]>
git.vpit.fr Git - perl/modules/Variable-Magic.git/log
Vincent Pit [Wed, 26 Oct 2011 15:57:32 +0000 (17:57 +0200)]
Reorder some code
Vincent Pit [Wed, 26 Oct 2011 15:50:13 +0000 (17:50 +0200)]
Rename vmg_wizard_* to vmg_wizard_sv_* and {MGWIZ,mgwiz}_* to vmg_wizard_*
Vincent Pit [Wed, 26 Oct 2011 14:52:27 +0000 (16:52 +0200)]
Clone wizard objects through dup magic
ptable.h is no longer needed.
Vincent Pit [Wed, 26 Oct 2011 14:12:20 +0000 (16:12 +0200)]
Clean up threads tests a bit
Vincent Pit [Wed, 26 Oct 2011 13:37:55 +0000 (15:37 +0200)]
Share the vtables with threaded perls
When a magical variable is cloned into a new thread, the associated vtables
aren't cloned by mg_dup(), but simply shared. This means that if a wizard
is casted upon a variable in the main interpreter, then a thread is created,
then the wizard is destroyed, and then the cloned variant of the magical
variable is manipulated, perl can no longer access the vtable (it was freed
when the wizard was destroyed) and memory misreads happen.
To solve this, the only solution is to make the vtables associated with
wizards shared and reference-counted.
Vincent Pit [Wed, 26 Oct 2011 10:02:11 +0000 (12:02 +0200)]
Wrap MUTEX_{LOCK,UNLOCK} in macros so can we don't have to #ifdef them
Vincent Pit [Wed, 26 Oct 2011 09:40:16 +0000 (11:40 +0200)]
MGf_COPY and MGf_DUP should always be set for the perls we consider
Vincent Pit [Tue, 25 Oct 2011 22:32:18 +0000 (00:32 +0200)]
Abstract the MGWIZ creation into a new vmg_mgwiz_alloc() function
Vincent Pit [Tue, 25 Oct 2011 22:25:38 +0000 (00:25 +0200)]
Pass the 'data' callback after the 'op_info' flag when calling _wizard()
Vincent Pit [Mon, 24 Oct 2011 19:57:55 +0000 (21:57 +0200)]
Threads tests may not be able to spawn all the threads
To work around this :
- move all the threads boilerplate to a new helper test module ;
- capture the warnings/errors that threads->create may throw and reprint
them properly as diagnostics ;
- don't set a plan in threads tests.
Vincent Pit [Mon, 24 Oct 2011 19:42:20 +0000 (21:42 +0200)]
Remove trailing whitespace
Vincent Pit [Sun, 23 Oct 2011 15:31:20 +0000 (17:31 +0200)]
Make sure code examples fit in a 80 columns terminal
Vincent Pit [Sun, 23 Oct 2011 14:55:07 +0000 (16:55 +0200)]
Preserve the PATH environment variable when running a sub-perl on cygwin
Vincent Pit [Mon, 5 Sep 2011 17:46:40 +0000 (19:46 +0200)]
Silence an "unused" compiler warning
Vincent Pit [Sun, 4 Sep 2011 21:32:06 +0000 (23:32 +0200)]
Tweak some optional module loading in tests
Vincent Pit [Wed, 24 Aug 2011 17:42:12 +0000 (19:42 +0200)]
Add META.json
Vincent Pit [Wed, 24 Aug 2011 12:52:54 +0000 (14:52 +0200)]
Encode the README file in UTF-8
Vincent Pit [Sun, 7 Aug 2011 16:17:22 +0000 (18:17 +0200)]
Replace $] by "$]"
An old bug could cause $] to be initialized to a wrong decimal value.
Since its stringified value was always right, we use it instead.
Occurences of $^V were also replaced by $].
Vincent Pit [Sun, 7 Aug 2011 16:05:30 +0000 (18:05 +0200)]
Ignore MYMETA.json
Vincent Pit [Sun, 7 Aug 2011 16:04:39 +0000 (18:04 +0200)]
Update vmg_opclass() for perl 5.15
Even if magic is no longer called by the aelem(fast).
Vincent Pit [Sun, 23 Jan 2011 16:44:38 +0000 (17:44 +0100)]
This is 0.46
Vincent Pit [Wed, 19 Jan 2011 17:45:36 +0000 (18:45 +0100)]
Switch to qw<>
Vincent Pit [Wed, 19 Jan 2011 17:29:15 +0000 (18:29 +0100)]
Update magical flags after dispelling magic
This fixes RT #64866.
