This Week on perl5-porters (30 June / 6 July 2003)

This Week on perl5-porters (30 June / 6 July 2003)

As the next maintenance release of perl is getting closer, the porters are still fixing bugs. Among the subjects that have been investigated this week, we can remember some hash-ordering-dependent bugs, process name problems, and more syntactic issues.

Hashes that bite

The hash seed randomization patch that was introduced last week (see our previous summary) uncovered some nasty bugs in perl and in its test suite.

Craig Berry discovered a bug in the order of destruction of weak references. This was fixed by Dave Mitchell.

Enache Adrian noticed that rebuilds were made quite more often. This was caused by some shuffling of the keys in the generated Config.pm.

There were some hash-order dependent tests as well, sometimes the cause being deeply hidden in the harness, or in the internals of Test::Builder.

Paul Johnson says that he expects some very impressive fireworks when this gets out into the wild. And in fact Slaven Rezic obtains random failures with HTML::Mason (that would be caused by Exception::Class) and SOAP::Lite.

Setting $0

Andreas Koenig and Jarkko Hietaniemi were (once again) working on the problem of setting the $0 variable and having it change the name by which the operating system knows the perl process. The problem is, of course, how to cope with the ideas the various platforms have about it.

perl581delta

Jarkko began to work (with the help of the porters) on the next incarnation of the perldelta manpage, to be bundled with the upcoming perl 5.8.1. And he released a large number of maintperl snapshots. Release candidates are obviously getting closer.

More syntax weirdness with indirect objects

Dave Mitchell gives a snippet of code :

  sub f { f a; } f a;

where the first f a is an indirect method call, using the indirect object notation (it's equivalent to a->f()), whereas the second f a is equivalent to f(a) (with a being a bareword).

I'd say this particular syntax oddity is sub-optimal, -- but the indirect object syntax is hairy.

    http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2003-07/msg00066.html

Harness speed

Abigail complained that Test::Harness prints a counter at each test, thus augmenting the amount of time needed to run large number of tests. This ended up in a patch. As says Andy Lester, Test::Harness' maintainer, now, when you run a test harnessed, the numbers don't fly by one at a time, one update per second. (In the same thread, Gurusamy Sarathy recalls the HARNESS_NOTTY environment variable, that suppresses progress messages.)

    http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2003-07/msg00212.html

Briefly

Jarkko, after a suggestion by Robin Barker, added a new utility in perl's source distribution, Porting/Modules, to list the official maintainers of the modules (this is mostly useful for the modules that have a dual life on CPAN). Michael Schwern pointed out that he was reinventing the Module::CoreList wheel ; and Jarkko removed this Modules.

Dan Kogai explained that FreeBSD comes with an implementation of malloc() that is optimized for paged memory, and safe from duplicate free() calls. But the downside is that realloc() is very slow. That's usually not a big deal, because most programs don't use realloc() very often -- but perl does. (The default configuration of perl on FreeBSD is to use perl's internal malloc, that hasn't this realloc limitation.)

James Jurach provided a patch to use 32 bit integers instead of 16 bit integers for line numbers. As it didn't seem to increase the size of cops on several platforms, it went in.

Steve Grazzini produced a patch for the my SLICE syntax we discussed last week. Its implementation was criticized ; and it wasn't applied.

As Randal L. Schwartz reported, the initial value of @INC (and directory installation scheme) of perl on MacOS X has to be cleaned up, as it causes problems with make install UNINST=1 (a sub-optimally documented MakeMaker feature, used to uninstall previous versions found in @INC).

Vadim Konovalov provided more WinCE portability patches, with threads and fork emulation.

Tels pre-released Math::BigInt v1.65, Math::BigRat v0.10, and bignum v0.14. Michael Schwern released MakeMaker 6.10_07.

Perl 5 turned 20000 patches old.

About this summary

This week's summary was written by Rafael Garcia-Suarez. Weekly summaries are published on http://use.perl.org/ and on a mailing list, which subscription address is perl5-summary-subscribe@perl.org. Comments, corrections, additions, and suggestions are (as always) welcome.