Try to avoid strange failures when 1 == 1 doesn't give us PL_sv_yes.

[p5sagit/Devel-Size.git] / CHANGES

Revision history for Perl extension Devel::Size.

0.80_50 2017-06-21 nicholas
  * Attempt to fix strange 1 == 1 failures on some older threaded perls.

0.80 2015-03-31 nicholas
 [no changes]

0.79_54 2015-03-27 nicholas
  * Fix the definition of padlist_size() for post-5.20 perls.
  * Avoid tests failing because Test::More lazy-loading changes the size of %::
  * Fix the build under 5005 threads. [Not that anyone is using them :-)]

0.79_53 2015-03-25 nicholas
  * Avoid t/recurse.t failing on earlier versions on 32 bit platforms
    (subroutines are smaller than it assumes)
  * Count the size of padlist names in CVs (for v5.18 and later - earlier TODO)
  * Handle PADNAMELIST/PADNAME introduced in v5.21.7
    two patches from Zefram:
  * Add handling of children of METHOP and UNOP_AUX ops [CPAN #102911]
  * More complete handling of padlists for XSUBs [CPAN #102909]

0.79_52 2015-03-20 nicholas
    two patches from Zefram:
  * handle the new METHOP. [CPAN #101071]
  * handle the multideref op, and to a limited extent the UNOP_AUX op class in
    general. [CPAN #102911]

  * As of eacbb37937698a03 (v5.21.6), can't call CvPADLIST on an XSUB.
    More usefully, the slot is now used for XS versioning protection, which
    means that we can be confident that nothing out there is using it for data.
    Reported as [CPAN #102909].
  * Avoid t/recurse.t failing because Test::More::is() has been refactored and
    is now larger than it was. [CPAN #102910].

0.79_51 2015-02-28 nicholas
  * as of 5.20.0, s/// is no longer a reliable test for OOK [CPAN #95493]

0.79_50 2015-02-28 nicholas
    patches from Zefram:
  * wrong test for magicalness [CPAN #88180]
  * reading tied scalar shrinks it [CPAN #88182]

0.79 2013-05-12 nicholas
 [no changes]

0.78_52 2013-05-10 nicholas
 * Fix the bug in the previous fix. Oops :-)

0.78_51 2013-05-10 nicholas
 * v5.17.1 and later handle CopSTASH differently under ithreads

0.78_50 2013-05-10 nicholas
 * Alternative test for CvOUTSIDE, and one for CvPADLIST
 * v5.17 changes the type of PADLIST [CPAN #83904]
 * stubbed subs no longer have CvOUTSIDE set [CPAN #77913]
 * Fix for v5.17.2's OP slab allocator [CPAN #83903]
 * Spelling fix to docs from gregor herrmann [CPAN #78766]

0.78 2012-07-26 nicholas
 [no changes]

0.77_51 2012-07-17 nicholas
 * Fix for MSVC builds from bulk 88 [CPAN #77589]

0.77_50 2012-02-10 nicholas
 * t/globs.t was failing on 5.15.6 and later due to side effects of a change
   to strict.pm [CPAN #73998]
 * skip tests in t/magic.t that use formline on 5.8.1 and 5.8.2, as those
   versions have a buggy formline that can trigger an assertion failure.

0.77 2011-05-16 nicholas
 [no changes]

0.76_50 2011-05-12 nicholas
 * Split out HEK size calculation into hek_size(). Add the shared HE overhead.
 * Handle shared hash key scalars correctly.
 * GvNAME() is shared from 5.10 onwards.
 * Count HvNAME(), the HV "aux" struct, the mro_meta struct, and ENAMEs.

0.76 2011-05-11 nicholas
 * Just fix the version number in the line below.

0.75_52 2011-05-09 nicholas
 * Use a table for SV body sizes. These incorporate the space saving post 5.8.x
 * Correctly handle SvOOK scalars. 5.12 and later don't use SvIVX().
 * Magic vtables aren't freed when magic is freed, so don't count them.
   (They are static structures. Anything that assumes otherwise is buggy.)
 * Reinstate 5.6.x "support". (i.e. it compiles and passes tests.)
 * Reinstate 5.005_xx "support". (i.e. it compiles and passes tests.)

