This is a live mirror of the Perl 5 development currently hosted at
13 years agoeliminate next_op from struct block_loop
David Mitchell [Wed, 8 Sep 2010 11:11:41 +0000 (12:11 +0100)]
eliminate next_op from struct block_loop

This field is only used in non-threaded builds, and the comments imply
that this is because in non-threaded builds this value may be
modified. But nothing in core modifies it.

13 years agoPerl releases are tagged v5.XX.YY since v5.11.0. Updated the example
Abigail [Wed, 8 Sep 2010 13:09:26 +0000 (15:09 +0200)]
Perl releases are tagged v5.XX.YY since v5.11.0. Updated the example
to reflect this. Now it can be cut-and-pasted and actually give useful

13 years agoReorder interpreter struct to remove alignment hole created by 9a87bd09eea1d037
Nicholas Clark [Wed, 8 Sep 2010 10:42:05 +0000 (11:42 +0100)]
Reorder interpreter struct to remove alignment hole created by 9a87bd09eea1d037

nice_chunk_size was U32, and had been paired with another 32 bit value to
ensure that all 64 bit quantities were naturally 64 bit aligned. There already
was a 32 bit "hole" elsewhere - reorder the two unpaired 32 bit values next to
each other.

13 years agoRemove offer_nice_chunk(), PL_nice_chunk and PL_nice_chunk_size.
Nicholas Clark [Wed, 8 Sep 2010 09:24:04 +0000 (10:24 +0100)]
Remove offer_nice_chunk(), PL_nice_chunk and PL_nice_chunk_size.

These provided a non-public API for the hash and array code to donate free
memory direct to the SV head allocation routines, instead of returning it
to the malloc system with free().

I assume that on some older mallocs this could offer significant benefits.
However, my benchmarking on a modern malloc couldn't detect any significant
effect (positive or negative) on removing the code. Its (continued) presence,
however, has downsides

a: slightly more code complexity
b: slightly larger interpreter structure
c: in the steady state, if net creation of SVs is zero, 1 chunk of allocated
   but unused memory will exist (per thread)

So I think it best to remove it.

13 years agomake qw(...) first-class syntax
Zefram [Wed, 8 Sep 2010 07:51:29 +0000 (09:51 +0200)]
make qw(...) first-class syntax

This makes a qw(...) list literal a distinct token type for the
parser, where previously it was munged into a "(",THING,")" sequence.
The change means that qw(...) can't accidentally supply parens to parts
of the grammar that want real parens.  Due to many bits of code taking
advantage of that by "foreach my $x qw(...) {}", this patch also includes
a hack to coerce qw(...) to the old-style parenthesised THING, emitting
a deprecation warning along the way.

13 years agoStop from updating DESCRIPTION
Florian Ragwitz [Wed, 8 Sep 2010 07:43:48 +0000 (09:43 +0200)]
Stop from updating DESCRIPTION

13 years agoThis is a function, not a method
Florian Ragwitz [Wed, 8 Sep 2010 06:50:15 +0000 (08:50 +0200)]
This is a function, not a method

13 years agoMove test for #76540 to op/gv.t
Florian Ragwitz [Wed, 8 Sep 2010 03:03:23 +0000 (05:03 +0200)]
Move test for #76540 to op/gv.t

Commit e5c69c9b added the test to dist/constant/t/constant.t, probably because
the bug was initially reported in combination with However, it was
a core bug in gv_init and can be reproduced easily without

With this change dist/constant can be safely synced back to CPAN without having
to SKIP or TODO this particular test.

13 years agoFix build for threaded perls
Florian Ragwitz [Wed, 8 Sep 2010 02:18:55 +0000 (04:18 +0200)]
Fix build for threaded perls

Ooops. I broke it with commit ad7cce9f36d376e55b45dd79ec28a7f795b5ae4e.

13 years agoMove magicalize_{isa,overload} out into functions
Florian Ragwitz [Wed, 8 Sep 2010 02:06:23 +0000 (04:06 +0200)]
Move magicalize_{isa,overload} out into functions

This way c++ compilers like us again, as we don't do jumps that skip
initialisations anymore.

