This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perl5.git
3 years agoConfigure: assume C89 headers in d_c99_variadic_macros probe
Aaron Crane [Thu, 12 Oct 2017 15:44:29 +0000 (17:44 +0200)]
Configure: assume C89 headers in d_c99_variadic_macros probe

This matches a recent change to metaconfig.

3 years agoRely on C89 snprintf()
Aaron Crane [Thu, 12 Oct 2017 15:35:52 +0000 (17:35 +0200)]
Rely on C89 snprintf()

3 years agoRely on C89 vsnprintf()
Aaron Crane [Thu, 12 Oct 2017 15:22:43 +0000 (17:22 +0200)]
Rely on C89 vsnprintf()

We assume it has the standard behaviour that, if the buffer is too short, it
returns the desired buffer length.

3 years agoDon't use VOL internally, because "volatile" works just fine
Aaron Crane [Thu, 12 Oct 2017 14:26:56 +0000 (16:26 +0200)]
Don't use VOL internally, because "volatile" works just fine

However, we do preserve it outside PERL_CORE for the use of XS authors.

3 years agoAssume C89 "volatile" keyword
Aaron Crane [Thu, 12 Oct 2017 13:46:20 +0000 (15:46 +0200)]
Assume C89 "volatile" keyword

The Configure changes here were generated using a version of metaconfig
that makes U/perl/perlxv.U assume that the keyword exists, and prevents
U/modified/d_volatile.U from promising to define a "volatile" keyword;
otherwise, those units would bring in the relevant Configure probe anyway.

3 years agoAssume we have C89 memchr()
Aaron Crane [Thu, 12 Oct 2017 13:18:42 +0000 (15:18 +0200)]
Assume we have C89 memchr()

3 years agoAssume we have sane C89 memcmp()
Aaron Crane [Thu, 12 Oct 2017 12:31:56 +0000 (14:31 +0200)]
Assume we have sane C89 memcmp()

"Sane" means that it works correctly on bytes with their high bit set, as
C89 also requires.

We therefore no longer need to probe for and/or use BSD bcmp().

3 years agoDrop support for PowerUX / Power MAX OS
Aaron Crane [Thu, 12 Oct 2017 13:04:00 +0000 (15:04 +0200)]
Drop support for PowerUX / Power MAX OS

It is almost impossible to find any information about this platform on the
internet, which strongly suggests that it's as dead as dead can be.

3 years agoAssume we have C89 memcpy() and memmove()
Aaron Crane [Thu, 12 Oct 2017 12:00:03 +0000 (14:00 +0200)]
Assume we have C89 memcpy() and memmove()

We can therefore also avoid probing for and/or using BSD bcopy().

3 years agoDon't look for a "safe" memcpy()
Aaron Crane [Thu, 12 Oct 2017 11:47:22 +0000 (13:47 +0200)]
Don't look for a "safe" memcpy()

C89 says that, if you want to copy overlapping memory blocks, you must use
memmove(), and that attempt to copy overlapping memory blocks using memcpy()
yields undefined behaviour. So we should never even attempt to probe for a
system memcpy() implementation that just happens to handle overlapping
memory blocks. In particular, the compiler might compile the probe program
in such a way that Configure thinks overlapping memcpy() works even when it
doesn't.

This has the additional advantage of removing a Configure probe that needs
to execute a target-platform program on the build host.

3 years agoAssume we have C89 memset()
Aaron Crane [Thu, 12 Oct 2017 09:02:27 +0000 (11:02 +0200)]
Assume we have C89 memset()

This means we also never need to consider using BSD bzero().

3 years agoRemove USE_STRUCT_COPY in favour of C89 struct assignment
Aaron Crane [Thu, 12 Oct 2017 08:33:33 +0000 (10:33 +0200)]
Remove USE_STRUCT_COPY in favour of C89 struct assignment

At least for now, we retain the StructCopy() macro, but its definition
always just uses struct assignment.

3 years agoConfigure: fix nl_langinfo_l() probe
Aaron Crane [Sat, 21 Oct 2017 15:42:51 +0000 (16:42 +0100)]
Configure: fix nl_langinfo_l() probe

This corresponds to metaconfig ad82f5bfbeafc125c0aed2b83391ed8d43bf31b2.

