Nicholas Clark [Fri, 28 Aug 2009 21:16:31 +0000]
Use & function-call notation to disable prototype checking on dl_load_file()
This causes the test for insufficient arguments to work on Win32 too, where
(for some reason) DynaLoader.xs is built *with* prototypes on the XS code.
On *nix, where it is build by MakeMaker in ext/DynaLoader/, it is built without
prototypes on the XS code, so the check on the number of parameters for
dl_load_file() is purely run-time, which is what the test script used to assume.
Nicholas Clark [Fri, 28 Aug 2009 21:23:35 +0000]
Remove the function mn(), made redundant by runnig the tsts from inside ext/
Craig A. Berry [Fri, 28 Aug 2009 19:18:37 +0000]
Use ../../t/perl rather than ../../perl, as the latter doesn't always exist.
(For example, when perl is built for the VMS debugger.)
Frank Wiegand [Fri, 28 Aug 2009 09:12:59 +0000]
Add ~~ and // to perlcheat.pod
Rafael Garcia-Suarez [Fri, 28 Aug 2009 08:33:31 +0000]
There is no longer an "err" low-prec version of //, so remove it from B::Deparse
Nicholas Clark [Thu, 27 Aug 2009 23:44:15 +0000]
Make extensions in ext run their tests from the extension's own directory.
Inspired by, and in parts borrows from, Schwern's branch on github, but takes a
slightly different approach in places.
Not quite perfect yet - ext/File-Glob still runs from t, at least one FIXME
needs fixing, and the changes to dual-life modules' tests need to be filtered
back upstream, and possibly modified to suit their respective authors.
But it works.
Nicholas Clark [Thu, 27 Aug 2009 19:37:49 +0000]
Don't use t/TEST to test a "read only file", repeatedly chmod-ing it a-w.
David Golden [Thu, 27 Aug 2009 15:41:10 +0000]
Update Module::Build to 0.35
0.35 - Thu Aug 27 09:12:02 EDT 2009
Bug fixes:
- Fix t/destinations.t segfault on 5.6.2
0.34_06 - Sat Aug 22 21:58:26 EDT 2009
Bug fixes:
- Multiple test fixes for OS2 [Ilya Zakharevich]
- Generated.ppd files use :utf8 if possible (RT#48827) [Olivier Mengue]
- Fixed preservation of custom install_paths on resume (RT#41166)
[David Golden]
- Warn instead of crashing when Pod::Man tries to create files with
colons on vfat partitions on unix (RT#45544) [David Golden]
0.34_05 - Sun Aug 9 22:31:37 EDT 2009
Bug fixes:
- When auto_configure_requires is true (the default), Module::Build will
only add last 'major' version of Module:Build (e.g. 0.XX) to
configure_requires to avoid specifying a minor development release not
available on CPAN [David Golden]
0.34_04 - Sat Aug 8 11:02:24 EDT 2009
Other:
- Added documentation warning that 'get_options' should be capitalized
to avoid conflicting with future Module::Build options and changed
the examples accordingly.
0.34_03 - Sat Aug 8 07:39:16 EDT 2009
Bug fixes:
- Fixed failing xs.t if /tmp is mounted noexec (RT#47331) [David Golden]
- Fixed failing debug.t on VMS (RT#48362) [Craig Berry]
- Prevent par.t from dying on error in .zip extraction [David Golden]
- Fixed potential runthrough.t failure on 5.6.2 [David Golden]
Other:
- Archive::Tar changed from 'requires' to 'recommends' so non-authors
without IO::Zlib can still use Module::Build to install modules
[reported by Matt Trout, fix by David Golden]
Nicholas Clark [Thu, 27 Aug 2009 13:26:20 +0000]
t/harness and t/TEST can both use the same _run_test() to invoke tests.
Now the logic to invoke tests is in exactly one place, we can refactor it with
impunity.