13 years agoperlfunc: list each/keys/values as ARRAY functions
David Golden [Wed, 8 Sep 2010 00:26:38 +0000 (20:26 -0400)]
perlfunc: list each/keys/values as ARRAY functions

As of Perl 5.12.0, the each(), keys() and values() functions also
operate on arrays, but were not listed in the table of contents section
titled "Functions for real @ARRAYs".  This patch fixes that oversight.

The delete() and exists() functions were not added, as their use on
array elements has been deprecated.

13 years agoMore tests for when(slice)
Rafael Garcia-Suarez [Tue, 7 Sep 2010 14:18:14 +0000 (16:18 +0200)]
More tests for when(slice)

13 years agoFix RT #77468: Smart matching on slices
David Leadbeater [Sat, 28 Aug 2010 21:39:58 +0000 (22:39 +0100)]
Fix RT #77468: Smart matching on slices

ref_array_or_hash did not take aslice or hslice OPs into account; wrap
them in an anonlist so that smart matching has a reference as it

13 years agoTests for caller_cx, cop_hints_*.
Ben Morrow [Tue, 31 Aug 2010 08:01:03 +0000 (09:01 +0100)]
Tests for caller_cx, cop_hints_*.

It seems that the runtime hinthash isn't returned correctly when running
under the debugger, so mark those tests TODO for now.

13 years agomake regen.
Ben Morrow [Tue, 31 Aug 2010 08:00:51 +0000 (09:00 +0100)]
make regen.

13 years agoAPI functions for accessing the runtime hinthash.
Ben Morrow [Tue, 31 Aug 2010 06:10:20 +0000 (07:10 +0100)]
API functions for accessing the runtime hinthash.

Add hinthash_fetch(sv|pv[ns]) as a replacement for refcounted_he_fetch,
which is not API (and should not be). Also add caller_cx, as the correct
XS equivalent to caller(). Lots of modules seem to have copies of this,
so a proper API function will be more maintainable in future.

13 years agoUpdate Module::CoreList Changes and entry in for the CPAN release
Chris 'BinGOs' Williams [Tue, 7 Sep 2010 07:01:20 +0000 (08:01 +0100)]
Update Module::CoreList Changes and entry in for the CPAN release

13 years agoThese files are no more
Florian Ragwitz [Tue, 7 Sep 2010 04:38:30 +0000 (06:38 +0200)]
These files are no more

13 years agoFix a pod/buildtoc warning
Florian Ragwitz [Mon, 6 Sep 2010 23:23:11 +0000 (01:23 +0200)]
Fix a pod/buildtoc warning

This was introduced with the perlXXXdelta.pod -> perldelta.pod move. buildtoc
assumed the copy target (perldelta.pod) should appear in perl.pod. Now
perldelta.pod is the copy source, but should still appear in perl.pod, while
perlXXXdelta.pod should not.

13 years agoFix some more typos and bad wording
Florian Ragwitz [Mon, 6 Sep 2010 22:41:01 +0000 (00:41 +0200)]
Fix some more typos and bad wording

Pointed out by pjcj++ and, again, ilmari++

13 years agoFix typo spotted by ilmari++
Florian Ragwitz [Mon, 6 Sep 2010 21:55:08 +0000 (23:55 +0200)]
Fix typo spotted by ilmari++

13 years agoTry to encourage sending perldelta entries
Florian Ragwitz [Mon, 6 Sep 2010 20:03:09 +0000 (22:03 +0200)]
Try to encourage sending perldelta entries

13 years agoRegenerate headers after last patch
Rafael Garcia-Suarez [Mon, 6 Sep 2010 21:51:44 +0000 (23:51 +0200)]
Regenerate headers after last patch

13 years agofunction interface to parse Perl statement
Zefram [Sat, 21 Aug 2010 17:54:04 +0000 (18:54 +0100)]
function interface to parse Perl statement

yyparse() becomes reentrant.  The yacc stack and related resources
are allocated in yyparse(), rather than in lex_start(), and they are
localised to yyparse(), preserving their values from any outer parser.

yyparse() now takes a parameter which determines which production it
will parse at the top level.  New API function parse_fullstmt() uses this
facility to parse just a single statement.  The top-level single-statement
production that is used for this then messes with the parser's head so
that the parsing stops without seeing EOF, and any lookahead token seen
after the statement is pushed back to the lexer.