3 years agoAdd Configure changes to perldelta
H.Merijn Brand [Sat, 21 Oct 2017 15:41:04 +0000 (17:41 +0200)]
Add Configure changes to perldelta

3 years agoAdd new Configure probes
H.Merijn Brand [Sat, 21 Oct 2017 15:29:47 +0000 (17:29 +0200)]
Add new Configure probes

This is a merge off several new probes in metaconfig done in the new
shared developing environment by several authors

Thanks to all that contributed!

3 years agoToday's M::CL is up on CPAN already
Steve Hay [Fri, 20 Oct 2017 23:17:00 +0000 (00:17 +0100)]
Today's M::CL is up on CPAN already

3 years agoPrepare Module::CoreList for 5.27.6
Steve Hay [Fri, 20 Oct 2017 23:11:33 +0000 (00:11 +0100)]
Prepare Module::CoreList for 5.27.6

3 years agoBump version to 5.27.6
Steve Hay [Fri, 20 Oct 2017 22:54:15 +0000 (23:54 +0100)]
Bump version to 5.27.6

3 years agoperldelta - Fix version for the next run of Porting/acknowledgements.pl
Steve Hay [Fri, 20 Oct 2017 22:45:38 +0000 (23:45 +0100)]
perldelta - Fix version for the next run of Porting/acknowledgements.pl

And fix the template file so that the impending version bump will leave it
correct for next time too.

3 years agoNew perldelta for 5.27.6
Steve Hay [Fri, 20 Oct 2017 22:41:00 +0000 (23:41 +0100)]
New perldelta for 5.27.6

3 years agoTick off 5.27.5
Steve Hay [Fri, 20 Oct 2017 22:27:06 +0000 (23:27 +0100)]
Tick off 5.27.5

3 years agoAdd epigraph for 5.27.5
Steve Hay [Fri, 20 Oct 2017 22:26:13 +0000 (23:26 +0100)]
Add epigraph for 5.27.5

3 years agoPerl 5.27.5 today v5.27.5
Steve Hay [Fri, 20 Oct 2017 20:20:22 +0000 (21:20 +0100)]
Perl 5.27.5 today

3 years agoFinalize perldelta
Steve Hay [Fri, 20 Oct 2017 20:16:59 +0000 (21:16 +0100)]
Finalize perldelta

3 years agoUpdate Module::CoreList for 5.27.5
Steve Hay [Fri, 20 Oct 2017 20:09:37 +0000 (21:09 +0100)]
Update Module::CoreList for 5.27.5

3 years agoperldelta - Fix broken links
Steve Hay [Fri, 20 Oct 2017 20:06:10 +0000 (21:06 +0100)]
perldelta - Fix broken links

3 years agoperldelta - Remove boilerplate
Steve Hay [Fri, 20 Oct 2017 19:52:32 +0000 (20:52 +0100)]
perldelta - Remove boilerplate

3 years agoperldelta - Update selected bug fixes etc
Steve Hay [Fri, 20 Oct 2017 19:47:46 +0000 (20:47 +0100)]
perldelta - Update selected bug fixes etc

3 years agoperldelta - Update modules
Steve Hay [Fri, 20 Oct 2017 18:45:46 +0000 (19:45 +0100)]
perldelta - Update modules

3 years agoSuppress spurious corelist-perldelta.pl warning
Steve Hay [Fri, 20 Oct 2017 18:45:04 +0000 (19:45 +0100)]
Suppress spurious corelist-perldelta.pl warning

3 years agoTeach corelist-perldelta.pl about libnet, podlators and Sub::Util
Steve Hay [Fri, 20 Oct 2017 18:43:41 +0000 (19:43 +0100)]
Teach corelist-perldelta.pl about libnet, podlators and Sub::Util

3 years agoPorting/Maintainers.pl - version's vxs.inc is now CUSTOMIZED as well
Steve Hay [Fri, 20 Oct 2017 07:25:35 +0000 (08:25 +0100)]
Porting/Maintainers.pl - version's vxs.inc is now CUSTOMIZED as well

3 years agoPorting/Maintainers.pl - NEXT is CUSTOMIZED
Steve Hay [Fri, 20 Oct 2017 07:22:33 +0000 (08:22 +0100)]
Porting/Maintainers.pl - NEXT is CUSTOMIZED

3 years agoDeparse.t tweak
Father Chrysostomos [Fri, 20 Oct 2017 04:35:51 +0000 (21:35 -0700)]
Deparse.t tweak

The revert two commits ago causes B::Deparse to go through a different
code path when dumping the constant.  Ideally the output should be the
same either way (i.e., consistently with or without a semicolon, not
sometimes one way and sometimes the other).  But for now, just make
the test more lenient.