0.75_51 2011-05-05 nicholas
 * Only use a static array of vtables on gcc.

0.75_50 2011-05-04 nicholas
 * The core's magic vtables are global constants, so aren't part of the size.
 * Follow mg_obj and mg_ptr.

0.75 2011-05-04 nicholas
 [no changes]

0.74_53 2011-05-01 nicholas
 * Fix typo error in bit-vector tracking mechanism. On 64 bit platforms which
   allocate on 8 byte alignment (rather than 16), a low pointer bit could get
   lost, resulting in new pointers being considered already "seen".
   "Found" by BinGOs' smoker, fixed on spectre.mongueurs.net. Thanks.

0.74_52 2011-04-23 nicholas
 * Fix potential SEGVs for OP_AELEMFAST on a lexical (eg $foo[3])
 * Fix likely SEGVs for PVOPs (missing break)
 * Fix potential SEGVs for PVBMs on 5.10 and later
 * Move hash and array traversal from total_size() to sv_size()
   - this allows total_size() and size() to be XS ALIASes.

0.74_51 2011-04-22 nicholas
 * Don't count PL_sv_{undef,no,yes} in the size returned
 * total_size() was double-counting entries in typeglobs
 * sv_size() was double-counting the PVGV size if GvEGV() looped back
 * fix classic off-by-one error - the answer is strlen() + 1, not just strlen()

0.74_50 2011-04-19 nicholas
 * Ensure that size() doesn't add the referent's size for non SVt_RV references

0.74 2011-04-19 nicholas
 * Correct the Makefile.PL - LICENSE was added to ExtUtils::MakeMaker in 6.31

0.73_51 2011-04-17 nicholas
 * Refactor the C code to accumulate the size inside the tracking structure
   - this means that the C *_size() functions now return void
   - The XS functions size() and total_size() now return UV, not IV

0.73_50 2011-04-17 nicholas
 * Avoid using assert() on ithreaded 5.8.8 and earlier, as it needs my_perl
 * Fix the test added in 0.73, as it tripped over a combination of bugs

0.73 2011-04-16 nicholas
 * Revert a bad assertion introduced in 0.72_50, which was logically wrong.
   - and a test to demonstrate one case that would trigger it

0.72_52 2011-04-15 nicholas
 * Add a test for the non-exceptional warnings.

0.72_51 2011-04-15 nicholas
 * Add PERL_NO_GET_CONTEXT to improve performance under multiplicity

0.72_50 2011-04-14 nicholas
 * Exception handling is totally MSVC specific, so only use it there
   - this means that we don't need to use a C++ compiler anywhere
 * Rework bit-vector tracking mechanism to use a 256-way tree. This avoids
   making assumptions about 64-bit platforms' memory layouts, and eliminates
   the fatal error introduced in 0.72 when the assumption was violated.
 * Convert to XSLoader
 * Resolve CPAN #49437 (Devel::Size adds magic in Perl 5.10)
 * Resolve CPAN #58484 and #58485 (related to CVs that are XSUBs)

0.72 2008-10-14 BrowserUk 70 tests
 * Added bit-vector pointer tracking mechanism.
   - new fatal error (64-bit platforms only)
 * Added exception handling
    - 4 new warnings (disabled by default)
 * Updated POD to reflect above
 * Added basic.t test 13 
 * replaced Module::Install with hand crafted EU::MM Makefile.pl
   (With many thanks to Sisyphus)
   because we couldn't work out how to add C++ options to allow 
   exception handling.

0.71 2008-08-24 Tels 69 tests
  * adapt patch from Reini Urban to fix failing RV under 5.10 and 5.11. AV
    and HV were pushed directly onto the pending_array, and not the RV,
    which caused #33530. So he rewrote the logic to deref the RV inside
    the array traversal. Applied this with one small omission, which
    caused test faiures.
  * Fixed 5.11 RV/IV logic. (Thanx Reini Urban!)
  * Removed one duplicate total_size arrayref test. (Thanx Reini Urban!)
  * changed //printf to dbg_printf() (Thanx Reini again!)