Nicholas Clark [Thu, 27 Aug 2009 12:23:38 +0000]
Share common code in t/TEST and t/harness, by having harness require ./TEST
The logical way to do this would be to have the common code in a file both
require or use. However, t/TEST needs to still work, to generate test results,
even if require isn't working, so we cannot do that. t/harness has no such
restriction, so it is quite acceptable to have it require t/TEST.
Vincent Pit [Thu, 27 Aug 2009 09:13:09 +0000]
Speed up repeatcpy() by at least 40% for 1-char or numerous repeats
And don't make it receive the interpreter anymore.
For 1-char repeats, use memset(). Otherwise, use the old implementation up
to some (small) length, and then use memcpy() in a binary manner, based on
what we previously copied.
Note that we use memcpy() so both strings shouldn't overlap. The previous
implementation didn't allow this as well. This would be a good place to use
the restrict keyword from C99. I'm not sure if Configure has a probe for it.
Vincent Pit [Thu, 27 Aug 2009 08:38:34 +0000]
delimcopy(), ibcmp(), ibcmp_locale(), instr(), ninstr() and rninstr() from util.c don't need the interpreter as well
Reini Urban [Thu, 27 Aug 2009 08:19:26 +0000]
B - The Perl Compiler (wrong name)
The Perl Compilers are for some time in B::C and not B. B is only the
backend.
Vincent Pit [Wed, 26 Aug 2009 22:11:38 +0000]
Don't pass the the interpreter to is_ascii_string(), is_utf8_char(), is_utf8_string(), is_utf8_string_loclen() as they don't need it
Chip Salzenberg [Wed, 26 Aug 2009 22:17:55 +0000]
fix threaded build after
8ff3e83b
Chip Salzenberg [Wed, 26 Aug 2009 21:33:15 +0000]
In C<use utf8; a=>'b'>, do not set utf8 flag on 'a' [perl #68812]
Chip Salzenberg [Wed, 26 Aug 2009 21:05:15 +0000]
set utf8 bit on inferred method names when C<use utf8>
Chip Salzenberg [Wed, 26 Aug 2009 20:26:49 +0000]
Merge branch 'blead' of ssh://perl5.git.perl.org/perl into blead
Chip Salzenberg [Wed, 26 Aug 2009 20:25:52 +0000]
New debugging flag -DB now dumps subroutine definitions,
leaving -Dx for its original purpose of dumping syntax trees.
Nicholas Clark [Wed, 26 Aug 2009 19:55:36 +0000]
No need to use File::Spec, as all names remain internal to perl.
(This will make merging the code in t/TEST and t/harness easier.)
Nicholas Clark [Wed, 26 Aug 2009 19:43:11 +0000]
Break out the code that finds tests in MANIFEST into _tests_from_manifest().
Nicholas Clark [Wed, 26 Aug 2009 18:39:20 +0000]
%OVER was removed in October 2001 with
6ff868ee and 73ddec28.
Nicholas Clark [Wed, 26 Aug 2009 18:06:55 +0000]
t/TEST shouldn't use -M options until we've tested that they work.
Also, t/TEST doesn't need to set $ENV{PERL_CORE} as t/TestInit.pm does this.
t/TestInit.pm should also add '.' to @INC (except under tainting) to correctly
emulate the include path of this perl binary once it and its libraries are
installed.
Fix t/run/cloexec.t to not rely on -I../lib being passed as a command-line
arguments. t/run/*.t *test* that command line arguments work, so they can't
be called with command line arguments that are mandatory for them to work.
Nicholas Clark [Wed, 26 Aug 2009 17:42:29 +0000]
Ensure $ENV{PERL_CORE} is always set, but correctly tainted.