3 years agoperldelta for 14062320f
Father Chrysostomos [Fri, 20 Oct 2017 03:14:29 +0000 (20:14 -0700)]
perldelta for 14062320f

3 years agoTemporarily revert CV-in-stash optimisation
Father Chrysostomos [Fri, 20 Oct 2017 01:18:37 +0000 (18:18 -0700)]
Temporarily revert CV-in-stash optimisation

This reverts two hunks from 6881372e19 to allow CPAN modules some time
to conform to the optimisation.

3 years agoUpdate win32/config_H.[gv]c canned configuration files
Steve Hay [Thu, 19 Oct 2017 12:55:03 +0000 (13:55 +0100)]
Update win32/config_H.[gv]c canned configuration files

This is done using the regen_config_h targets in the Windows makefiles,
being careful to restore a couple of things otherwise lost from the
config_H.gc file (a couple of STRINGIFY #undefs, comments above the
HASATTRIBUTE_* #defines, and the #define of HAS_MKSTEMP).

The files are now in sync with the top-level master configuration file,
config_h.SH.

3 years agoUse SvLEN_set/SvCUR_set in a few extra locations
Nicolas R [Mon, 25 Sep 2017 18:37:06 +0000 (13:37 -0500)]
Use SvLEN_set/SvCUR_set in a few extra locations

SvLEN was set without using the generic macro SvLEN_set.
Use it in three extra locations, and also use SvCUR_set
instead of SvCUR.

3 years agoMerge branch 'smoke-me/ilmari/exec-empty-argv' into blead
Dagfinn Ilmari Mannsåker [Wed, 18 Oct 2017 16:04:57 +0000 (17:04 +0100)]
Merge branch 'smoke-me/ilmari/exec-empty-argv' into blead

3 years agoPerldelta for exec PROGRAM LIST with empty LIST
Dagfinn Ilmari Mannsåker [Wed, 18 Oct 2017 11:48:30 +0000 (12:48 +0100)]
Perldelta for exec PROGRAM LIST with empty LIST

3 years ago[perl #131730] Fix exec PROGRAM LIST with empty LIST on VMS
Dagfinn Ilmari Mannsåker [Fri, 1 Sep 2017 11:11:53 +0000 (12:11 +0100)]
[perl #131730] Fix exec PROGRAM LIST with empty LIST on VMS

Only when being Unixish, because I have no idea if
setup_argstr()/vms_do_exec() handle it.

3 years ago[perl #131730] Fix exec PROGRAM LIST with empty LIST
Dagfinn Ilmari Mannsåker [Sat, 22 Jul 2017 18:32:40 +0000 (19:32 +0100)]
[perl #131730] Fix exec PROGRAM LIST with empty LIST

This should call execvp() with an empty argv array (containing only the
terminating NULL pointer), but was instead just returning false (and not
setting $!).

Executing a program with an empty argv array is valid ISO C: it merely
requires argc to be nonnegative and argv[argc] to be NULL.  POSIX states
that that argv[0] "should point to a filename string that is associated
with the process being started", but "should" only applies to
applications claiming strict POSIX conformance.  Perl does not, and
certainly should not impose it on perl programs.

This also requires handling the case where both PROGRAM and LIST are
empty, to avoid calling execvp(NULL, …), which _is_ invalid.  I've made
it return ENOENT, which it the error POSIX specifies for execvp("", …).