13 years agoUpgrade to Test-Simple-0.97_01
Karl Williamson [Mon, 6 Sep 2010 16:22:03 +0000 (10:22 -0600)]
Upgrade to Test-Simple-0.97_01

13 years agoperldelta up to f4beb78
Florian Ragwitz [Mon, 6 Sep 2010 16:00:47 +0000 (18:00 +0200)]
perldelta up to f4beb78

13 years agoA couple of perldelta fixes
Father Chrysostomos [Mon, 6 Sep 2010 15:30:23 +0000 (17:30 +0200)]
A couple of perldelta fixes

13 years ago$ref++, $ref-- leaked referent
David Mitchell [Mon, 6 Sep 2010 15:24:47 +0000 (16:24 +0100)]
$ref++, $ref-- leaked referent

[perl #9466]

pp_postinc and pp_postdec used a pad TARG to return a copy of the
original value. When that value was a reference, it meant a copy
of the reference would hang out in the pad forever and so the referent
would leak. Fix this by using a mortal instead.

13 years ago[perl #76138] perl inadvertently destroys signal handlers as of f746176000
Father Chrysostomos [Sun, 5 Sep 2010 20:03:53 +0000 (22:03 +0200)]
[perl #76138] perl inadvertently destroys signal handlers as of f746176000

Stop magic applied to $!, %SIG, et al. from applying to similarly-
named variables in other packages.

13 years agort74170: test case that triggered the problem
Father Chrysostomos [Mon, 6 Sep 2010 13:41:01 +0000 (23:41 +1000)]
rt74170: test case that triggered the problem

13 years agort74170: handle the stack changing in the custom sort functions
Tony Cook [Mon, 6 Sep 2010 13:40:24 +0000 (23:40 +1000)]
rt74170: handle the stack changing in the custom sort functions

Based on work by Father Chrysostomos <>, but all
my fault.

13 years agoCleanup perl5135delta.pod instead of perldelta.pod
Jan Dubois [Sun, 5 Sep 2010 19:20:04 +0000 (12:20 -0700)]
Cleanup perl5135delta.pod instead of perldelta.pod

Amends 37ee6528c

13 years agoRemove duplicate command to rebuild uudmap.h
Jan Dubois [Sun, 5 Sep 2010 19:13:21 +0000 (12:13 -0700)]
Remove duplicate command to rebuild uudmap.h

Fixes 6999193b
(was only broken for win32/Makefile, but not for win32/

13 years agoUpdate Unicode-Collate to CPAN version 0.59
Chris 'BinGOs' Williams [Sun, 5 Sep 2010 16:53:23 +0000 (17:53 +0100)]
Update Unicode-Collate to CPAN version 0.59


  0.59  Sun Sep  5 17:03:52 2010
    - U::C::Locale newly supports locales: az, fil, ha, lt, mt, tr, wo, yo.
    - precomposites of a-uml, o-uml, and u-uml are tailored as well.
        (affected locales: da, et, fi, fo, is, kl, nb, nn, sk, sv)

13 years agoRemove NEXT's README and Changes files
Florian Ragwitz [Sun, 5 Sep 2010 14:46:00 +0000 (16:46 +0200)]
Remove NEXT's README and Changes files

Those already live in NEXT's upstream repository. There's no point having them
in core as well.