Hopefully a better solution than
a19106168e3a7e2a. (change 19964)
Rafael Garcia-Suarez [Wed, 26 Aug 2009 13:44:44 +0000]
Cargo-cult addition of op_say to the MAD code
(and make some more TODO tests pass)
Rafael Garcia-Suarez [Wed, 26 Aug 2009 13:39:03 +0000]
This MAD test now passes here, so unmark it as TODO
Rafael Garcia-Suarez [Wed, 26 Aug 2009 13:30:35 +0000]
Make MAD understand the "..." operator
Rafael Garcia-Suarez [Wed, 26 Aug 2009 12:32:22 +0000]
Use less constness in MAD code
This suppresses warnings, because sometimes the constness was taken away
Rafael Garcia-Suarez [Wed, 26 Aug 2009 09:56:00 +0000]
Note the change to Socket in perldelta.
Rafael Garcia-Suarez [Wed, 26 Aug 2009 09:53:15 +0000]
Bump version of Socket.pm
Lubomir Rintel [Tue, 25 Aug 2009 15:09:53 +0000]
Add support for Abstract namespace sockets
Abstract namespace sockets are Linux-specific socket type that live in
AF_UNIX family, slightly abusing it to be able to use arbitrary
character arrays as addresses: They start with nul byte and are not
terminated by nul byte, but with the length passed to the socket()
system call.
Added regression test for the correct address length computation.
Signed-off-by: Lubomir Rintel <lkundrak@fedoraproject.org>
John P. Linderman [Wed, 26 Aug 2009 07:14:13 +0000]
Improve perlhack instructions about gprof
Nicholas Clark [Tue, 25 Aug 2009 21:06:28 +0000]
abd39864 broke t/harness when $ENV{TEST_JOBS} was not set. Oops. Fixed.
Michael G. Schwern [Wed, 4 Mar 2009 04:42:55 +0000]
Refactor t/harness to always use TAP::Harness.
No reason to screw around with half Test::Harness, half TAP::Harness.
This normalization will make it easier to adapt the ext/Module tests.
[Edited to remove the code related to the fork option, which was dropped from
TAP::Harness with version 3.17. The now-unknown fork argument to a constructor
generates an error.]
Michael G. Schwern [Tue, 3 Mar 2009 00:12:33 +0000]
Document the purpose of TestInit
[Ammended with a grammar fix]
Michael G. Schwern [Tue, 3 Mar 2009 00:11:25 +0000]
Refactoring to consoldate some hard coded paths.
I was going to do more extensive work here but it wasn't necessary. Left
the refactorings in.
Michael G. Schwern [Tue, 3 Mar 2009 00:09:21 +0000]
There's no reason to load TestInit in TEST.
Since TestInit executes code on loading can be actively bad.
Michael G. Schwern [Tue, 3 Mar 2009 00:08:38 +0000]
Quote "strict"->import and "warnings"->import to quiet warnings.
[Ammended slightly to invert the change in
3fd4b35989]
Michael G. Schwern [Mon, 2 Mar 2009 01:03:55 +0000]
Refactor the code to run the test into _run_test()
Also turn $valgrind_log into $Valgrind_Log, because it's really a global config variable.
Michael G. Schwern [Mon, 2 Mar 2009 00:55:01 +0000]
Refactoring to move the code to read the test for special options into its own function.
Get the hell out of the way so I can read this mess.
[ammended slightly by Nicholas Clark to keep require strict commented out]
Michael G. Schwern [Mon, 2 Mar 2009 00:40:59 +0000]
This is reading the first line of the test, not running it.
Nicholas Clark [Tue, 25 Aug 2009 14:13:57 +0000]
Purge 3 items of documentation change, which are insufficiently noteworthy.
Nicholas Clark [Tue, 25 Aug 2009 14:10:31 +0000]
Syle nit - we're not marking things (XS).
Nicholas Clark [Tue, 25 Aug 2009 14:03:57 +0000]
Tidy up some more Pod nits.
Nicholas Clark [Tue, 25 Aug 2009 09:38:41 +0000]
Run t/run/*.t before t/cmd/*.t, making the order base, cmd, run, and note why.
base first, as TEST bails out if that can't run
then comp, to validate that require works
then run, to validate that -M works
then we know we can -MTestInit for everything else, making life simpler)
Nicholas Clark [Tue, 25 Aug 2009 08:31:44 +0000]
Remove File::Spec from t/TEST - VMS is fine without it. Win32 doesn't use it.