3 years agoUpgrade Test-Simple from version 1.302101 to 1.302103
Steve Hay [Wed, 18 Oct 2017 07:29:16 +0000 (08:29 +0100)]
Upgrade Test-Simple from version 1.302101 to 1.302103

3 years agothe unopened test should use a file handle distinct from "OTHER"
Tony Cook [Tue, 17 Oct 2017 03:39:44 +0000 (14:39 +1100)]
the unopened test should use a file handle distinct from "OTHER"

used a name and added a comment so this isn't broken again

3 years ago[perl #127063] suppress warning on assigning 0 to $[
Tony Cook [Mon, 4 Jan 2016 00:13:19 +0000 (11:13 +1100)]
[perl #127063] suppress warning on assigning 0 to $[

3 years agoPERL5LIB documentation: '.' is no longer in @INC by default
Lukas Mai [Mon, 16 Oct 2017 20:39:45 +0000 (22:39 +0200)]
PERL5LIB documentation: '.' is no longer in @INC by default

3 years agocustomise Pod::Perldoc to fix output misbehaviour
Zefram [Sat, 7 Oct 2017 04:17:19 +0000 (05:17 +0100)]
customise Pod::Perldoc to fix output misbehaviour

Pod::Perldoc has, since version 3.20, exhibited various kinds of
misbehaviour relating to a bad default choice of formatter.  Output has
sometimes appeared mangled due to the newly-default formatter emitting
unportable escape sequences, and sometimes there has been a more
severe output failure due to perldoc making unportable changes to pager
configuration in an attempt to make the escape sequences work.  This is
discussed in [perl #131762].  In the upstream instance of the module
there have been tweaks to the unportable behaviour, but not an actual fix.

In order to make the core distro ship a reliably-working version of
perldoc, this patch customises Pod::Perldoc to implement the obvious
fix for the portability problems.  The fixed version defaults to the
ToText formatter, which produces properly plain text that will go through
any pager.  It never attempts any change to pager configuration.

3 years ago[perl #124349] deprecation warning on hostname() with arguments
Tony Cook [Tue, 17 Nov 2015 00:56:12 +0000 (11:56 +1100)]
[perl #124349] deprecation warning on hostname() with arguments

3 years agoUse preprocessor check for some DEBUG_X_TEST
Nicolas R [Mon, 25 Sep 2017 17:18:28 +0000 (12:18 -0500)]
Use preprocessor check for some DEBUG_X_TEST

Most of the DEBUG_?_TEST calls are already protected
by one '#idef DEBUGGING' check, but noticed a few of them
which were not protected in sv.c and toke.c

We should avoid these extra 'if' statements if perl
is not compiled with debug option: -DDEBUGGING.

3 years ago(perl #131982) update the set/getpriority documentation too
Tony Cook [Wed, 27 Sep 2017 05:27:14 +0000 (15:27 +1000)]
(perl #131982) update the set/getpriority documentation too

3 years ago(perl #131982) provide constants for get/setpriority in POSIX.pm
Tony Cook [Wed, 27 Sep 2017 05:18:17 +0000 (15:18 +1000)]
(perl #131982) provide constants for get/setpriority in POSIX.pm

<sys/time.h> isn't necessarily needed, but apparently increases
portability, since the rusage structure defined in <sys/resource.h>
has struct timeval members.

3 years agoDocument that POSIX::abs, ::alarm, ::atan2, ::chdir, ::chmod need explicit argument.
James E Keenan [Thu, 28 Sep 2017 00:34:09 +0000 (20:34 -0400)]
Document that POSIX::abs, ::alarm, ::atan2, ::chdir, ::chmod need explicit argument.

Reformat code in one foreach block for readability.  Make podcheck happy.

For: RT #132145

3 years agopp.c: delete dead cpp-conditional declaration
Aaron Crane [Sun, 15 Oct 2017 10:01:08 +0000 (12:01 +0200)]
pp.c: delete dead cpp-conditional declaration

This was added in commit dfe9444ca7881e716e9e8feaf20b55da491363ca (February
1998, for Perl 5.004_60) by Andy Dougherty, and its comment says that, even
then, he thought it was unneeded. But the perl5 repo has ever defined the
NEED_GETPID_PROTO cpp symbol that guards this declaration, so this ability
has clearly never been used.

3 years agoRemove out-of-date mention of USA president
Smylers [Thu, 12 Oct 2017 13:18:40 +0000 (14:18 +0100)]
Remove out-of-date mention of USA president

Avoid having to continue to update this every time the president changes
by replacing Barack Obama with Ada Lovelace.

