict as warnings
Ricardo Signes [Fri, 28 Sep 2012 02:27:18 +0000]
todo: strict as warnings

23 months agoDocument exportable subroutines longmess() and shortmess().
jkeenan [Fri, 27 Jul 2012 01:33:03 +0000]
Document exportable subroutines longmess() and shortmess().

Per recommendation by srezic in RT #114280.

23 months agoClarify rationale for consulting L<feature>.
James E Keenan [Thu, 27 Sep 2012 23:05:15 +0000]
Clarify rationale for consulting L<feature>.

For RT #90178.

23 months agoRevert "Add 5.14.3-RC1 to perlhist"
Dominic Hargreaves [Thu, 27 Sep 2012 17:31:45 +0000]
Revert "Add 5.14.3-RC1 to perlhist"

RC releases shouldn't be in perlhist

This reverts commit bfec0e075afbe1005446f328adda6fefd2dc88ee.

23 months agoperlreguts: Fit long verbatim lines to 79 cols
Karl Williamson [Thu, 27 Sep 2012 03:40:22 +0000]
perlreguts: Fit long verbatim lines to 79 cols

23 months agomktables: Mention USourceData in generated pod
Karl Williamson [Thu, 27 Sep 2012 16:12:41 +0000]
mktables: Mention USourceData in generated pod

These files were included by Unicode for the first time in the final
version of its version 6.2.  They document proposals for encoding
Han characters in Unicode.  As far as I can tell, they have no real use
except to people working on such proposals.  They are considered part of
the Unicode Character Database, however, and should be mentioned in
perluniprops as data that Perl ignores from that database.

23 months agomktables: Nits in comments, generated pod
Karl Williamson [Thu, 27 Sep 2012 16:12:06 +0000]
mktables: Nits in comments, generated pod

23 months agoperlpolicy.pod: clarify module deprecation/removal
David Golden [Thu, 27 Sep 2012 15:08:09 +0000]
perlpolicy.pod: clarify module deprecation/removal

23 months agomove perlrun to a more prominent place in perl.pod
Jesse Luehrs [Thu, 27 Sep 2012 15:03:13 +0000]
move perlrun to a more prominent place in perl.pod

the list of command line options in the synopsis is really not very
useful without knowing what they mean, and listing perlrun way down in
the table of contents doesn't make it very easy to figure out what they

23 months agoUse official Unicode 6.2
Karl Williamson [Thu, 27 Sep 2012 02:41:58 +0000]
Use official Unicode 6.2

Unicode 6.2 has been officially released, and is delivered by this
commit.  There are no substantive changes from the final 6.2 beta, which
this replaces.

23 months agoEliminate now superfluous counter.
James E Keenan [Thu, 27 Sep 2012 00:15:30 +0000]
Eliminate now superfluous counter.

23 months agoRefactor t/op/lex_assign.t to use test.pl, and not make TAP by hand.
Colin Kuskie [Mon, 10 Sep 2012 02:08:55 +0000]
Refactor t/op/lex_assign.t to use test.pl, and not make TAP by hand.

23 months agoAdd 5.14.3-RC1 to perlhist
Dominic Hargreaves [Wed, 26 Sep 2012 22:48:48 +0000]
Add 5.14.3-RC1 to perlhist

23 months agoFlush PL_stashcache when assigning a file handle to a typeglob.
Nicholas Clark [Sun, 23 Sep 2012 20:48:42 +0000]
Flush PL_stashcache when assigning a file handle to a typeglob.

File handles take priority over stashes for method dispatch. Assigning a
file handle to a typeglob potentially creates a file handle where one did
not exist before. As PL_stashcache only contains entries for names which
unambiguously resolve to stashes, such a change may mean that PL_stashcache
now contains an invalid entry. As it's hard to work out exactly which entries
might be affected, simply flush the entire cache and let it rebuild itself.

23 months agoRestore use of PL_stashcache, the stash name lookup cache for method calls.
Nicholas Clark [Sun, 23 Sep 2012 20:21:14 +0000]
Restore use of PL_stashcache, the stash name lookup cache for method calls.

