This Week on perl5-porters (Oct 11-17 2004)

This Week on perl5-porters (Oct 11-17 2004)

"This Week on perl5-porters" was started by Mark-Jason Dominus five years ago on October 17, exactly five years after the release of perl 5.000.

Debugger handles threads

Richard Foley and Yitzchak Scott-Thoennes added support for debugging programs using ithreads. This added the e and E commands for displaying the current thread ID and list of thread IDs, respectively, when -dt is specified on the command-line or $ENV{PERL5DB_THREADED} is set.

Handling failure in source code filters

Steve Hay wondered what Perl source code filters should do on failure. Perl isn't aware that there was a failure, so it's necessary to croak from the filter. A Perl_filter_read comment suggests that it should return a negative number on failure, whereas S_filter_gets just returns null, so it's not clear which should be correct.

%INC caching failure-case problem

Nicholas Clark found a problem (#31924) in bleadperl:

  $ echo die >
  $ ./perl -Ilib -e 'eval "use fail q{:std};"; $INC{""} = ""'
  Modification of non-creatable hash value attempted, subscript ""
    at -e line 1.

Rick Delaney said this is because of compile failures being cached wrongly, and he submitted a patch to make keys %INC only return names of files successfully loaded. Yitzchak Scott-Thoennes objected to having hidden entries in %INC.

Exporting constants

David Radunz offered a patch to Exporter so that it would declare constants. Michael Schwern fears that this will make Exporter much slower and that it might not be that useful anyway because not all constants are declared with

Pod::Man =head1 capitalization

David Wheeler proposed having Pod::Man force uppercasing of =head1 headings. Russ Allbery resisted this, because Pod::Man doesn't necessarily only generate regular man pages, though he'd be open to an option.

ETA for 5.8.6

Nicholas Clark mentioned that he plans to go to a four month release schedule (previously it was three months), which means a code freeze on Halloween for 5.8.6.

In brief

Craig Berry fixed a Devel::PPPort test on VMS.

Moshe Kaminsky fixed warnings::warn($obj, $msg) where $obj has overloaded stringification.

Tels made not load unnecessarily, saving a bit of memory.


The thread for bug number $BUGNUM can be found at$BUGNUM

About this summary

This summary was written by Scott Lanning. Summaries are published weekly at and posted to a mailing list whose subscription address is The archive is at Comments and corrections welcome.