3 years agoRT#131000: splice doesn't honour read-only flag
Aaron Crane [Thu, 16 Mar 2017 12:33:59 +0000 (12:33 +0000)]
RT#131000: splice doesn't honour read-only flag

The push and unshift builtins were correctly throwing a "Modification of a
read-only value attempted" exception when modifying a read-only array, but
splice was silently modifying the array. This commit adds tests that all
three builtins throw such an exception.

One discrepancy between the three remains: push has long silently accepted
a push of no elements onto an array, whereas unshift throws an exception in
that situation. This seems to have been originally a coincidence. The
pp_unshift implementation first makes space for the elements it unshifts
(which croaks for a read-only array), then copies the new values into the
space thus created. The pp_push implementation, on the other hand, calls
av_push() individually on each element; that implicitly croaks, but only one
there's at least one element being pushed.

The pp_push implementation has subsequently been changed: read-only checking
is now done first, but that was done to fix a memory leak. (If the av_push()
itself failed, then the new SV that had been allocated for pushing onto the
array would get leaked.) That leak fix specifically grandfathered in the
acceptance of empty-push-to-readonly-array, to avoid changing behaviour.

I'm not fond of the inconsistency betwen push on the one hand and unshift &
splice on the other, but I'm disinclined to make empty-push-to-readonly
suddenly start throwing an exception after all these years, and it seems
best not to extend that exemption-from-exception to the other builtins.

3 years agoUse library functions instead of manual TAP in t/io/tell.t.
Dmitry Ulanov [Fri, 13 Oct 2017 15:14:49 +0000 (17:14 +0200)]
Use library functions instead of manual TAP in t/io/tell.t.

Committer: minor corrections to descriptions.

Dmitry Ulanov is now a Perl author.

For: RT #132276

3 years agohints/aix.sh: remove trailing spaces
Lukas Mai [Sat, 14 Oct 2017 16:03:17 +0000 (18:03 +0200)]
hints/aix.sh: remove trailing spaces

3 years agominiperl also needs -Ilib for perl.exp on AIX etc
H.Merijn Brand [Sat, 14 Oct 2017 15:01:40 +0000 (17:01 +0200)]
miniperl also needs -Ilib for perl.exp on AIX etc

3 years agoDocument that -I. is now needed with -MTestInit
Dominic Hargreaves [Sat, 14 Oct 2017 14:19:51 +0000 (16:19 +0200)]
Document that -I. is now needed with -MTestInit

3 years agog++ is like gcc on AIX too
H.Merijn Brand [Sat, 14 Oct 2017 13:55:08 +0000 (15:55 +0200)]
g++ is like gcc on AIX too

note that we cannot use g*, as AIX/vac ships gxlc as xlc with gcc-like
option handling

3 years agodelete unused file from SDBM_File
Lukas Mai [Sat, 14 Oct 2017 12:45:41 +0000 (14:45 +0200)]
delete unused file from SDBM_File

3 years agoget rid of "implicit fallthrough" warnings in SDBM_File
Lukas Mai [Sat, 14 Oct 2017 12:42:20 +0000 (14:42 +0200)]
get rid of "implicit fallthrough" warnings in SDBM_File

3 years agoSDBM_File: tabs -> spaces
Lukas Mai [Sat, 14 Oct 2017 12:39:36 +0000 (14:39 +0200)]
SDBM_File: tabs -> spaces

3 years agoDelete dead mktemp() call
Aaron Crane [Sat, 14 Oct 2017 12:56:23 +0000 (14:56 +0200)]
Delete dead mktemp() call

Commit 9c12f1e5a87cce227357eea4b0780c0323f952f0 provided support for Cray
XT4 Catamount/Qk. It added some code with a call to (insecure) mktemp(),
guarded with "#ifdef HAS_MKTEMP". But nothing has ever attempted to define
that preprocessor symbol (including the hints file that was added in that
commit), so this is clearly unused.