13 years agoremove dist/Filter-Simple-Changes from MANIFEST
Florian Ragwitz [Sun, 5 Sep 2010 14:40:04 +0000 (16:40 +0200)]
remove dist/Filter-Simple-Changes from MANIFEST

13 years agoUpdate release manager guide for perldelta move
Florian Ragwitz [Wed, 1 Sep 2010 19:01:28 +0000 (21:01 +0200)]
Update release manager guide for perldelta move

13 years agoMove the latest perldelta to pod/perldelta.pod
Florian Ragwitz [Sun, 22 Aug 2010 08:43:37 +0000 (10:43 +0200)]
Move the latest perldelta to pod/perldelta.pod

This way patches including perldelta entries will apply properly, no matter when
they are applied. If there's conflicts, they'll at least be in the right file.

13 years agoUpdate CPANPLUS-Dist-Build to CPAN version 0.48
Chris 'BinGOs' Williams [Sun, 5 Sep 2010 14:23:24 +0000 (15:23 +0100)]
Update CPANPLUS-Dist-Build to CPAN version 0.48


  0.48 Sun Sep  5 15:14:17 BST 2010
  - Explicitly set the location of cpanp-run-perl when under PERL_CORE
    in t/inc/
  - Make sure we find cpanp-run-perl in ../../utils when PERL_CORE
  - Correct spelling in POD, pointed out by H.Merijn Brand

  These changes have already been in blead for a while, the CPAN release
  brings CPAN and blead into sync

13 years agoperldelta up to be34c38
Florian Ragwitz [Sun, 5 Sep 2010 13:31:55 +0000 (15:31 +0200)]
perldelta up to be34c38

13 years agoUpgrade Filter::Simple to 0.85
Steffen Mueller [Sun, 5 Sep 2010 14:08:58 +0000 (16:08 +0200)]
Upgrade Filter::Simple to 0.85

Purely maintenance. No functional changes.

13 years agoFilter::Simple Changelog needn't be part of the core
Steffen Mueller [Sun, 5 Sep 2010 14:06:50 +0000 (16:06 +0200)]
Filter::Simple Changelog needn't be part of the core

13 years agoremove removed file dist/Cwd/Changes from the MANIFEST
Tony Cook [Sun, 5 Sep 2010 13:10:57 +0000 (23:10 +1000)]
remove removed file dist/Cwd/Changes from the MANIFEST

13 years agoDo not ship the PathTools Changes
Steffen Mueller [Sun, 5 Sep 2010 10:21:40 +0000 (12:21 +0200)]
Do not ship the PathTools Changes

There is no reason (I can think of) to track and ship the PathTools CPAN
change log in core perl.

13 years agoFix core-cpan-diff for dists that extract weirdly
Florian Ragwitz [Sun, 5 Sep 2010 00:08:52 +0000 (02:08 +0200)]
Fix core-cpan-diff for dists that extract weirdly

The recent "Foo-Bar-1.23-withoutworldwriteables.tar.gz" dists are an example of
that. They'll extract to Foo-Bar-1.23/.

Caching of the untarred dists probably doesn't work for those dists, but at
least it works at all.

13 years agoUpdate's META.yml
Florian Ragwitz [Sat, 4 Sep 2010 23:11:00 +0000 (01:11 +0200)]
Update's META.yml

It's there for Rafael's convenience when doing CPAN releases, as there's no
separate base repository.

13 years agoSynchronize with bignum 0.24
Florian Ragwitz [Sat, 4 Sep 2010 13:03:47 +0000 (15:03 +0200)]
Synchronize with bignum 0.24

13 years agoSynchronize with Math::BigInt::FastCalc 0.21
Florian Ragwitz [Fri, 3 Sep 2010 21:32:23 +0000 (23:32 +0200)]
Synchronize with Math::BigInt::FastCalc 0.21

13 years agoUpstream for Math::BigRat is CPAN
Florian Ragwitz [Fri, 3 Sep 2010 18:14:11 +0000 (20:14 +0200)]
Upstream for Math::BigRat is CPAN

13 years agoMath::BigInt 1.90 just shipped to CPAN
Florian Ragwitz [Fri, 3 Sep 2010 15:34:20 +0000 (17:34 +0200)]
Math::BigInt 1.90 just shipped to CPAN

