This Week on perl5-porters (31 May / 6 June 2004)

This Week on perl5-porters (31 May / 6 June 2004)

As I've advertised the weekly P5P summaries during the French Perl Workshop, I feel compelled to continue to write them each week... Read on for the latest batch of commented links.

Smaller tarball

Nicholas Clark, enhancing Jeff "japhy" Pinyan's improvements to the Unicode data tables, makes them generated at perl compile time. This way, an unpacked clean source tree is 1.6 MB lighter.

    http://groups.google.com/groups?selm=20040531151254.GX1147%40plum.flirble.org

However, this produced some problems with builds on Windows using the dmake utility: the build now works, thanks to Steve Hay, but dmake tends to build things several times unnecessarily.

Data::Dumper and hooks

Gaal Yahas proposed to allow Data::Dumper to call methods on objects it dumps, if these methods exists, to produce a class-specific customized dump. Comments were posted.

    http://groups.google.com/groups?selm=20040604141439.GI647%40sike.forum2.org

$SIG{FOO} = undef

Dave Mitchell proposed that assigning undef to a signal handler does the same thing as assigning 'DEFAULT': restore the default handler. Nick Ing-Simmons remembers that in the past, this question was discussed: shouldn't it set the signal handler to 'IGNORE' instead?

    http://groups.google.com/groups?selm=20040604151423.GA1896%40iabyn.com

fork() at compile-time

Robin Barker reported that on Solaris, forking in a BEGIN block seems to repeat the remaining parts of the script twice, even though the child process exited immediately (bug #30040). Nicholas Clark's advice is to use POSIX::_exit() instead of perl's exit(). Ton Hospel suggests an explanation.

    http://groups.google.com/groups?selm=c9sa2u$uk4$3%40post.home.lunix

Thread leaks

Two thread-related memory leaks were reported: one involving join(), by Andrew Savige (#30063), and one involving nested arrays, by Eric Garland (#30066).

    http://groups.google.com/groups?selm=rt-3.0.9-30063-89801.3.14458478683115%40perl.org
    http://groups.google.com/groups?selm=rt-3.0.9-30066-89809.5.74912811298226%40perl.org

Double DESTROY

Ton Hospel produces a small code snippet where an object, stored in a hash, is destroyed twice -- that's certainly one time too many. (Bug #30061).

    http://groups.google.com/groups?selm=rt-3.0.9-30061-89788.12.6990632560854%40perl.org

Memory wrapping

Calle Dybedahl finds that the new error messages emitted for large allocations of strings are not emitted in some cases. Jarkko Hietamieni provided patches.

    http://groups.google.com/groups?selm=86smdb8y27.fsf%40ulthar.bisexualmenace.org

In Brief

Ken Williams announced a new beta of Module::Build on CPAN. Testers (and patches, but you know that) are welcome.

    http://groups.google.com/groups?selm=F5F32715-B452-11D8-9FAD-000A95BD9874%40mathforum.org

Sadahiro Tomoyuki fixed bug #29841 (see last week's summary), about the -T taint checks flag modifying the behaviour of perl regarding utf8::decode(). This function, however, is considered as experimental and better suited for internal use.

    http://groups.google.com/groups?selm=20040606003344.57B2.BQW10602%40nifty.com

Oddly enough, this week there were several bugs reports concerning bugs that were already fixed, either in perl 5.8.4, in perl 5.8.5-to-be, or bleadperl.

About this summary

This summary was written by Rafael Garcia-Suarez, from the French Perl Workshop, a.k.a Journées Perl 2004. Weekly summaries are published on http://use.perl.org/ and posted on a mailing list, which subscription address is perl5-summary-subscribe@perl.org. Comments and corrections welcome.