3 years agofix comment on GCC_DIAG_IGNORE/GCC_DIAG_RESTORE
Lukas Mai [Sat, 14 Oct 2017 12:24:02 +0000 (14:24 +0200)]
fix comment on GCC_DIAG_IGNORE/GCC_DIAG_RESTORE

3 years agoactually fix "implicit fallthrough" warnings in gcc 7
Lukas Mai [Sat, 14 Oct 2017 12:21:25 +0000 (14:21 +0200)]
actually fix "implicit fallthrough" warnings in gcc 7

3 years agoremove the GCC_DIAG_RESTORE pragma as well
Yves Orton [Sat, 14 Oct 2017 12:06:20 +0000 (14:06 +0200)]
remove the GCC_DIAG_RESTORE pragma as well

3 years agoremove GCC_DIAG_IGNORE from sbox32_hash.h
Yves Orton [Sat, 14 Oct 2017 11:55:53 +0000 (13:55 +0200)]
remove GCC_DIAG_IGNORE from sbox32_hash.h

3 years agofixups for portability
Yves Orton [Sat, 14 Oct 2017 11:42:47 +0000 (13:42 +0200)]
fixups for portability

a) some platforms want #define to start at col 0
b) some compilers want to know whether fallthrough is expected or not.

3 years agoDelete unused files from ext/SDBM_File
Aaron Crane [Sat, 14 Oct 2017 10:42:20 +0000 (12:42 +0200)]
Delete unused files from ext/SDBM_File

3 years agoskip bench.t if valgrind is installed but broken
Lukas Mai [Fri, 13 Oct 2017 21:22:11 +0000 (23:22 +0200)]
skip bench.t if valgrind is installed but broken

3 years agouse standard STMT_START/STMT_END macros in STATIC_ASSERT_STMT
Lukas Mai [Fri, 13 Oct 2017 20:58:40 +0000 (22:58 +0200)]
use standard STMT_START/STMT_END macros in STATIC_ASSERT_STMT

3 years agoavoid gcc 7 warning about * in boolean context
Lukas Mai [Fri, 13 Oct 2017 20:34:25 +0000 (22:34 +0200)]
avoid gcc 7 warning about * in boolean context

3 years agoget rid of "implicit fallthrough" warnings in ext/re/re_exec.c
Lukas Mai [Fri, 13 Oct 2017 20:16:30 +0000 (22:16 +0200)]
get rid of "implicit fallthrough" warnings in ext/re/re_exec.c

3 years agoget rid of "implicit fallthrough" warnings with gcc 7
Lukas Mai [Fri, 13 Oct 2017 17:50:24 +0000 (19:50 +0200)]
get rid of "implicit fallthrough" warnings with gcc 7

3 years agoFix typo in comment about Configure modification
Dominic Hargreaves [Fri, 13 Oct 2017 14:31:03 +0000 (16:31 +0200)]
Fix typo in comment about Configure modification

3 years agoClarify policy on updates to Configure
Dominic Hargreaves [Fri, 13 Oct 2017 14:16:51 +0000 (16:16 +0200)]
Clarify policy on updates to Configure

3 years agoFurther metaconfig updates to Porting/pumpkin.pod
Dominic Hargreaves [Fri, 13 Oct 2017 13:01:32 +0000 (15:01 +0200)]
Further metaconfig updates to Porting/pumpkin.pod

3 years agoAdd Porting/mksample
Dominic Hargreaves [Fri, 13 Oct 2017 12:49:03 +0000 (14:49 +0200)]
Add Porting/mksample

This script was previously in the metaconfig repository, and also dealt
with updating Porting/Glossary, but it wasn't being regularly used.
It does not use any data from that repository, so is better placed
here.

There are some outstanding ambiguities to resolve in when this script
should be run in the documentation, since the generated files
Porting/config.sh and Porting/config_H are also updated by other
means.

3 years agoActually test for memrchr in VMS
H.Merijn Brand [Fri, 13 Oct 2017 13:01:56 +0000 (15:01 +0200)]
Actually test for memrchr in VMS

3 years agoUpdate Config-Perl-V to CPAN version 0.29
Chris 'BinGOs' Williams [Fri, 13 Oct 2017 12:25:29 +0000 (13:25 +0100)]
Update Config-Perl-V to CPAN version 0.29

  [DELTA]

0.29  - 12 Oct 2017, H.Merijn Brand
    * It's 2017
    * Add test for 5.26.1-RC1
    * Fix signature with argument
    * Ignore ccache and compiler path in signature

3 years agoFinish adding memrchr Configure probe
Dagfinn Ilmari Mannsåker [Fri, 13 Oct 2017 12:40:09 +0000 (14:40 +0200)]
Finish adding memrchr Configure probe

Commit 1e436e33 accidentally added the probe to Configure, this finishes
the job by regenerating Glossary, config_h.SH and friends.