0.70 2008-08-23 Tels 69 tests
  * fix SEGFAULTS under v5.10 (Thanx Reini Urban!)
  * fix compilation under blead (Thanx Reini Urban!)
  * require Perl 5.006 (to stop CPANTESTERS sending bogus reports
    about failed compiling on ancient Perls)
  * update the bundles Module::Install to v0.77

0.69 2007-08-11 Tels 69 tests
  * fix compilation under Perl v5.9.5 and v5.10 (Thanx Steve Peters!)
  * clarify the license by specifying Perl v5.8.8's license
  * small doc fixes, add a README file

0.68 2007-06-12 Tels 69 tests
  * remove a bit of duplicate code in op_size, the second instance
    was a no-op anyway
  * fix the failing tests by rewriting the test logic to be sane
  * the count of array slots was off by one

0.67 2007-03-15 Tels 20 tests
  * rip out Build.PL since it doesn't actually work
  * fix bug #1653: All SVPV subtype may contain an RV instead of a PV
    (Thanx a lot Jan, and sorry for the truly horrible long delay!)
  * fix bug #24846: Does not correctly recurse into references in PVNV
    (Thanx STEVIEO)
  * add t/recurse.t that tests the two bugfixes in combination

0.66 2007-03-02 Tels 15 tests
  * fix the failing test (the code was right, but the test wrong)

0.65 2007-02-24 Tels 15 tests
  * fix the test failings under 64 bit (bugs #21037, #18596, #21404)
  * fix wrong size for strings (bug #17586)
  * reverse Changelog (bug #16636)
  * Size.xs: remove unused variable "count"
  * Size.xs: fix warnings about missing ()
  * Size.xs: fix size of array refs when AvALLOC == 0
  * Modernize package:
    + maintained by me, signed with my key
    + bundle Module::Install and use it
    + add POD tests
    + rewrite tests to use Test::More; add more tests

0.64  Mon Dec 19 18:46:00 2005
        - Patch to make Devel::Size compile on bleadperl (Courtesy of
      Nick Clark)
    - Fix up the PVIV string size calcs (Courtesy of Andrew Shirrayev)

0.63  Thu Jun 30 14:29:43 2005
        - Turns out that OP_TRANS is a basic op, not a loop op. This
      is a bug in versions of perl 5.8.x before 5.8.7. Work around
      it.

0.62  Tue Jun 28 11:59:00 2005
        - Took out // comments
    - Added in copyright notice
    - Some small amount of regex parsing
    - Suppress multiple copies of each warning on each call

0.61  Mon Jun 27 16:19:00 2005
        - Added more checks for formats
    - Got CVs sizing right

0.59  Sat Nov 27 16:42:42 2004
        - Applied documentation and sane warning patch from Nigel Sandever
        - Taught Devel::Size how to size up IO and globs properly

0.58  Fri Jul 18 11:42:32 2003
        - Fix for problems triggered by perl 5.8.0 and up, more tests, and
      removal of an "our" for better backwards compatibility. (Courtesy
      of Marcus Holland-Moritz <mhx-perl@gmx.net>)

0.57  Thu Mar 20 13:21:14 2003
        - setting $Devel::Size::warn to 0 disables not complete warnings

0.56  Mon Feb 24 12:10:13 2003
        - Chopped out some C++ comments. D'oh! Version incremented for CPAN

0.55  Sat Feb 22 17:21:00 2003
        - Fixed a bad size calculation (we were overestimating by one byte)
    - Updated the docs to show some of the places that there might be 'dark'
      memory that Devel::Size can't see.
    - Added in tests from Ken Williams

0.54  Sat Oct 12 14:11:00 2002
    - Applied a patch to get it working on 5.8.0 under Tru64

0.53  Thu Oct 10 12:30:00 2002
        - Finally started updating Changes file
    - Applied doc patch from Ann Barcomb
    - Got globs sizing right

0.01  Mon Oct  7 01:05:32 2002
    - original version; created by h2xs 1.2 with options
        -A -n Devel::Size