13 years agoperldelta up to 45b279e
Florian Ragwitz [Fri, 3 Sep 2010 15:02:31 +0000 (17:02 +0200)]
perldelta up to 45b279e

13 years agoBump $Errno::VERSION after change be54382
Florian Ragwitz [Fri, 3 Sep 2010 14:45:47 +0000 (16:45 +0200)]
Bump $Errno::VERSION after change be54382

13 years agoIn uudmap.h rule, nothing is different from the absence of anything.
Craig A. Berry [Fri, 3 Sep 2010 02:44:57 +0000 (21:44 -0500)]
In uudmap.h rule, nothing is different from the absence of anything.

Follow-up to 6999193ba9069e620831bed0f85efa43ff6e283a, which broke
the build on VMS because the rule to build uudmap.h was unknown.

13 years agoBump Math::BigInt{,::Calc}::VERSION for bnok fixes
Florian Ragwitz [Fri, 3 Sep 2010 01:44:11 +0000 (03:44 +0200)]
Bump Math::BigInt{,::Calc}::VERSION for bnok fixes

13 years agoFix Math::BigInt's bnok($k) for $k==0 and $k==$n-1
Piotr Fusik [Fri, 3 Sep 2010 00:41:00 +0000 (02:41 +0200)]
Fix Math::BigInt's bnok($k) for $k==0 and $k==$n-1

Signed-off-by: Florian Ragwitz <>
[ fixed bigflt tests and test plans]

13 years agoblead is upstream for bignum
Florian Ragwitz [Thu, 2 Sep 2010 23:12:17 +0000 (01:12 +0200)]
blead is upstream for bignum

13 years agoblead is upstream for Math-BigInt-FastCalc
Florian Ragwitz [Thu, 2 Sep 2010 22:12:22 +0000 (00:12 +0200)]
blead is upstream for Math-BigInt-FastCalc

13 years agoblead is upstream for Math-BigInt
Florian Ragwitz [Thu, 2 Sep 2010 21:11:26 +0000 (23:11 +0200)]
blead is upstream for Math-BigInt

13 years agoIn XS_VERSION_BOOTCHECK, correct the flags argument of get_sv() to 0.
Nicholas Clark [Thu, 2 Sep 2010 15:21:43 +0000 (16:21 +0100)]
In XS_VERSION_BOOTCHECK, correct the flags argument of get_sv() to 0.

It should not be FALSE, because it's a bitmap, not a boolean.

13 years agoChange the first argument of Perl_fetch_cop_label() to COP *
Nicholas Clark [Thu, 2 Sep 2010 14:57:24 +0000 (15:57 +0100)]
Change the first argument of Perl_fetch_cop_label() to COP *

From a suggestion from Ben Morrow.

The first argument used to be struct refcounted_he *, which exposed an
implementation detail - that the COP's labels are (now) stored in this way.

Google Code Search and an unpacked CPAN both fail to find any users of this
API, so the impact should be minimal.

13 years agoFor the generated, avoid a runtime lexical which is constant.
Nicholas Clark [Thu, 2 Sep 2010 10:44:51 +0000 (11:44 +0100)]
For the generated, avoid a runtime lexical which is constant.

XSLoader_pm.PL had been resolving $Config::Config{dlext}, and writing it as
the constant initialiser for a lexical variable in In turn,
that lexical was used only once, in string interpolation. So the interpolation
can be done instead at build time.

13 years agoFix test (see <>)
Abigail [Thu, 2 Sep 2010 09:55:07 +0000 (11:55 +0200)]
Fix test (see <>)

13 years agoSanity check on Errno values.
Curtis Jewell [Tue, 24 Aug 2010 20:07:51 +0000 (14:07 -0600)]
Sanity check on Errno values.

On a few machines (Win32/gcc using mingw64 headers)
will find a value that is not numeric for a proposed error
key. This change adds a sanity check to discard such keys.

13 years agot/lib/ must convert $^X to an absolute path before changing directory.
Nicholas Clark [Wed, 1 Sep 2010 20:24:54 +0000 (21:24 +0100)]
t/lib/ must convert $^X to an absolute path before changing directory.