3 years agoStrip ccache from $Config{cc}
H.Merijn Brand [Fri, 13 Oct 2017 11:56:47 +0000 (13:56 +0200)]
Strip ccache from $Config{cc}

When building with -Dcc="ccache cc", $Config{cc} would also end up with
this setting. If that perl is then distributed, no XS will be able to
build if the target machine has no ccache installed/available

With this change, the whole build of perl itself will use ccache, but
$Config{cc} will have it stripped, so all modules (lib/cpan/ext) will
not use ccache. A correct installation of ccache however does not need
$CC to include the ccache prefix, so it is essentially a setup problem
on the building host.

3 years agofix comment for upstream meta
H.Merijn Brand [Fri, 13 Oct 2017 09:25:03 +0000 (11:25 +0200)]
fix comment for upstream meta

3 years agoUpdate metaconfig related information
H.Merijn Brand [Fri, 13 Oct 2017 09:20:12 +0000 (11:20 +0200)]
Update metaconfig related information

3 years agoUpgrade Test-Simple from version 1.302100 to 1.302101
Steve Hay [Fri, 13 Oct 2017 07:18:23 +0000 (08:18 +0100)]
Upgrade Test-Simple from version 1.302100 to 1.302101

3 years agohopefully better macros for building on x86
Yves Orton [Thu, 12 Oct 2017 15:57:21 +0000 (17:57 +0200)]
hopefully better macros for building on x86

3 years agoperl no longer builds on MacOS Classic
Tom Hukins [Wed, 11 Oct 2017 19:37:02 +0000 (19:37 +0000)]
perl no longer builds on MacOS Classic

3 years agoUpgrade Test-Simple from vesion 1.302098 to 1.302100
Steve Hay [Thu, 12 Oct 2017 07:22:40 +0000 (08:22 +0100)]
Upgrade Test-Simple from vesion 1.302098 to 1.302100

3 years agoUpgrade Encode from version 2.92 to 2.93
Steve Hay [Tue, 10 Oct 2017 07:30:01 +0000 (08:30 +0100)]
Upgrade Encode from version 2.92 to 2.93

3 years agoEliminate recently introduced compilation warnings.
James E Keenan [Mon, 9 Oct 2017 11:42:02 +0000 (13:42 +0200)]
Eliminate recently introduced compilation warnings.

Warnings like the following started to appear within last 24 hours while
running 'make' on Linux/gcc and FreeBSD-10.3/clang.

op.c: In function ‘Perl_newATTRSUB_x’:
op.c:8593:28: warning: suggest parentheses around ‘&&’ within ‘||’
[-Wparentheses]
        :   IN_PERL_RUNTIME && PL_curstash != CopSTASH(PL_curcop)

3 years agoIncrease B::Xref::VERSION to 1.07
Father Chrysostomos [Mon, 9 Oct 2017 05:23:38 +0000 (22:23 -0700)]
Increase B::Xref::VERSION to 1.07

3 years agoFix B::Xref to handle sub refs
Father Chrysostomos [Mon, 9 Oct 2017 05:22:30 +0000 (22:22 -0700)]
Fix B::Xref to handle sub refs

This only applies to non-threaded builds.  Threaded builds were
already fine.

3 years agoB: Really remove unused var
Father Chrysostomos [Sun, 8 Oct 2017 22:58:01 +0000 (15:58 -0700)]
B: Really remove unused var

6a4fc5265ba102 removed the useless assignment to it, but not the
var itself.

3 years ago[Merge] [perl #129916] Allow sub-in-stash outside of main
Father Chrysostomos [Sun, 8 Oct 2017 21:54:13 +0000 (14:54 -0700)]
[Merge] [perl #129916] Allow sub-in-stash outside of main

The sub-in-stash optimization introduced in 2eaf799e only applied to
subs in the main stash, not in other stashes, due to a problem with
the logic in newATTRSUB.

This branch includes various commits to fix the issue and other prob-
lems that the fix uncovered.

