Hey everyone,

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

January 4th-10th

Corrections

Last week's summary accidentally noted discouraging the use of *glob{FORMAT]. It was actually the use of *glob{FILEHANDLE}. Thanks to H.Merijn Brand for the correction and my apology for the mistake.

News and updates

The biggest news of the week were the reworked context stack system Dave Mitchell had been working on. Read more under Discussion.

Tony Cook provided the 8th and 9th reports of his sixth running grant. Over 34 hours in total with 22 tickets reviewed or worked on, and 4 patches applied.

Tony also provides a more detailed report of his grant work - a great read.

Dave Mitchell provided more reports on his grant work.

Karl Williamson updates the list on feedback request from the Unicode Technical Committee on pending additions to the draft repertoire of characters.

Karl Williamson applied his proposed text for utf8.

Karl Williamson provides details on the tailored implementation of qr/\b{wb}/, now committed.

Karen Etheridge updated podlators in blead to 4.04.

The release of podlators 4.04 accidentally changed the generation of the scripts shebang line on Windows. Craig A. Berry noticed, notified Russ Allbery, and Russ will revert those changes.

Ricardo Signes will delete all RC versions of perl from his CPAN directory. They will remain available on the BACKPAN mirrors.

Bugs

Reported bugs

• Perl #127149: A local PerlIO::encoding::fallback setting + encoding causes an infinite loop.
• Perl #127189: B::Deparse doesn't emit /n regex flag.
• Perl #127182: One digit short to correctly stringify a double
• Perl #127183: Non-canonical hexadecimal floats are parsed prematurely.
• Perl #127231: Bleadperl breaks App::JobLog.
• Perl #127212: ABI incompatibilities with -DDEBUGGING vs. without, caused by memory_debug_header.
• Perl #127232: Segmentation fault with Storable.

Resolved bugs

• Perl #126845: Perl_newATTRSUB_x: Assertion block failed.
• Perl #127131: In Configure: libswanted=nm suddenly harmful.
• Perl #126847: Use the new fdclose(3) function in FreeBSD.
• Perl #127080: Perldoc for waitpid missing description for argument value -1.
• Perl #126877: Improve build performance on Win32 substantially by using a real getcwd.
• Perl #126042: Segmentation fault in Perl_pp_multiply (and other functions).
• Perl #126922: crypt uses uninit/unalloced/freed memory as salt.
• Perl #127189: B::Deparse doesn't emit /n regex flag.
• Perl #126240: Panic happening after freeing the env after a fork.
• Perl #126633: Bleadperl breaks ddb.
• Perl #123737: S_no_op assertion failure.
• Perl #123788: Unloading Mouse class crashes Perl on 32-bit MSWin32.
• Perl #126862: XS File::Spec's canonpath loses taint. (CVE-2015-8607)
• Perl #126755: OOB access in VDir::MapPathW (Win32-specific). (CVE-2015-8608)

Proposed patches

Tony Cook provided a patch in Perl #127063 for a warning generated by setting $[ to zero.

Tony also provided a patch in Perl #127122 for making sure unless ( $a = 2 ) issues a warning just like if ($a = 2) does.

Yet another patch from Tony in Perl #126991 for integer overflow in line number tracking reported by caller().

Tony tried a different approach (with a suggestion from Karl Williamson) to handle a VC 2003 crash in Perl #126045.

Douglas Bell provided a Github Pull Request for Perl #127028.

Tom Hukins provided a (now merged) patch in Perl #127229 for fixing module paths in INSTALL.

Niko Tyni provided a patch in Perl #127234 to fix a Configure escape when usecrosscompile is used but there is no targethost.

Discussion

Dave Mitchell has reworked the context stack system to allow some subroutines to be much faster now. It breaks several CPAN modules (notably Scope::Upper) but there seems to be possible fixes in place.

Tony Cook writes in Perl #127054 (Bleadperl breaking Test::MemoryGrowth) that he is seeing inconsistent behavior if the test is run using perl or using make test and adds some details on how he ran checks and what he found. Tony was not able to find the cause yet.

Dan Kogai discovered an interesting problem (and opened Perl #127182) with having one digit short to correctly stringify a double. Dan then provided outputs from several languages, showing which also get it wrong (Rakudo Perl 6, PHP 5, Python 2, and Swift) and which get it right (Go, Haskell, Node.js, Python 3, and Ruby). Craig A. Berry added that changing it to 18 fixes a loss of precision he had on FreeBSD, and Zefram linked this issue to Perl #41202.

Karl Williamson suggested not generating EBCDIC non-1047 tables, since there seems to be no user of those. This will save time when regenerating and the Perl distribution size will decrease. Ze'ev Atlas (maintainer of PCRE for z/OS) provides different experience.

Steve Hay tried to update Pod::Checker, but since it is now derived from Pod::Simple instead of Pod::Parser, there is a failing test that requires additional work. Karl Williamson added that he will work on the changes necessary when time permits, as it is not urgent for Perl 5.24.

Bulk88 emailed the list about an idea he has for a new threaded memory allocator for Win32 perl.

An email was sent asking about a "secure" version of perl, without any FS access.

Ricardo Signes asks for help on making the release of PathTools simpler, since it carries its own ppport.h file.