Not every operating system has /proc/self/exe

13 years agoTests using t/lib/ need to run in separate directories.
Nicholas Clark [Wed, 1 Sep 2010 18:05:52 +0000 (19:05 +0100)]
Tests using t/lib/ need to run in separate directories.

Commit 8f776eae73090661 turned out to be a bit optimistic with
"should be capable of running in parallel", as the temporary files and
modules written out by the various test scripts have clashing names.
Hence run each test a private subdirectory.

13 years agoRT 74444 - emits warnings when installperl is run with destdir
Todd Rinaldo [Wed, 1 Sep 2010 16:18:43 +0000 (11:18 -0500)]
RT 74444 - emits warnings when installperl is run with destdir

running the following produces 7 warnings like this on my system:
'LD_LIBRARY_PATH=/usr/home/build/perl-5.12.0  ./perl installperl --destdir=/var/local/tmp/perl-root'

Use of uninitialized value $dev2 in numeric eq (==) at ./ line 123.

replacing == with ~~ (the smart match operator) seems to be the best DWIM fix.

13 years agoAdd alternate email address for Todd
Florian Ragwitz [Wed, 1 Sep 2010 16:41:58 +0000 (18:41 +0200)]
Add alternate email address for Todd

13 years agoIgnore droppings from lib/h2ph.t
Florian Ragwitz [Wed, 1 Sep 2010 16:14:18 +0000 (18:14 +0200)]
Ignore droppings from lib/h2ph.t

13 years agoAdd test for deprecation warnings from h2ph
Florian Ragwitz [Wed, 1 Sep 2010 14:54:38 +0000 (16:54 +0200)]
Add test for deprecation warnings from h2ph

This makes sure the fix for RT#74404 is working.

13 years agoRT 74404 - h2ph using deprecated goto
Todd Rinaldo [Wed, 1 Sep 2010 16:05:26 +0000 (11:05 -0500)]
RT 74404 - h2ph using deprecated goto

It looks like h2ph makes copious use of global variables. As a result, I can
commit this evil for a minimal patch. No code was changed moving it into the
sub other than passing $proto, which is the only non-global used (but not

I tried a 5.12.0 build with this patch and it seems to be passing tests.
There is a lib/h2ph.t file but I don't know what it tests exactly.

13 years agoStop Tie::Hash->TIEHASH() looping forever.
Nicholas Clark [Wed, 1 Sep 2010 15:54:56 +0000 (16:54 +0100)]
Stop Tie::Hash->TIEHASH() looping forever.

This change is analogous to bc3707112523ba7a in Tie::Scalar, but with far fewer

13 years agopp_match: fix confusing layout in an if()
David Mitchell [Wed, 1 Sep 2010 16:05:22 +0000 (17:05 +0100)]
pp_match: fix confusing layout in an if()

    if ((    !global
rather than
    if (    (!global
made it hard to follow the precedence.
Tidy up the rest of the expression while I'm at it.

13 years agoCopy RE capture buf on overload as well as TEMP
David Mitchell [Wed, 1 Sep 2010 15:59:03 +0000 (16:59 +0100)]
Copy RE capture buf on overload as well as TEMP

Partial fix for [perl #77084]. Sometimes pp_match makes a copy of the
original SV's string for the later use of $1 et al; in particular if the
SV is TEMP (so will soon go away).

Make it do the same if the SV is overloaded, as the string return is most
certainly temporary!

(Also tweak the tests to make them more likely to fail on badness by
creating new stings that will likely reallocate freed buffer).

13 years agoperldelta up to 137127d
Florian Ragwitz [Wed, 1 Sep 2010 15:27:46 +0000 (17:27 +0200)]
perldelta up to 137127d

13 years agot/harness can run the tests lib/*.t in parallel with each other.
Nicholas Clark [Wed, 1 Sep 2010 13:32:47 +0000 (14:32 +0100)]
t/harness can run the tests lib/*.t in parallel with each other.

