Hey everyone,

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

November 10th-15th

Updates

Steve Hay has released Perl 5.22.1-RC2!

The hoary frost has a white sheen Strewn all over my hair, So I thought I was an old man And thought life dealt me fair.

Yet soon was thawed my old white mane, And I have my black hair again. How I abhor my young fair years, How long to wait for death and biers?

From setting sun to morning's hue Many a head turns white. Who'll credit it? My hair did not In all this lifelong plight!

-- Wilhelm Müller, trans. Anon., "The Old Head" (No. 14 in Schubert's song-cycle, "Winterreise")

Aaron Crane has updated Perl #119437 on the status of the implicit dereference in push to a reference as an unsuccessful experiment.

ExtUtils::MakeMaker released a new version which reverted many changes, including a commit by Sisyphus to build when sh.exe is in the path (but following cmd.exe) and several changes by mohawk. Therefore, both Sisyphus and Bulk88 resubmitted the "lost" patches to include in EUMM (ExtUtils::MakeMaker).

Tony Cook provides another grant report. About 16 hours spent, 13 tickets reviewed or worked on, and 5 patches applied.

Dave Mitchell had merged his fast arithmetic branch. Expect even faster basic arithmetic in the next perl version! You can read more about it in his grant report mentioned below.

Dave provides a grant report. 98:49 total hours. He worked mainly on the aforementioned arithmetic optimizations and the overhaul of the perl context stack and dynamic scope implementation.

Dave also mentioned he applied for an extension of his Perl core grant. The Perl Foundation is looking for comments. If you want to show appreciation (other than a thank-you email to p5p and/or Dave), you can provide comments on the grant extension request.

Ricardo Signes is making a few CPAN releases (PathTools and base) in order to fix the installation path on perl 5.12 and above.

Ricardo wants to remove encoding.pm. He asks about any blockers.

Ed Avis noticed the docs sometimes use lookahead and sometimes look-ahead (with or without a hyphen) and provided a patch to solve it in Perl #126608. Tony Cook provided additional comments.

Module::Metadata is passing several TODO tests, adding more output to the build process. Karen Etheridge updates it will be resolved soon.

Mohammed El-Afifi provided a patch in Perl #126611 for File::Find to prevent death in case an empty array is provided.

Dagfinn Ilmari Mannsåker provided a patch in Perl #126609 to include the error message when IPC::Open3 fails. It was applied by Tony Cook.

Bulk88 submitted patches in Perl #126632 (which he intends to revise further) to add parallelness to Win32/GNUmakefile.

Bugs

Reported bugs

Perl #126602, reported by Brian Carpenter, is a fuzzer-found double-free segfault caused by the code:

s//*_=0;s|0||;00.y0/e

Tony Cook debugged and provided submitted a patch for review.

Perl #126613, reported by Dan Collins, also found by a fuzzer, causes a segfault with the code:

/(?{s0})|(?{s0:})/|s|||eg|y|||

Perl #126614, reported by Dan as well by fuzzing, segfaults with:

/(?{})|(??{U:0})/|s|||g

Perl #126615, another fuzzer-discovered fuzzer by Dan with the code:

00./(?[[0]+()+])/

This was already fixed by Karl Williamson.

Andreas J. Koenig opened several tickets for bleadperl breaking the following modules:

• Perl #126621 breaking Tie::DBIx::Class.
• Perl #126633 breaking ddb.
• Perl #126635: breaking Data::Integer.
• Perl #126637: breaking Data::Rmap.

Resolved bugs

• Perl #126600: Update perlexperiment for removal of lexical topic and autoderef.
• Perl #124080: DEBUGGING versions of the MARK macros to display arity of op and function calls.
• Perl #124068: Various improvements to DynaLoader.
• Perl #124069: XSLoader improvement.
• Perl #126325: Segmentation fault in pack.
• Perl #126179: Rejected regression report with \N{}.

Discussion

The ticket Perl #126602 sparked a response from Zefram on a slew of problems with stack refcounting. He proposes a way to handle them which will require some major overhauls. Dave Mitchell proposes another way which wouldn't require the same amount of work and might prevent breakage in API.

A breakage caused by a commit which required a version number bump raised a rant by Jarkko Hietaniemi on tests which require such version bumps which conflict with how one might run tests. The thread continues here.

Jarkko also mentioned a Python change suggestion to provide a faster directory iterator. This also evolved into a conversation on the stat caching feature.

Steve Hay suggests maintenance releases should also include back-ported fixes to features introduced in those stable releases. Aaron Crane provided a patch to address this in the policy.

Peter John Acklam wondered about results when working with infinite numbers and exponents. Additional details provided by Sisyphus and Glenn Golden, it would appear that the computing part of these operations were marked as contentious by IEEE during the early 754-2008 process.