Commit da6b625f78f5f133 in Aug 2011 inadvertently broke the code that looks
up values in PL_stashcache. As it's a only cache, quite correctly everything
carried on working without it.

Restoring it re-introduces two bugs first introduced when PL_stashcache was
added, by commit 081fc587427bbcef in Apr 2003.

23 months ago-Do now also reports updates and use of PL_stashcache.
Nicholas Clark [Sun, 23 Sep 2012 20:05:16 +0000]
-Do now also reports updates and use of PL_stashcache.

23 months agoTest the resolution behaviour for file handles and package names.
Nicholas Clark [Sun, 23 Sep 2012 20:01:14 +0000]
Test the resolution behaviour for file handles and package names.

Historical behaviour is that file handles take priority over package names,
and the use of PL_stashcache shouldn't change this.

23 months agoRestore special blocks to working order
Father Chrysostomos [Wed, 26 Sep 2012 20:12:57 +0000]
Restore special blocks to working order

I accidentally broke these in commit 85ffec3682, yet everything passed
for me under threads+mad.

PL_compcv is usually restored to its previous value at the end of
newATTRSUB when LEAVE_SCOPE is called.  But BEGIN blocks are called
before that.  I needed PL_compcv to be restored to its previ-
ous value before it was called, so I added LEAVE_SCOPE before

But that caused the name to be freed before S_process_special_blocks
got a chance to look at it.

So I have now added a new parameter to S_process_special_blocks to
allow *it* to call LEAVE_SCOPE after it determines that it is a BEGIN
block, but before it calls it.

23 months agoperlreapi.pod: Reflow verbatim lines to 79 cols.
Karl Williamson [Wed, 26 Sep 2012 17:23:05 +0000]
perlreapi.pod: Reflow verbatim lines to 79 cols.

23 months agoporting/podcheck.t: Add test for removed files
Karl Williamson [Wed, 26 Sep 2012 17:21:13 +0000]
porting/podcheck.t: Add test for removed files

If a file is removed from Perl that is in podcheck's data base,
it should be removed from the db at that time.  This adds a check
to make sure that happens.

23 months agopodcheck.t: Return success/failure from ok()
Karl Williamson [Wed, 26 Sep 2012 17:19:20 +0000]
podcheck.t: Return success/failure from ok()

This program has a private version of ok() to avoid messages that would
come from the standard one that would confuse things in these tests.
It did not return true/false prior to this commit

23 months agoregen podcheck after MPE port removed
Karl Williamson [Wed, 26 Sep 2012 16:42:29 +0000]
regen podcheck after MPE port removed

Commit b5afd3466ff5e5b70ea2921169f138f02727183e removed the MPE/iX
port, which included a file that had known (to podcheck.t) pod errors
podcheck doesn't currently check if a pod has been removed, so its data
base still contained the offending entry, which would be removed the
next time it is regenerated for another unrelated reason.  This might
confuse spelunkers of perl history, so I'm regenerating it separately

23 months agoregcomp.c: Add a less confusing #define alias
Karl Williamson [Wed, 12 Sep 2012 16:57:54 +0000]
regcomp.c: Add a less confusing #define alias

ALNUM (meaning \w) is too close to ALNUMC ([[:alnum:]]) for comfort

23 months agoregcomp.c: Properly handle no isblank(), isascii()
Karl Williamson [Wed, 12 Sep 2012 16:04:29 +0000]
regcomp.c: Properly handle no isblank(), isascii()

Configure probes whether or not these two C library functions are
present or not.  (However until commit
1c6eef9acffe4b512210edba79119e423ea4874a it didn't find isblank() even
if present.)

However the code changed by this commit always presumed both functions
were present.  That there were no failure reports from the field
indicates that Perl is being run on systems where they are present.

23 months agopod/perlrecharclass.pod: Small corrections, typos
Karl Williamson [Mon, 17 Sep 2012 21:47:17 +0000]
pod/perlrecharclass.pod: Small corrections, typos

23 months agoperlvar.pod: Document UTF8CACHE
Karl Williamson [Thu, 13 Sep 2012 01:43:02 +0000]
perlvar.pod: Document UTF8CACHE

23 months agomg_get should be called before access
Peter Martini [Tue, 25 Sep 2012 12:45:43 +0000]
mg_get should be called before access

23 months agoTest XS registration of state subs
Father Chrysostomos [Wed, 26 Sep 2012 15:53:36 +0000]
Test XS registration of state subs

my subs do not currently work yet.  I am not sure what the API
should be.