All these tests are for different modules, and should be capable of running
in parallel. Hopefully this will reduce the chance of seeing lib/warnings.t
running all alone on a multi-core machine.

13 years agoClarify CPU cache improvements entry in 5133delta
Florian Ragwitz [Wed, 1 Sep 2010 13:40:04 +0000 (15:40 +0200)]
Clarify CPU cache improvements entry in 5133delta

13 years agoPL_my_cxt_list is only available with PERL_IMPLICIT_CONTEXT
Tony Cook [Wed, 1 Sep 2010 13:14:51 +0000 (23:14 +1000)]
PL_my_cxt_list is only available with PERL_IMPLICIT_CONTEXT

13 years agofix some 64-bit casts under DEBUG_LEAKING_SCALARS
David Mitchell [Wed, 1 Sep 2010 11:52:32 +0000 (12:52 +0100)]
fix some 64-bit casts under DEBUG_LEAKING_SCALARS

13 years agofix indentation of MY_CXT #defines
David Mitchell [Wed, 1 Sep 2010 11:07:32 +0000 (12:07 +0100)]
fix indentation of MY_CXT #defines

Only whitespace changes

13 years agomerge two similar MY_CXT code branches
David Mitchell [Wed, 1 Sep 2010 10:59:01 +0000 (11:59 +0100)]
merge two similar MY_CXT code branches

No functional changes

13 years agoMY_CXT macros: make the two sets of defs similar
David Mitchell [Wed, 1 Sep 2010 10:49:02 +0000 (11:49 +0100)]
MY_CXT macros: make the two sets of defs similar

Three years ago there was a cut and paste of all the MY_CXT macros
into a second #ifdef PERL_GLOBAL_STRUCT_PRIVATE branch with minor
modifications (Bad programmer! No cookie!). Make the two branches more
similar in preparation for a partial merge. No functional changes.

13 years agoPL_my_cxt_list leaks
Jirka Hruška [Wed, 1 Sep 2010 10:18:43 +0000 (11:18 +0100)]
PL_my_cxt_list leaks

[perl #77352]

PL_my_cxt_list was never freed

13 years agoMemory leak cloning PVGVs
Jirka Hruška [Wed, 1 Sep 2010 09:59:35 +0000 (10:59 +0100)]
Memory leak cloning PVGVs

[perl #77352]

In S_sv_dup_common, Perl_rvpv_dup was called twice on a non-GP PVGV value,
causing the first duped value to be leaked

13 years agoMove the call to Perl_cv_clone() into S_doform(), from its two callers.
Nicholas Clark [Wed, 1 Sep 2010 09:56:47 +0000 (10:56 +0100)]
Move the call to Perl_cv_clone() into S_doform(), from its two callers.

This reduces code duplication slightly, and reduces the object code size.

13 years agoRefactor Perl_store_cop_label() to avoid exposing struct refcounted_he *.
Nicholas Clark [Wed, 1 Sep 2010 06:24:59 +0000 (07:24 +0100)]
Refactor Perl_store_cop_label() to avoid exposing struct refcounted_he *.

Instead pass in a COP, as suggested by Ben Morrow. Also add length and flags
parameters, and remove the comment suggesting this change. The underlying
storage mechanism can honour length and UTF8/not, so there is no harm in
exposing this one level higher.

13 years agoBump blib::VERSION after 3f9bb6b
Florian Ragwitz [Wed, 1 Sep 2010 02:41:39 +0000 (04:41 +0200)]
Bump blib::VERSION after 3f9bb6b

13 years ago[perl #77492] open $fh, ">", \*glob causes SEGV
Father Chrysostomos [Wed, 1 Sep 2010 01:00:38 +0000 (03:00 +0200)]
[perl #77492] open $fh, ">", \*glob causes SEGV

PerlIO_layer_from_ref must not treat a real glob as a scalar. This
function was not updated when SVt_PVGV was moved before SVt_PVLV.

