Revision history for IPC::MMA
0.8 Feb 27 2010
* Added an Example to the pod documentation (a user suggested
this, it should also raise IPC::MMA's Qwalitee leverl :-)
* Undid some changes to C compiler testing in Makefile.PL,
that were a bad idea.
0.7 Feb 21 2010
* Improved Makefile.PL to check more deeply for the presence
and completeness of the mm library, to work around an Unknown
test result on one CPAN tester's system.
* Improved C compiler testing in Makefile.PL to deal with
compound names like 'ccache cc'.
* Corrected two method names in .pod docs from 'mm[a]_hash_' to
'mm[a]_array_'.
0.6 Nov 23 2009
* Added NaN filter to random DOUBLE generation in 04_fixedArray.t.
* Improved reliability/consistency of 09_lock.t operation by
** reworking delays/timeouts for slow/heavily-loaded systems,
** using a local copy of the global variable, and
** always checking the expected value of the local copy,
as a condition for setting the next global value.
0.59009 Nov 19 2009
* Reworked test 09_lock.t to try to eliminate a FAIL on a CPAN
tester's slow system.
* Reworked optimization selection in Makefile.PL to be more
compatible with various toolsets.
* Eliminated error messages when NaN or inf values are stored
in DOUBLE_ARRAYs. Random number generation in test
04_fixedArray.t will occasionally produce NaNs and infs.
0.59007 Nov 17 2009
* Modified Makefile.PL to work around some problems with the
'inline' keyword under the Sun C compiler.
* Built a Perl 5.6.2 with 64-bit integers and worked thru test
problems using it.
* Improved test 04_fixedArray to test option MM_CSTRING as well
as MM_FIXED_REC.
* Updated Makefile.PL prerequisites, abstract, author.
* Added test for .pod file.
0.59005 Nov 13 2009
* Found/fixed a nasty problem involving list operands of array
push, unshift, splice.
* Improved UINT value-testing mechanism for perl 5.6.
* Revised test 4_fixedArray to use proper equality operator (==
or eq) based on the type of array (perl 5.6 again)
0.59004 Nov 11 2009
* Another try to fix the problem under perl 5.6, wherein storing
into UINT arrays gives unwarranted errors for large values.
* Improved the mechanism for eliminating 'inline' on cc's that
don't support it.
* Yet another tweak to t/9_lock.t, which has still had some
"hang at state 23" errors on some smoke-testers' systems.
0.59003 Nov 6 2009
* Fixed numerous C warnings in MMA.xs, that were revealed by
smoke-test systems with more rigorous C compilers than gcc-x86.
* Tried to fix a problem under perl 5.6, wherein storing into
UINT arrays gave unwarranted errors for values with the MS bit
set.
* Removed -s option from 'which mm-config' command in Makefile.PL.
Some which'es don't implement -s.
* Restricted cc '-Os' option (optimize for size) to those systems
that can handle it (e.g., gcc).
* Restricted 'inline' functions to compilers that can handle them.
0.58001 Oct 23 2009
* Changed tests 5, 7, 8 to use /dev/urandom (or perl rand) rather
than /dev/random to avoid entropy delays on newly-started Linux
systems. Thanks to Oliver Paukstadt for this suggestion.
* Extended final timeout in test 9 to accomodate heavily-loaded
smoke-test systems.
* Reworked detection of mm library in Makefile.PL. New version
tries to balance the needs of smoke-testers vs. regular users.
Thanks to David Golden for his advice on this.
* Added license=>perl to Makefile.PL to make YAML reviewers happy
0.58 Aug 15 2009
Included revised t/2_mmArray.t to try to fix 0.54 failure
(inadvertently omitted from 0.55-57)
0.57 Aug 13 2009
* Used Devel::PPPort to add ppport.h file to distribution
* Included (in C preprocessor sense) ppport.h in MMA.xs
* With help of script in ppport.h, defined two symbols in MMA.xs
that make ppport.h define required "IS_NUMBER..." symbols when
they're not provided by an older Perl
* Added prerequisite for latest version of ExtUtils::ParseXS, an
attempt to avoid some nasty-looking warnings in 'make' step,
that I had just battled through on my own system
* revised t/9_lock.t in response to FAIL on 0.56
* added Devel::CheckOS to distribution and Makefile.PL, to
restrict IPC::MMA install to OSes that mm runs on
(Still no change to IPC::MMA itself since 0.54)
0.56 Aug 11 2009:
Fixed some bad code in Makefile.PL of 0.55
It shouldn't have worked but it seemed to for me. Forget 0.55!
0.55 Aug 10 2009:
Tests changed to avoid some problems found by CPAN testers
Improved Makefile.PL to ask for mm library (to reduce UNKNOWN tests)
(No change to IPC::MMA)
0.54 Feb 26 2009:
* Fixed boolean array operation on platforms with 64 bit integers
* Corrected the .pod docs some more
0.53 Feb 23 2009:
* Fixed 2 bugs introduced with the "minor improvement" in 0.51:
** the mm_sizeof routine wrongly computed when an allocated
block could remain in place
** mm_hash_store used its return value in the wrong polarity
* Revised installation tests and the mm_alloc_size function
so that they should succeed on more platforms.
* Also for CPAN testers, include Test::Warn as a prerequisite.
* Further tweaking of return values in functions that return lists.
* Documentation updated in several places.
0.52 Feb 15 2009:
* Set cc Optimize flag in Makefile.PL to optimize (0.51 was
set for debugging).
* Return empty lists for exceptional conditions in functions
that return lists.
0.51 Feb 14 2009:
* Added use tags (:basic :scalar etc) to .pod doc.
* Minor improvement in keeping shared memory blocks in place,
when they don't have to be re-allocated.
0.5 Feb 13 2009: New package with largely IPC::MM compatible names.
Retains direct call and tied interfaces for scalars and hashes.
Adds direct-call and tied array functions with multiple storage
types, including general purpose, signed and unsigned integer,
floating point, boolean, and fixed-length records
(a.k.a. maximum-length strings).
BTree functions are retained as aliases for Hash functions.
Hash functions return keys in sorted order, which was the
main difference of IPC::MM's BTree mode compared to Hash mode.