23 months agoMake PL_compcv visible to BEGIN blocks
Father Chrysostomos [Wed, 26 Sep 2012 15:47:28 +0000]
Make PL_compcv visible to BEGIN blocks

This allows BEGIN { XS_func(); } to access the currently-com-
piling pad.

BEGIN blocks were unlike any other subroutine or special block in that
PL_compcv would be set to the BEGIN block itself at run time.

23 months ago[MERGE] make regex engine handle non-null-terminated strings
David Mitchell [Wed, 26 Sep 2012 09:13:03 +0000]
[MERGE] make regex engine handle non-null-terminated strings

23 months agostop regex engine reading beyond end of string
David Mitchell [Fri, 21 Sep 2012 09:29:04 +0000]
stop regex engine reading beyond end of string

Historically the regex engine has assumed that any string passed to it
will have a trailing null char. This isn't normally an issue in perl code,
since perl strings *are* null terminated; but it could cause problems with
strings returned by XS code, or with someone calling the regex engine
directly from XS, with strend not pointing at a null char.

The engine currently relies on there being a null char in the following

First, when at the end of string, the main loop of regmatch() still reads
in the 'next' character (i.e. the character following the end of string)
even if it doesn't make any use of it. This precludes using memory mapped
files as strings for example, since the read off the end would SEGV.

Second, the matching algorithm often required the trailing character to be
\0 to work correctly: the test for 'EOF' was "if next char is null *and*
locinput >= PL_regeol, then stop". So a random non-null trailing char
could cause an overshoot.

Thirdly, some match ops require the trailing char to be null to operate
correctly; for example, \b applied at the end of the string only happens
to work because the trailing char (\0) happens to match \W.

Also, some utf8 ops will try to extract the code point at the end, which
can result in multiple bytes past the end of string being read, and
possible problems if they don't correspond to well-formed utf8.

The main fix is in S_regmatch, where the 'read next char' code has been
updated to set it to a special value, NEXTCHR_EOS instead, if we would be
reading past the end of the string.

Lots of other random bits in the regex engine needed to be fixed up too.

To track these down, I temporarily hacked regexec_flags() to make a copy
of the string but without trailing \0, then ran all the t/re/*.t tests
under valgrind to flush out all buffer overruns. So I think I've removed
most of the bad code, but by no means all of it. The code within the
various functions in regexec.c is far too complex to be able to visually
audit the code with any confidence.

23 months agoregmatch(): fix typo in TRIE commentary text
David Mitchell [Sun, 16 Sep 2012 16:39:06 +0000]
regmatch(): fix typo in TRIE commentary text

23 months agoregmatch() annotate ops and separate out branches
David Mitchell [Sun, 16 Sep 2012 16:33:08 +0000]
regmatch() annotate ops and separate out branches

Annotate each 'case OP:' in the main switch in regmatch() to show
what regex pattern this implements. About half the ops had already been
done. Also add a blank line between each 'case' statement for readability.
(no code changes)

23 months agoregmatch(): do nextchr=*locinput at top of loop
David Mitchell [Fri, 14 Sep 2012 15:19:10 +0000]
regmatch(): do nextchr=*locinput at top of loop

Currently each branch in the main regmatch() loop is responsible
re-initialising nextchar to UCHARAT(locinput) if locinput is modified.

By adding
    nextchr = UCHARAT(locinput);
to the head of the loop, we can remove most of the nextchar assignments
in the individual branches. We lose slightly for the zero-width assertions
like \b which will re-read the same nextchar, but this will make it
easier to handle non-null-terminated strings.

23 months agoregmatch(): nextchar should always be positive
David Mitchell [Fri, 14 Sep 2012 14:46:47 +0000]
regmatch(): nextchar should always be positive

Remove the one bit of code that tests for < 0, and put in a
general assert.

