Hey everyone,

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

October 27th - November 1st

Updates

Perl 5.22.1-RC1 is now available. Thanks to Steve Hay for the work!

In the deepest rocky crevice A will-o'-the wisp lured me; How I could find my way from here, For me it's easy memory!

For I am used to straying ways, Every path to th'end a way, All our joys and all our suffering,-- To a will-o'-the wisp it 's all play!

Through the dried-up bed of torrents I quite calmly downward stroll; Every stream its sea will enter, Every suffering finds its goal!

-- Wilhelm Müller, trans. Anon., "Will-o'-the Wisp" (No. 9 in Schubert's song-cycle, "Winterreise")

More agreement with Karl Williamson's suggestion to include fixes to \s{wb} and \s{sb} in perl 5.22.1 and it went in. Due to performance hit concerns raised by Yves Orton, the suggested fix to /(?(?!))/ mentioned in Perl #126222 was left out of 5.22.1.

Bulk88 offered an updated patch for Win32 parallel builds in Perl #126452, which was merged.

Jarkko Hietaniemi resolved a few older IRIX compilation tickets (mentioned below).

Tony Cook provided a solution for some Coverity issues (mentioned last week) Jarkko has been working through.

Dave Mitchell has another grant report for us. Dave continued working on the context stack, but also took a break to see about optimizing some arithmetic operations, making the nbody benchmarks (focusing heavily on floating point vector arithmetic) about a third faster. Great news! Bulk88 provided several improvement points to the code which Dave had created another branch for.

Matthew Horsfall suggests the problem in Perl #126070 is likely to be with TryCatch.

Aaron Crane provides insight in Perl #126170 on BEGIN blocks with subroutine attributes.

Bugs

Reported bugs

Perl #126480, reported by Dan Collins, shows an assertion failure with the code:

pipe$$5,0

Perl #126482, reported by Dan, shows another assertion failure with the code:

sub(){sub ub(){0}ub ub

Perl #126481, reported by Dan, is one more assertion failure with the code:

0=/(?[!!(\w])/

Which had been fixed the next day by Karl Williamson.

Perl #126469, reported by Ben Bullock, asks to document the second argument of sv_reftype. Patches by Tony Cook added.

Perl #126474, reported by Kazuho Oku, mentions that unsafe signal handlers use newSVsv, thus calling malloc. He opened the bug since perl 5.17 introduces this call at all times, which wasn't the case prior.

Perl #126484, reported by Andreas J. Koening, mentions that perl 5.23.3 breaks Lexical::SealRequireHints.

Perl #126472, also reported by Andreas, shows a problem in 5.23.3 breaking additional modules which run very explicit checks for a behavior which does not break the module, but the specific test. Dave Mitchell advised updating the test.

Perl #126502, reported by masterchiefaragorn, shows that once you freeze and thaw floating point numbers using Storable they will no longer be equal using ==. Zefram explains in the ticket why this happens.

Perl #126515, reported by John Imrie, regarding the missing Unicode character Block=CJK_Unified_Ideograph. This was the wrong name and it only came up as an error now since Karl Williamson added a compile-time check for for valid Unicode properties.

Resolved bugs

• Perl #125298: warnings.t one failure in IRIX 6.5.
• Perl #126453: Don't distribute version::vpp/EUMM::version::vpp.
• Perl #116062: Installing perl 5.16.2 in SGI IRIX 6.2.
• Perl #39797: Unable to build Perl under IRIX with -Duseshrplib.
• Perl #125314: Configuring with Accflags DAPPLLIB_EXP broke in 5.22.0.
• Perl #126468: (Duplicate of [Perl #125314].)
• Perl #126309: x operator on list causes segfault, confuses valgrind (64-bit).
• Perl #126481: Assert fail with 0=/(?[!!(\w])/
• Perl #124212: porting/extrefs.t fails in tru64.

Discussion

More conversation happening in Perl #126414 about number rounding in perl (mentioned last week). Tony Cook suggests using EOVERFLOW (as both Solaris and FreeBSD do) and to warn if the value of inodes changes because of rounding. Aristotle Pagaltzis supports stringifying and Bulk88 notes it might be the only way to handle 128 bit inodes.

Ricardo Signes suggests opening a meta ticket in order to track all the fuzzer-related bugs. Dan Collins wants a way to triage the tickets and have multiple users be able to edit them, even if only a spreadsheet.

Paul "LeoNerd" Evans wonders if it's possible to warn about POD in the middle of an expression. Unfortunately, as shown by Eric Brine and H.Merijn "Tux" Brand, it's a problematic idea, partly because of functions and partly because of indirect notation support.

The discussion on optimizing read (mentioned last week) in Perl #126403 is continuing. Craig A. Berry wonders whether the suggested solutions really speed things up. Apparently on GNU/Linux and Macs (or more accurately, non-Windows) it does. Craig will try to provide a clean patch, but no promises on when.

Bulk88 is continuing work on cleaning up dependencies on having DynaLoader available for miniperl. He wrote about his plan, then wrote more verbosely on his journey down the rabbit-hole. Perl #126533 and Perl #126534 are the result of his work, provided as patches.

Karl Williamson continues his research and possible solution suggestions on handling Unicode characters beyond 72-bits (past 0xFF).