13 years agoCorrect errors in the use of multiple targets for bitcount.h and uudmap.h
Nicholas Clark [Tue, 24 Aug 2010 14:25:36 +0000 (15:25 +0100)]
Correct errors in the use of multiple targets for bitcount.h and uudmap.h

The fixes of Makefile.SH in 0f13ebd5d71f8177 also need to be applied on Win32
and VMS.

13 years agoMerge the two 5134delta entries for uniary functions
Florian Ragwitz [Tue, 31 Aug 2010 20:16:29 +0000 (22:16 +0200)]
Merge the two 5134delta entries for uniary functions

13 years ago[perl #77234] Change 649d02de (unary prototypes) changes precedence
Father Chrysostomos [Tue, 31 Aug 2010 20:13:06 +0000 (22:13 +0200)]
[perl #77234] Change 649d02de (unary prototypes) changes precedence

This patch retroactively adds a description of the breakage to
perl5134delta so it will be copied eventually into perl5140delta.

13 years agoremove dead listexprcom production from grammar
Zefram [Mon, 23 Aug 2010 19:49:45 +0000 (20:49 +0100)]
remove dead listexprcom production from grammar

The third production of <listexprcom>, "expr ','", could never be invoked,
because <expr> can already end with arbitrarily many commas.

13 years agobetter silencing of APItest deprecation warning
David Mitchell [Tue, 31 Aug 2010 18:04:18 +0000 (19:04 +0100)]
better silencing of APItest deprecation warning

Commit 95ddc6755c1ff41d06e5afc2212c85f918ebcd28 attempted to add
-Wno-deprecated-declarations' to the ccflags for APItest.xs;
however it ended up deleting existing flags rather than appending.

Under -DDEBUG_LEAKING_SCALARS this caused the PerlInterpreter structure
to have different field alignments in core and APItest.xs. Ouch!

13 years agoAdd Zefram as our release manager victim for 20 December
Jesse Vincent [Tue, 31 Aug 2010 18:11:23 +0000 (14:11 -0400)]
Add Zefram as our release manager victim for 20 December

13 years agoAdd two missing break in Perl_magic_set's big switch
Vincent Pit [Tue, 31 Aug 2010 11:10:26 +0000 (13:10 +0200)]
Add two missing break in Perl_magic_set's big switch

This fixes $^A being reset when $1..$2 are localized before any regexp match

13 years agoRemove unneeded use of INT2PTR() in the XS for Tie::Hash::NamedCapture.
Nicholas Clark [Mon, 30 Aug 2010 17:51:25 +0000 (18:51 +0100)]
Remove unneeded use of INT2PTR() in the XS for Tie::Hash::NamedCapture.

13 years agoStart perl5135delta
Florian Ragwitz [Tue, 31 Aug 2010 01:59:10 +0000 (03:59 +0200)]
Start perl5135delta

13 years agoModule::CoreList updates for 5.12.2
Jesse Vincent [Tue, 31 Aug 2010 01:43:01 +0000 (21:43 -0400)]
Module::CoreList updates for 5.12.2

13 years agoAvoid needless use of UTF8=1 format [RT#56336]
Eric Brine [Sun, 22 Aug 2010 05:17:26 +0000 (22:17 -0700)]
Avoid needless use of UTF8=1 format [RT#56336]

Some literals (e.g. q'abc') don't set the UTF8 flag for pure ASCII literals.
Others (e.g. -abc) do. This should be consistent.

13 years agoTODO tests for avoid needless use of UTF8=1 format [RT#56336]
Eric Brine [Sun, 22 Aug 2010 03:14:08 +0000 (20:14 -0700)]
TODO tests for avoid needless use of UTF8=1 format [RT#56336]

Some literals (e.g. q'abc') don't set the UTF8 flag for pure ASCII literals.
Others (e.g. -abc) do. This should be consistent.

13 years agoadd more tests to lib/mauve.t so it tests also that mauve::reftype can return "LVALUE"
Zsbán Ambrus [Mon, 30 Aug 2010 17:24:18 +0000 (19:24 +0200)]
add more tests to lib/mauve.t so it tests also that mauve::reftype can return "LVALUE"