Also remove one complex alternation regexp, and replace it with a hash lookup.
At this stage do we trust the regexp engine that much?
[Really TEST shouldn't be relying on require working, and hence being able to
pull in modules to do essential parts of its work. And modules that rely on
OO dispatch - :-(]
Bo Borgerson [Sun, 16 Aug 2009 15:07:39 +0000]
String with NULL auto-increment bug fix
Check position relative to end of string length rather than whether
char is NULL at end of initial alphanumerics in Perl_sv_inc
Rafael Garcia-Suarez [Tue, 25 Aug 2009 07:36:42 +0000]
Bump Safe version to 2.19
Rafael Garcia-Suarez [Tue, 25 Aug 2009 07:34:18 +0000]
Merge branch 'blead' into dual/Safe
Jesse Vincent [Mon, 24 Aug 2009 20:34:29 +0000]
Added two removed platforms to the Perl 5.11.0 delta
Nicholas Clark [Mon, 24 Aug 2009 17:32:14 +0000]
Add an encoding line, as the file is in UTF-8.
Nicholas Clark [Mon, 24 Aug 2009 17:30:38 +0000]
Pod copy editing - whitespace cleanup, and F<>/C<> corrections.
Nicholas Clark [Mon, 24 Aug 2009 17:14:49 +0000]
Note that File::Copy is actually at 2.16, and that cp needs describing.
Nicholas Clark [Mon, 24 Aug 2009 17:11:02 +0000]
constant has been upgraded to 1.19. Describe the improvements.
Nicholas Clark [Mon, 24 Aug 2009 17:07:47 +0000]
The attrs pragma has been removed, not upgraded.
Nicholas Clark [Mon, 24 Aug 2009 17:00:27 +0000]
Correct some 5.10.1s that should be 5.11.0, and one 5.10.2 to 5.11.1.
Nicholas Clark [Mon, 24 Aug 2009 16:59:15 +0000]
Note that the smartmatch changes in 5.11.0 are the same as in 5.10.1.
Jesse Vincent [Mon, 24 Aug 2009 16:09:21 +0000]
Merge branch 'blead' of ssh://perl5.git.perl.org/gitroot/perl into blead
* 'blead' of ssh://perl5.git.perl.org/gitroot/perl:
Note that linearising C3 MRO is now 40% faster for single inheritance.
Jesse Vincent [Mon, 24 Aug 2009 16:08:44 +0000]
More work toward a usable 5.11.0 delta file
Nicholas Clark [Mon, 24 Aug 2009 15:50:23 +0000]
Note that linearising C3 MRO is now 40% faster for single inheritance.
Jesse Vincent [Mon, 24 Aug 2009 15:35:55 +0000]
Merge branch 'blead' of ssh://perl5.git.perl.org/gitroot/perl into blead
* 'blead' of ssh://perl5.git.perl.org/gitroot/perl:
Updated CPANPLUS::Dist::Build to CPAN version 0.38
Jesse Vincent [Mon, 24 Aug 2009 15:34:41 +0000]
Forward ported perl 5.10.1 changes into perl5110delta.pod per nicholas
Chris 'BinGOs' Williams [Mon, 24 Aug 2009 15:16:52 +0000]
Updated CPANPLUS::Dist::Build to CPAN version 0.38
Hi,
Attached is a patch to update the CPANPLUS::Dist::Build in core
to the CPAN version 0.38
Highlights from the Changelog:
==============================
0.38 Mon Aug 24 13:59:11 BST 2009
- Cosmetic change to the test output message so it
matches CPANPLUS::Dist::MM (bingos)
==============================
Many thanks,
--
Chris Williams
aka BinGOs
PGP ID 0x
4658671F
http://www.gumbynet.org.uk
==========================
From
2c3e6d8f82c08a5817d74a8407ebcdbbd86ebb15 Mon Sep 17 00:00:00 2001
From: Chris Williams <chris@bingosnet.co.uk>
Date: Mon, 24 Aug 2009 16:09:47 +0100
Subject: [PATCH] Updated CPANPLUS::Dist::Build to CPAN version 0.38
Signed-off-by: H.Merijn Brand <h.m.brand@xs4all.nl>
Nicholas Clark [Mon, 24 Aug 2009 13:17:11 +0000]
Fix a Deparse bug - constants and PCSs were appearing as subroutine stubs.