23 months agoregmatch(): consolidate locinput++
David Mitchell [Fri, 14 Sep 2012 11:37:33 +0000]
regmatch(): consolidate locinput++

There are several places in the code that increment locinput by 1 char
(which may or may not be 1 byte) then update nextchr.

Consolidate these into a single code block with the others goto'ing it.
This actually reduces the code more than it appears, since the CCC_TRY*
macros expand into several branches, each of which repeatthe
increment code.

23 months agoregmatch(): use nextchar where available
David Mitchell [Fri, 14 Sep 2012 10:28:08 +0000]
regmatch(): use nextchar where available

In a couple of places the code was using *locinput, where
nextchar already equalled *locinput

23 months agoStop declaring non-exported externs to non-core XS modules [perl #114516]
Steve Hay [Wed, 26 Sep 2012 07:33:20 +0000]
Stop declaring non-exported externs to non-core XS modules [perl #114516]

Hide the perl.h declarations of gete?[ug]id and getlogin on Win32 since
they are already declared in win32/win32.h, nearer to their definitions
(stub functions for UNIX compatibility) in win32/win32.c.

Also only declare them, and kill(pg)?, sbrk, chown and mkstemp, under
PERL_CORE anyway since they are not exported: including declarations for
non-exported functions just hides compiler errors about the symbols being
undefined, which doesn't help when trying to fix subsequent errors from
the linker about the symbols being unresolved. (Actually, all but sbrk,
chown and mkstemp get indirected through the perlhost layer normally
anyway, but it doesn't hurt to still hide the declarations, and helps in
the case of PERL_IMPLICIT_SYS not being defined, where only kill is
redefined to something which is exported.)

The declarations of the set[ug]id stub functions remain for now because
those two symbols are currently exported.

23 months agoTest for errors requiring .ph files.
Paul Johnson [Wed, 26 Sep 2012 00:44:44 +0000]
Test for errors requiring .ph files.

23 months agoCorrect fm vtable in perlguts.pod
Father Chrysostomos [Tue, 25 Sep 2012 21:46:35 +0000]
Correct fm vtable in perlguts.pod

fm magic uses want_vtbl_fm, which is #defined as want_vtbl_regexp.
The definition in regen/mg_vtable.pl does not affect anything except
the documentation.  It was listed as using regdata which was wrong.

23 months agoRestore perly.o build with -DDEBUGGING
Father Chrysostomos [Wed, 26 Sep 2012 01:00:42 +0000]
Restore perly.o build with -DDEBUGGING

23 months agoAdd t/test_pl* to MANIFEST
Father Chrysostomos [Tue, 25 Sep 2012 21:16:13 +0000]
Add t/test_pl* to MANIFEST

Soon, we’ll be testing that these tests work. :-)

23 months agoMake t/test_pl/tempfile.t produce more diagnostics
Brad Gilbert [Tue, 25 Sep 2012 18:12:54 +0000]
Make t/test_pl/tempfile.t produce more diagnostics

23 months agoAdded t/test_pl/tempfile.t
Brad Gilbert [Sun, 16 Sep 2012 20:28:01 +0000]
Added t/test_pl/tempfile.t

23 months agoRework tempfile() in t/test.pl to use _num_to_alpha()
Brad Gilbert [Sun, 16 Sep 2012 20:23:36 +0000]
Rework tempfile() in t/test.pl to use _num_to_alpha()

23 months agoAdded test names to some tests in t/test_pl/_num_to_alpha.t
Brad Gilbert [Sun, 16 Sep 2012 20:15:02 +0000]
Added test names to some tests in t/test_pl/_num_to_alpha.t

23 months agoAdded optional char limit to _num_to_alpha() in test.pl
Brad Gilbert [Sun, 16 Sep 2012 19:25:19 +0000]
Added optional char limit to _num_to_alpha() in test.pl

23 months agoAdd _num_to_alpha() to test.pl
Brad Gilbert [Sun, 16 Sep 2012 19:06:59 +0000]
Add _num_to_alpha() to test.pl

Also added testing for _num_to_alpha()

23 months agoMove @letters in test.pl earlier
Brad Gilbert [Tue, 25 Sep 2012 16:39:45 +0000]
Move @letters in test.pl earlier

