Hey everyone,

Following is the p5p (Perl 5 Porters) mailing list summary for the past week. Enjoy!

November 23rd-29th

Updates

Peter Rabbitson (ribasushi) has volunteered to maintain base.pm to continue support for perl 5.6.

Christian Jaeger announced the alpha release of Functional Perl.

Another grant report from Tony Cook. In about 19 hours of work, Tony reviewed or worked on 17 tickets and applied 3 patches. More details available in the grant report. Tony provided a full report for October as well, covering over 51 hours of work.

Jarkko Hietaniemi reports pushing a branch that adds clang thread safety analysis annotations to the perl source code. This helps clang evaluate thread safety in its analysis when it isn't certain. Jarkko provides more details about it in the thread.

Karl Williamson updated the porters on a proposal for a new character property (on which The Unicode Technical Committee is asking for feedback) for the class of prepended concatenation marks.

Per a request by Dennis Kaarsemaker, Dave Mitchell has worked on reducing the chatter of a smoke test.

Karen Etheridge submitted a now-merged patch to reduce additional noise from a Module::MetaData test when run under core testing environment.

Bugs

Reported bugs

Perl #126706, reported by Dominyk Tiller, is a compilation error on OS X El Capitan with the new security feature DYLD_LIBRARY_PATH.

Perl #126709, reported by Lukas Mai, demonstrates that local *PACKAGE:: can lead to segfaults and bizarre errors. Dave Mitchell was able to reduce this further and explain why it happens.

Perl #126710, reported by Tony Cook, mentions that pure-Perl modules that have no Makefile.PL (which then use make_ext.pl), cause a rebuild because make_ext.pl doesn't check for fresh files before touching pm_to_blib - which causes a rebuild.

Perl #126719, reported by Michael Adamcik, raises a problem with Encode consuming all memory. Tony Cook investigated and opened a ticket against Encode, RT #10955.

Perl #126735, reported by KES, indicates a perl debugger disconnect problem.

Resolved bugs

• IO::Select does not remove closed file. (Bug already fixed in 2011.)
• Perl #115116: Limit code points to IV_MAX.
• Perl #123710: Segmentation fault in Perl_sv_setpvn.
• Perl #126257: Invalid ARRAY attribute leads to Perl_op_lvalue_flags assertion fail.
• Perl #126443: Use-After-free in Parser.
• Perl #126480: Perl_pp_pipe_op assertion fail when first arg to pipe is definitely not a filehandle.
• Perl #126635: Bleadperl breaks Data::Integer.
• Perl #126687: Duplicate of Perl #126686.

Proposed patches

Tony Cook worked on several bugs, providing patches:

• Helping with the error handling of Perl #123737.
• Fixing Bleadperl module breaking (this time of App::test::travis) in Perl #126593.
• Fixing a memory corruption bug: Perl #124097.
• Null pointer dereferencing in Perl #123991.

Sergey Leschenko provides a patch in Perl #126707, for fixing a race condition in a test, possibly due to the VirtualBox environment.

Shlomi Fish provided a patch for the aforementioned Perl #126735.

Bulk88 proposed patches in Perl #126753 to move the Win32 $^X code to where the $^X code for other OSes live.

Discussion

Ed Avis opened Perl #126715 to seek comments on whether barewords-as-strings are still used in code, in order to understand whether they could be deprecated and removed from core. So far there seems to be little knowledge of many such patterns (few examples given), and the conversation split to the definition itself of "barewords".

Ricardo Signes inquired on why he is seeing a redefinition warning on a piece of code only when defining $^P (an internal variable for debugging support) and Tony Cook chimed in with an explanation.

Additional comments provided by Dave Mitchell for Christian Hansen's proposed patches for faster UTF-X validation and conversation ensued on the meaning of U32_ALIGNMENT_REQUIRED.

Conversation on how to handle incorrect parameters in File::Glob continues in Perl #126239.

Dave Mitchell asks whether PROTOTYPES are actually enabled by default.

Another question from Dave on whether the PerlIOMmap_close function, defined in ext/PerlIO-mmap/mmap.xs, can be deleted as unused or not. Craig A. Berry provides some context, mentioning he thinks it can be deleted.

Glenn Golden asks whether he should open a bug or not on some odd inconsistency he sees in %a formatting option between perl's and C's printf function.