(Because constants and other Proxy Constant Subroutines are stored in the
symbol table as references, and in 5.11.0 RVs merged with IVs, and B::Deparse
knows that an IV in the symbol table is a stub declaration for a subroutine
with no prototype, so B::Deparse "knew" that that's what it was.)
Nicholas Clark [Mon, 24 Aug 2009 12:37:23 +0000]
Remove the MacOS Classic support from deparse.t (as it's getting in the way).
Rafael Garcia-Suarez [Mon, 24 Aug 2009 08:24:45 +0000]
Merge commit 'shlomif/blead' into blead
Leon Brocard [Mon, 24 Aug 2009 08:22:44 +0000]
Fix URL to Unicode 5.1.0 notable changes
Tom Hukins [Mon, 24 Aug 2009 08:01:51 +0000]
Minor improvements to perldebug.pod
Minor improvements to the section on profiling.
Remove suggestions that Devel::DProf might be the most popular
profiler and reference perlperf.pod.
Alan Grover [Mon, 24 Aug 2009 07:38:31 +0000]
[perl #68708] Add "CARP_NOT" section to Carp's POD
I Carp's POD, @CARP_NOT is mentioned but it's use is not clear. It is
particularly not clear that @CARP_NOT is a variable in _your_ package.
I propose adding a section to the POD, before the %Carp::Internal
section.
David Mitchell [Sun, 23 Aug 2009 16:41:42 +0000]
release_managers_guide: lots of tweaks to post-release tasks
David Mitchell [Sun, 23 Aug 2009 16:36:42 +0000]
add 5.10.1 to perlhist in blead
David Mitchell [Sun, 23 Aug 2009 16:33:11 +0000]
add perl5101delta
David Mitchell [Sun, 23 Aug 2009 15:17:46 +0000]
add Porting/perldelta_template,
a skeleton outline of a perldelta file
Rafael Garcia-Suarez [Sun, 23 Aug 2009 13:28:22 +0000]
Fix parsing of readline(FH) [perl #68458]
This was broken by commit
984f9f66477bc722578262ae8520584e44e881af,
which was committed to solve bug [perl #53806].
Basically, to make everyone happy, we need only to enforce strictures
on barewords that follow the specially parsed keywords "print" and
"exec" (and similar) -- except when they have been already parsed.
David Mitchell [Sat, 22 Aug 2009 22:17:57 +0000]
more release_managers_guide tweaks
David Mitchell [Sat, 22 Aug 2009 21:51:15 +0000]
Update Module::CoreList for 5.10.1
Nicholas Clark [Sat, 18 Jul 2009 18:32:50 +0000]
Eliminate *_ALLOCATED_HEAD and *_HEAD macros which are now used only once.
Nicholas Clark [Sat, 18 Jul 2009 17:34:06 +0000]
Eliminate the remaining definitions for *_allocated structs.
David Mitchell [Sat, 22 Aug 2009 19:45:09 +0000]
mention perlivp in release_managers_guide
David Mitchell [Sat, 22 Aug 2009 19:43:00 +0000]
better document smart match overloading
David Mitchell [Sat, 22 Aug 2009 17:16:24 +0000]
re-apply boilerplate to CGI test files
David Mitchell [Sat, 22 Aug 2009 17:02:13 +0000]
upgrade GCI from 3.43 to 3.45
Note that this blows away some local mods:
* b<> to B<> fixes
* core boilerplate for upload.t and uploadInfo.t
David Mitchell [Sat, 22 Aug 2009 16:32:53 +0000]
add the CGI "default_value for popup_menu()" fix from 3.45
Nicholas Clark [Sat, 22 Aug 2009 17:08:06 +0000]
In pp_entersub, replace gv_fetchpv() with gv_fetchpvs().