Vincent Pit [Tue, 18 Jan 2011 12:33:36 +0000 (13:33 +0100)]
Bump copyright year
Vincent Pit [Tue, 18 Jan 2011 12:33:16 +0000 (13:33 +0100)]
Tighten the scope of two variables unused on 5.8
Vincent Pit [Sun, 21 Nov 2010 23:10:30 +0000 (00:10 +0100)]
This is 0.45
Vincent Pit [Sun, 21 Nov 2010 22:56:26 +0000 (23:56 +0100)]
A terser way to display versions for modules used in tests
Vincent Pit [Sun, 21 Nov 2010 22:43:16 +0000 (23:43 +0100)]
Only use Capture::Tiny when it is working on a simple example
Vincent Pit [Sun, 21 Nov 2010 13:43:04 +0000 (14:43 +0100)]
A bare glob that isn't part of a fetch doesn't call get magic anymore in 5.13.7
Vincent Pit [Sun, 21 Nov 2010 13:42:06 +0000 (14:42 +0100)]
Add more glob tests
Vincent Pit [Sun, 21 Nov 2010 13:23:09 +0000 (14:23 +0100)]
Wording fix
Vincent Pit [Thu, 28 Oct 2010 22:20:04 +0000 (00:20 +0200)]
Give a saner default to pPTBLMS
This silences a compiler warning.
Vincent Pit [Thu, 28 Oct 2010 22:19:32 +0000 (00:19 +0200)]
Skip NULL values when walking a table
ptable_fetch already returns NULL for nonexistent entries, so NULL can't
possibly have any other meaning than "this node is not currently in use".
Vincent Pit [Thu, 28 Oct 2010 22:18:08 +0000 (00:18 +0200)]
Add ptable_delete to ptable.h
Vincent Pit [Sat, 16 Oct 2010 12:34:04 +0000 (14:34 +0200)]
Explicitely state that C++ compilers aren't supported
Vincent Pit [Sun, 3 Oct 2010 00:22:17 +0000 (02:22 +0200)]
Ignore MYMETA.yml
Peter Rabbitson [Sat, 25 Sep 2010 18:18:36 +0000 (20:18 +0200)]
Clarify minimum perl 5.10.0 requirement for hash magic
Vincent Pit [Fri, 24 Sep 2010 19:10:06 +0000 (21:10 +0200)]
This is 0.44
Vincent Pit [Fri, 24 Sep 2010 18:00:03 +0000 (20:00 +0200)]
Correctly propagate exceptions from _wizard()
Vincent Pit [Fri, 24 Sep 2010 17:59:24 +0000 (19:59 +0200)]
Only load Carp.pm when throwing an error
Vincent Pit [Fri, 24 Sep 2010 17:08:21 +0000 (19:08 +0200)]
Describe why we don't have to guard against PL_dirty in vmg_mgwiz_free()
Vincent Pit [Fri, 24 Sep 2010 16:10:23 +0000 (18:10 +0200)]
Force threads tests with the PERL_VARIABLE_MAGIC_TEST_THREADS environment variable
Vincent Pit [Fri, 24 Sep 2010 15:58:29 +0000 (17:58 +0200)]
Skip threads tests unless perl version is 5.13.4 or greater
There was a long standing bug in the handling of GV <-> CV double linkage
that could (and explicitely did with a poisonous perl) cause segfaults at
thread destruction. It got fixed by Dave in commit
803f274 which went in
5.13.3, but the fix was amended for 5.13.4 in commit
09aad8f . Since it's
not really fair for the user to not be able to install the module because
of this, we skip the threads tests unless perl is at least 5.13.4.
Vincent Pit [Thu, 23 Sep 2010 21:58:40 +0000 (23:58 +0200)]
Remove the patchlevel test
Commit
d29a774d106d145036eda849ab57f42410100008 removed the only place it
was used, and perl left perforce almost three years ago.
Vincent Pit [Thu, 23 Sep 2010 21:55:16 +0000 (23:55 +0200)]
Remove the test for non-released perl 5.11.0
Who would be crazy enough to still use them? And the test seems to break
with some git checkouts.
Vincent Pit [Thu, 23 Sep 2010 16:52:42 +0000 (18:52 +0200)]
Force linking against the perl dll when using gcc 3.4 on Windows
When a Windows perl links an XS shared object to an import library using
gcc/g++ version 3.4.x, strange breakage happens at load time because the
mutex-related symbols seems not to be reachable. This situation happens
especially for AS perl and the default compiler suite it installs.