3 years agoMake pp_multideref handle local $::{subref}
Father Chrysostomos [Sun, 24 Sep 2017 23:48:48 +0000 (16:48 -0700)]
Make pp_multideref handle local $::{subref}

Based on a patch by Nicholas R.

3 years ago[perl #129916] Allow sub-in-stash outside of main
Father Chrysostomos [Thu, 21 Sep 2017 14:06:05 +0000 (07:06 -0700)]
[perl #129916] Allow sub-in-stash outside of main

The sub-in-stash optimization introduced in 2eaf799e only applied to
subs in the main stash, not in other stashes, due to a problem with
the logic in newATTRSUB.

This comment:

   Also, we may be called from load_module at run time, so
   PL_curstash (which sets CvSTASH) may not point to the stash the
   sub is stored in.

explains why we need the PL_curstash != CopSTASH(PL_curcop) check.
(Perl_load_module will fail without it.) But that logic does not work
properly at compile time (when PL_curcop == &PL_compiling).

The value of CopSTASH(&PL_compiling) is never actually used.  It is
always set to the main stash.  So if we check that PL_curstash !=
CopSTASH(PL_curcop) and forego the optimization in that case, we will
never optimize subs outside of the main stash.

What we really need is to check IN_PERL_RUNTIME && PL_curstash !=
opSTASH(PL_curcop).  I.e., forego the optimization at run time if the
stashes differ.  That is what this commit implements.

One observable side effect of this change is that deleting a stash
element no longer anonymizes the CV if the CV had no GV that it was
depending on to provide its name.  Since the main thing in such situa-
tions is that we do not get a crash, I think this change (arguably an
improvement) is acceptable.)

-----------

A bit of explanation of various other changes:

gv.c:require_tie_mod needed a bit of help, since it could not handle
sub refs in stashes.

To keep localisation of stash elements working the same way,
local($Stash::{foo}) now upgrades a coderef to a full GV before the
localisation.  (Changes in two pp*.c files and in scope.c:save_gp.)

t/op/stash.t contains a test that makes sure that perl does not crash
when a GV with a CV pointing to it gets deleted.  This commit tweaks
the test so that it continues to test that.  (There has to be a GV for
the test to test what it is meant to test.)

Similarly with t/uni/caller.t and t/uni/stash.t.

op.c:rv2cv_op_cv with the _MAYBE_NAME_GV flag was returning the cal-
ling GV in those cases where a GV-less sub is called via a GV.  E.g.,
*main = \&Foo::foo; main().  This meant that errors like ‘Not enough
arguments’ were giving the wrong sub name.

newATTRSUB was not calling mro_method_changed_in when storing a
sub as an RV.

gv_init needs to arrange for the new GV to have the file and line num-
ber corresponding to the sub in it.  These are taken from CvSTART,
which may be off by a few lines, but is the closest we have to the
place the sub was declared.

3 years agoIncrease $B::Deparse::VERSION to 1.43
Father Chrysostomos [Sun, 24 Sep 2017 21:15:01 +0000 (14:15 -0700)]
Increase $B::Deparse::VERSION to 1.43

3 years agoDeparse: Better constant-dumping heuristics
Father Chrysostomos [Sun, 24 Sep 2017 21:14:00 +0000 (14:14 -0700)]
Deparse: Better constant-dumping heuristics

Constants created via sub foo () { 1 } are stored in the stash as
simple scalar references, under the CV-in-stash optimisation.  That
optimisation currently only applies to the main package, but will
shortly be extended to other packages.  This means B::Deparse’s
heuristics for dumping the constants needs to be improved, to avoid
dumping B::Deparse’s own constants for every program.

The heuristic I am using (since CvFILE is not present on a scalar ref)
is to record whether other subroutines in the same package as the con-
stant are being dumped by virtue of having CvFILE pointing to a file
that is being dumped.  This assumption is that constants and subroutines
in the same package are likely to be in the same file.

3 years agoProvisional version bump for NEXT.pm
Father Chrysostomos [Mon, 11 Sep 2017 04:51:50 +0000 (21:51 -0700)]
Provisional version bump for NEXT.pm

A patch ha- been submitted upstream already, so hopefully this version
number will be short-lived.