23 months agoapparently this actually needs to be regenerated too
Jesse Luehrs [Tue, 25 Sep 2012 15:04:00 +0000]
apparently this actually needs to be regenerated too

23 months agofix regen_perly.pl for bison 2.6
Jesse Luehrs [Tue, 25 Sep 2012 08:51:14 +0000]
fix regen_perly.pl for bison 2.6

23 months ago[perl #56880] Allow v10 as a label or package name
Father Chrysostomos [Tue, 25 Sep 2012 01:18:42 +0000]
[perl #56880] Allow v10 as a label or package name

23 months agoUpdate Digest-SHA to CPAN version 5.72
Chris 'BinGOs' Williams [Tue, 25 Sep 2012 08:46:42 +0000]
Update Digest-SHA to CPAN version 5.72


  5.72  Mon Sep 24 15:22:08 MST 2012
    - adjusted module installation directory for later Perls
      -- As of 5.11 Perl searches 'site' first, so use that
        -- ref. INSTALLDIRS in Makefile.PL
      -- thanks to Robert Sedlacek for patch

23 months ago[perl #77240] Don’t warn for --subname
Father Chrysostomos [Mon, 24 Sep 2012 23:24:54 +0000]
[perl #77240] Don’t warn for --subname

23 months ago[perl #77094] Stop printf +() from reading past SP
Father Chrysostomos [Mon, 24 Sep 2012 23:08:07 +0000]
[perl #77094] Stop printf +() from reading past SP

printf with an empty list was reading past the end of the stack and
using whatever it found as its format.  If given an empty list, it
should treat the format as "".

23 months agoDon’t crash with existent but undefined &DB::DB
Father Chrysostomos [Mon, 24 Sep 2012 15:46:56 +0000]
Don’t crash with existent but undefined &DB::DB

This is a follow-up to 432d4561c48, which fixed *DB::DB without
&DB::DB, but not &DB::DB without body.

23 months agoA better fix for leaking array assignment
Father Chrysostomos [Mon, 24 Sep 2012 06:56:40 +0000]
A better fix for leaking array assignment

Instead of filling up the mortals stack with as many SVs as there are
elements, just call get-magic before creating the new SV.

(This is not so easy for hashes, as we have keys as well, and
hv_common always calls get-magic on keys.)

See commits 9c744f4f4d7 and 39984de3a8, which fixed leaking bugs, but

23 months agoFix C++ build
Father Chrysostomos [Mon, 24 Sep 2012 16:36:11 +0000]
Fix C++ build

23 months ago[perl #97958] Make reset "" match its docs
Father Chrysostomos [Mon, 24 Sep 2012 06:47:57 +0000]
[perl #97958] Make reset "" match its docs

According to the documentation, reset() with no argument resets pat-
terns.  But reset "" and reset "\0foo" were also resetting patterns.
While I was at it, I fixed embedded nulls, too, though it’s not likely
anyone is using this.  I could not fix the bug within the existing API
for sv_reset, so I created a new function and left the old one with
the old behaviour.  Call me pear-annoyed.

23 months agodon't crash with -d if DB::DB is seen but not defined [perl #114990]
Jesse Luehrs [Mon, 24 Sep 2012 05:29:06 +0000]
don't crash with -d if DB::DB is seen but not defined [perl #114990]

23 months ago[perl #107000] Don’t leak if hh copying dies
Father Chrysostomos [Sun, 23 Sep 2012 19:42:15 +0000]
[perl #107000] Don’t leak if hh copying dies

When %^H is copied on entering a new scope, if it happens to have been
tied it can die.  This was resulting in leaks, because no protections
were added to handle that case.

The two things that were leaking were the new hash in hv_copy_hints_hv
and the new value (for an element) in newSVsv.

By fixing newSVsv itself, this also fixes any potential leaks when
other pieces of code call newSVsv on explosive values.