We fix this by forcing g++ to link against the dll instead, which should
be in the same directory as the perl executable (at least for AS perl).
This (should) fix RT #51483.
Vincent Pit [Thu, 23 Sep 2010 14:45:12 +0000 (16:45 +0200)]
Clarify some logic in Makefile.PL
Vincent Pit [Thu, 19 Aug 2010 10:47:16 +0000 (12:47 +0200)]
SvREFCNT_dec already checks if the SV is non null
Vincent Pit [Fri, 25 Jun 2010 23:35:32 +0000 (01:35 +0200)]
This is 0.43
Vincent Pit [Fri, 25 Jun 2010 23:10:43 +0000 (01:10 +0200)]
More control tests
Vincent Pit [Fri, 25 Jun 2010 20:31:57 +0000 (22:31 +0200)]
Prettify the constants export list
Vincent Pit [Fri, 25 Jun 2010 20:30:18 +0000 (22:30 +0200)]
5.13.2 calls get magic on globs
Add the new compatibility macro VMG_COMPAT_GLOB_GET to reflect this change.
Vincent Pit [Fri, 25 Jun 2010 20:16:50 +0000 (22:16 +0200)]
Fix CLONE_PARAMS struct initialization for perl 5.13.2
Vincent Pit [Fri, 25 Jun 2010 20:05:22 +0000 (22:05 +0200)]
Always use a safe version of call_sv()
All the relevant code has been factored out of vmg_cb_free() into a new
vmg_call_sv() function.
Vincent Pit [Fri, 25 Jun 2010 18:34:14 +0000 (20:34 +0200)]
Exception propagation fixes
First, make sure that exceptions thrown in a free callback at the end of a
regular block are propagated.
Second, prevent duplication of error messages when dieing in a free callback
at BEGIN time.
Vincent Pit [Wed, 19 May 2010 00:13:23 +0000 (02:13 +0200)]
This is 0.42
Vincent Pit [Tue, 18 May 2010 23:46:47 +0000 (01:46 +0200)]
Fix tests for perl 5.13
Vincent Pit [Tue, 18 May 2010 22:27:59 +0000 (00:27 +0200)]
Fix t/18-opinfo.t failures with perl 5.13
Vincent Pit [Sat, 24 Apr 2010 21:46:47 +0000 (23:46 +0200)]
Complete coverage of the len callback
Vincent Pit [Sat, 24 Apr 2010 21:01:49 +0000 (23:01 +0200)]
Improve op_info coverage
Vincent Pit [Mon, 15 Mar 2010 17:28:47 +0000 (18:28 +0100)]
This is 0.41
Vincent Pit [Mon, 15 Mar 2010 17:22:47 +0000 (18:22 +0100)]
Compatibility fix for 5.12.0
Shlomi Fish [Sat, 6 Mar 2010 00:32:39 +0000 (01:32 +0100)]
A few doc corrections
Vincent Pit [Tue, 23 Feb 2010 13:42:19 +0000 (14:42 +0100)]
Tune the userdata example
Vincent Pit [Wed, 13 Jan 2010 01:07:57 +0000 (02:07 +0100)]
Correctly propagate the errors thrown when variable destruction happens at compile-time
Vincent Pit [Wed, 6 Jan 2010 23:16:43 +0000 (00:16 +0100)]
This is 0.40
Vincent Pit [Wed, 6 Jan 2010 22:58:37 +0000 (23:58 +0100)]
Indentation
Vincent Pit [Wed, 6 Jan 2010 22:57:03 +0000 (23:57 +0100)]
Bump copyright year
Vincent Pit [Wed, 6 Jan 2010 22:54:49 +0000 (23:54 +0100)]
Use SvREFCNT_inc_simple_void() whenever possible
Vincent Pit [Wed, 6 Jan 2010 22:48:11 +0000 (23:48 +0100)]
Use SvRV_const() whenever possible
Vincent Pit [Mon, 28 Dec 2009 23:01:25 +0000 (00:01 +0100)]
SvPV_const is only available since 5.9.3
Vincent Pit [Mon, 28 Dec 2009 22:56:45 +0000 (23:56 +0100)]
Cast the op_info value before applying bit shifts
Also move the opinfo member before the uvar one in the MGWIZ struct, as
it's used more frequently.