Nicholas Clark [Sat, 22 Aug 2009 16:28:25 +0000]
In Perl_newATTRSUB(), refactor the calls to gv_fetch*(), reducing if blocks.
Nicholas Clark [Sat, 22 Aug 2009 15:28:32 +0000]
In Perl_newATTRSUB(), we know the lengths, so can avoid gv_fetchpv().
Brought to you by the Campaign for the Elimination of strlen().
Nicholas Clark [Fri, 21 Aug 2009 21:11:47 +0000]
gv_efullname3() could return NULL, so mro::_nextcan() must cope (and croak()).
Nicholas Clark [Fri, 21 Aug 2009 20:50:26 +0000]
sv_newmortal() is faster than sv_2mortal(newSV(0))
Nicholas Clark [Fri, 21 Aug 2009 20:28:18 +0000]
Eliminate SV leak in mro::_nextcan - hv_store_ent()'s "key" doesn't take a ref.
Jerry D. Hedden [Fri, 21 Aug 2009 14:45:22 +0000]
Fix build warning in mro.c
Zefram [Thu, 20 Aug 2009 23:49:14 +0000]
Add clear magic to %^H so that the HE chain is reset when you empty it.
This fixes [perl #68590] : %^H not lexical enough.
Nicholas Clark [Fri, 21 Aug 2009 10:04:06 +0000]
Use the constant _CAN_PCS to optimise out things that can't happen pre 5.10
Nicholas Clark [Fri, 21 Aug 2009 08:53:47 +0000]
Remove a refactoring typo - don't post-increment $flush_mro at the sub's end.
Nicholas Clark [Fri, 21 Aug 2009 08:40:03 +0000]
In mro_get_linear_isa_c3() optimise even if AvARRAY(isa_lin) is NULL.
It can only be NULL if AvFILLp(isa_lin) is -1, and the code in the fast path
block will work cleanly in that case too.
Chip Salzenberg [Thu, 20 Aug 2009 21:37:01 +0000]
incremental improvement in ~~ documentation, already in maint-5.10
Nicholas Clark [Thu, 20 Aug 2009 20:12:44 +0000]
Instead of trusting mro::get_linear_isa(), test it against the expected output.
Nicholas Clark [Thu, 20 Aug 2009 20:09:03 +0000]
Optimise mro_get_linear_isa_c3() when there is a single parent. 40% speed up.
Idea blatantly copied from chromatic's analogous change to parrot, r38477.
David Mitchell [Thu, 20 Aug 2009 18:29:35 +0000]
add -DM flag to track smartmatch resolution
Nicholas Clark [Thu, 20 Aug 2009 15:02:40 +0000]
Optimise S_mro_get_linear_isa_dfs() when dealing with the first parent class.
Benchmarking with single inheritance suggests that this is 10% faster.
Nicholas Clark [Thu, 20 Aug 2009 14:56:18 +0000]
Perl_newHVhv() should copy immortal values as-is, such as PL_sv_undef
Currently it calls newSVsv() always, which copies the value, but the immortal
SVs are used as much for their addresses as their values. You can't get the
immortals into HVs from Perl-space, except for PL_sv_placeholder, and any hash
with those will take the else block, where the call to Perl_hv_iternext_flags()
won't be returning placeholders anyway. Hence If XS code has gone to the
trouble to get the "impossible" in there, they had a reason for it.
I am assuming that Perl_hv_copy_hints_hv() should stay as-is, as it is
documented that only strings and integers are supported values for %^H.