This Week on perl5-porters (5-11 July 2004)

This Week on perl5-porters (5-11 July 2004)

Perl 5.8.5 approaches, and the two release candidates of this week prove it.

Release candidates

Nicholas Clark released the first release candidate of 5.8.5 on Wednesday. Rafael noticed that the regression tests of Gtk2 weren't passing. This was because Gtk2 converts every string to UTF-8 internally, and the internal function to upgrade an SV to UTF-8 was no longer accepting undef as an argument. Rafael provided a patch, which promptly led to RC2 (on Friday). Later, it was found that RC2 caused problems with Tk, for more or less the same reasons; Nick Ing-Simmons was working on this second problem.


Bad warning

Yves Orton noticed the relatively new warning, Newline in left-justified string for sprintf, and disapproves it, arguing that it is of little value for developers and forces to disable the printf warning category only to silence it. However, it won't be removed from the upcoming 5.8.5, since code is frozen. Hugo van der Sanden posted some remarks on revamping the way warning categories work on perlmonks.

Debugger restart

Tim Bunce and Andrew Pimlott report that the R command of the debugger (restart) leaves filehandles open, and this causes problems with databases connections. Andrew sent a patch to fix the problem.

Some sort stuff

Perl had an optimization for reverse sort, i.e. this kind of block:

    @x = sort { $b cmp $a } @y;

John P. Linderman noticed that this optimization breaks sort stability. Nicholas removed it.

Meanwhile, Michael Schwern noted that sort subroutines are not autoloaded (bug #30661).

Selected bugs and fixes

Dave Rolsky found a bug with NEXT and eval(), and (impressively) fixed it.

Steve Hay worked with Ken Williams at improving Cwd's Windows support (bug #30132).

Wolfgang Laun provided a fix for the splice() bug reported last week (bug #30568).

Jeff Pinyan remarks that $1 et alii are not properly set in (?{...}) blocks in regular expressions during backtracking. (Bug #30608.) He tried to debug it, without success so far.

Tassilo von Parseval asks whether it would be a good idea to backport a dummy version of the bytes pragma to CPAN, so it can be used in pre-5.6 perls. Robert Spiers remarks that using the if pragma is a better alternative.

Peter J. Acklam reported a bug (#30609) on Math::BigFloat, regarding the assignment operators (+= etc.) Tels investigated.

Andy Lester proposed to remove references to the idiom

    sub new {
        my $proto = shift;
        my $class = ref($proto) || $proto;

from the perl docs, as it is unnecessary in most cases (bug #30663).

Help needed

Nicholas rewrote the perltodo manpage, with lots of things that would be cool to have. In addition to fixing bugs, you know.

About this summary

This summary was written by Rafael Garcia-Suarez. Weekly summaries are published on and posted on a mailing list, which subscription address is Comments and corrections welcome.