Vincent Pit [Mon, 28 Dec 2009 22:13:22 +0000 (23:13 +0100)]
Use size_t as the index type in the pointer table
Vincent Pit [Mon, 28 Dec 2009 21:40:08 +0000 (22:40 +0100)]
Fix expected constness for utf8_length() arguments
Vincent Pit [Mon, 28 Dec 2009 18:59:39 +0000 (19:59 +0100)]
Fix type coercions on values returned from callbacks
Vincent Pit [Mon, 28 Dec 2009 18:35:07 +0000 (19:35 +0100)]
call_sv() ought to take an I32
Vincent Pit [Mon, 28 Dec 2009 18:25:11 +0000 (19:25 +0100)]
Explicitely cap the op_info parameter to 255
Vincent Pit [Mon, 28 Dec 2009 18:19:31 +0000 (19:19 +0100)]
Wizard ids are IV, not UV
Vincent Pit [Mon, 28 Dec 2009 18:12:46 +0000 (19:12 +0100)]
Fix memory miswrite when passing data arguments to cast()
Vincent Pit [Tue, 1 Dec 2009 00:00:22 +0000 (01:00 +0100)]
This is 0.39
Vincent Pit [Mon, 30 Nov 2009 23:59:21 +0000 (00:59 +0100)]
Document passing a string reference as the callback
Vincent Pit [Mon, 30 Nov 2009 21:14:16 +0000 (22:14 +0100)]
Fix compiler warnings with the Intel CC
Vincent Pit [Mon, 30 Nov 2009 20:57:24 +0000 (21:57 +0100)]
Test string callbacks
Vincent Pit [Mon, 30 Nov 2009 20:45:24 +0000 (21:45 +0100)]
Test destruction of magic callbacks
Vincent Pit [Mon, 30 Nov 2009 20:22:10 +0000 (21:22 +0100)]
Also test getdata() in t/80-leaks.t
Vincent Pit [Mon, 30 Nov 2009 20:18:23 +0000 (21:18 +0100)]
Stop leaking objects stored in the data slot
Vincent Pit [Mon, 30 Nov 2009 20:02:04 +0000 (21:02 +0100)]
Stop passing arguments to the data callback through an AV
Vincent Pit [Mon, 30 Nov 2009 17:30:56 +0000 (18:30 +0100)]
Fix t/35-stash.t for 5.11.2
And make it more robust for other versions.
Vincent Pit [Mon, 30 Nov 2009 00:16:18 +0000 (01:16 +0100)]
Directly store the CVs into the MGWIZ structure
Instead of references to CVs.
Vincent Pit [Mon, 30 Nov 2009 00:12:05 +0000 (01:12 +0100)]
Free the thread local context from an atexit callback
They happen very late in the destruction process but that's not an issue
there. The MGWIZ structure and the SVs they contain are destroyed and
removed from the global pointer table when the wizard object is destroyed.
At the time the atexit callback fires, the pointer table is then empty and
the only thing left to do is to free its own (shared) memory.
Vincent Pit [Sun, 29 Nov 2009 23:14:30 +0000 (00:14 +0100)]
Stop leaking the SVs that reference the callbacks
Vincent Pit [Fri, 27 Nov 2009 11:16:16 +0000 (12:16 +0100)]
Fix a consting error
Vincent Pit [Fri, 27 Nov 2009 11:13:53 +0000 (12:13 +0100)]
Make the wizard store the MGWIZ address as an IV
Also, since we validated the wizard, use SvIVX to read the address.
That way, we no longer need to pass the interpreter to vmg_find().
This results also in some consting goodness.
Vincent Pit [Thu, 26 Nov 2009 21:10:19 +0000 (22:10 +0100)]
Empty the global pointer table at the end of a thread
Vincent Pit [Thu, 26 Nov 2009 20:53:13 +0000 (21:53 +0100)]
Remove unused variables
Vincent Pit [Thu, 26 Nov 2009 20:50:50 +0000 (21:50 +0100)]
Factor the V::M magic searching logic into a new vmg_find()
Vincent Pit [Thu, 26 Nov 2009 00:40:07 +0000 (01:40 +0100)]
Remove all signature-related features
From now on, the wizards are only stored in a global pointer table with
threaded perls.
Vincent Pit [Thu, 22 Oct 2009 18:54:50 +0000 (20:54 +0200)]
Compatibility fix for 5.11.1
Vincent Pit [Sun, 4 Oct 2009 16:06:33 +0000 (18:06 +0200)]
This is 0.38
Vincent Pit [Sun, 4 Oct 2009 15:54:49 +0000 (17:54 +0200)]
Add a cookbook section