23 months ago[perl #97466] Stop defined from propagating ref cx too far
Father Chrysostomos [Sun, 23 Sep 2012 13:02:58 +0000]
[perl #97466] Stop defined from propagating ref cx too far

23 months agoUpdate Unicode-Normalize to CPAN version 1.15
Chris 'BinGOs' Williams [Sun, 23 Sep 2012 10:03:28 +0000]
Update Unicode-Normalize to CPAN version 1.15


  1.15  Sun Sep 23 10:43:14 2012
      - perl 5.11.0 or later: Install to 'site' instead of 'perl'
        (see [rt.cpan.org #79801])

23 months agoUpdate Unicode-Collate to CPAN version 0.90
Chris 'BinGOs' Williams [Sun, 23 Sep 2012 10:01:23 +0000]
Update Unicode-Collate to CPAN version 0.90


  0.90  Sun Sep 23 10:42:26 2012
      - perl 5.11.0 or later: Install to 'site' instead of 'perl'
        (see [rt.cpan.org #79800])

23 months agoSync versions of Module-CoreList and Exporter in Maintainers.pl
Chris 'BinGOs' Williams [Sun, 23 Sep 2012 09:57:38 +0000]
Sync versions of Module-CoreList and Exporter in Maintainers.pl

23 months agoIncrease $B::Deparse::VERSION to 1.18
Father Chrysostomos [Sun, 23 Sep 2012 01:48:50 +0000]
Increase $B::Deparse::VERSION to 1.18

23 months ago[perl #94490] const fold should not trigger special split " "
Father Chrysostomos [Sun, 23 Sep 2012 00:54:12 +0000]
[perl #94490] const fold should not trigger special split " "

The easiest way to fix this was to move the special handling out of
the regexp engine.  Now a flag is set on the split op itself for
this case.  A real regexp is still created, as that is the most
convenient way to propagate locale settings, and it prevents the
need to rework pp_split to handle a null regexp.

This also means that custom regexp plugins no longer need to handle
split specially (which they all do currently).

23 months agoregexp.h: Correct comment
Father Chrysostomos [Sun, 23 Sep 2012 00:26:48 +0000]
regexp.h: Correct comment

RXf_SKIPWHITE is for split " ", which is special, *not* for split / /.

23 months agoStop array assignment from leaking on croak
Father Chrysostomos [Sat, 22 Sep 2012 19:06:45 +0000]
Stop array assignment from leaking on croak

This made a to-do test in sort.t pass, but adventitiously, so I modi-
fied it to fail again.

23 months agoStop hash assignment from leaking on croak
Father Chrysostomos [Sat, 22 Sep 2012 14:12:04 +0000]
Stop hash assignment from leaking on croak

23 months agoFree iterator when freeing tied hash
Father Chrysostomos [Sat, 22 Sep 2012 05:01:19 +0000]
Free iterator when freeing tied hash

The current iterator was leaking when a tied hash was freed or

Since we already have a mechanism, namely HvLAZYDEL, for freeing
HvEITER when not referenced elsewhere, we can use that.

23 months agoDon’t leak deleted iterator when tying hash
Father Chrysostomos [Sat, 22 Sep 2012 01:23:20 +0000]
Don’t leak deleted iterator when tying hash

23 months agoOops. make regen
Father Chrysostomos [Sat, 22 Sep 2012 21:46:29 +0000]
Oops. make regen

23 months agoAdd a #! line to config_h.SH
Father Chrysostomos [Sat, 22 Sep 2012 20:17:20 +0000]
Add a #! line to config_h.SH

It will get one automatically the next time config_h.pl is run, but
that time hasn’t come yet. :-)

23 months agoadd to authors
Sawyer X [Sat, 22 Sep 2012 13:12:26 +0000]
add to authors

23 months agoadd shebangs where missing
Sawyer X [Sat, 22 Sep 2012 12:48:09 +0000]
add shebangs where missing

23 months agotest for files that have exec bit but no shebang
Sawyer X [Fri, 21 Sep 2012 00:44:38 +0000]
test for files that have exec bit but no shebang

23 months ago[perl #96230] Stop qr// from reusing last pattern
Father Chrysostomos [Sat, 22 Sep 2012 14:13:36 +0000]
[perl #96230] Stop qr// from reusing last pattern

qr// should not be using the last-successful pattern, because it is
"(?^:)", not the empty pattern.  A stringified qr// does not use the
last-successful pattern.