This Week on perl5-porters (11-17 August 2003)

This Week on perl5-porters (11-17 August 2003)

The next maintenance release of Perl approaches, but the porters want to take the time to do it right. Meanwhile, discussions and bug reports continue to occur, as usual. Read about new and old documentation, valgrind, backwards [in]compatibility, and other stuff.


The perlreftut manpage is currently distributed with Perl under a restrictive license that doesn't make it completely freely redistributable and that is apparently not compliant with the Debian Free Software Guidelines. Its copyright is currently owned by The Perl Journal, but Mark-Jason Dominus (author of this document) is trying to obtain it. If he succeeds he will rerelease it under a more standard license.

Meanwhile, a new manpage, perlcheat (the perl cheat sheet), crafted by Juerd, was added to the perl core documentation.


Jarkko Hietaniemi added a test.valgrind make target, to run the core test suite under valgrind, a memory access debugger for Linux/x86. This shakes out some errors, indicating potential or actual bugs.
    valgrind :


Elizabeth Mattijsen is heavily testing the perl 5.8.1 release candidates with her thread (and non-thread) modules. She's able to produce error messages like Attempt to free unreferenced scalar or Scalars leaked -- notably with the Thread::Tie and Thread::Pool modules.

She also notes that, now that warnings are automatically enabled when running tests (unless otherwise specified), test suites that used to seem perfectly working could be now proven flawed. (Not mentioning potential failures due to random hash key ordering.)

PerlIO layers

Stas Bekman, facing a problem with the handling of STDIN and STDOUT with mod_perl 2 when there are PerlIO layers pushed on those handles, asks for a way to replace the worker PerlIO layer. Nick Ing-Simmons explains some of the details about the PerlIO internals.

Scripts for porters

Jarkko decided that it would be a good thing to bump up all version numbers of core modules that have changed since 5.8.0. Slaven Rezic provided a script to find in two Perl source trees modules that have changed but have the same version numbers. (Slaven also wrote another script to identify all modules with a dual life on CPAN, whose Makefile.PL's do not contain something like INSTALLDIRS => perl.

Jarkko also asked for a script to regenerate [parts of] all files that use a list of the standard documentation pod files (various makefiles, table of contents generator, etc.) Nicholas Clark studied the problem and posted the state of his unfinished work so far.

In Brief

Rafael Garcia-Suarez asks for more regression tests, notably for some of the basic parts of the interpreter that might be undertested and/or undocumented (for example, numeric/string conversion). This would most probably be useful for Ponie.

Abigail finds that the snippet $x=~(0,0) produces an obscure error message : Too many arguments for regexp internal reset. (Bug #23328). But should it be a syntax error at all ?

Alan Burlison, who set up an impressive smoke suite on several different Solaris boxes, gets random failures from them : some of them were apparently cured by using a non-parallel make for make _test.

Hugo explained why some long-deprecated features are not likely to be removed.

gcc 3.3.1 was released ; it doesn't seem to cause major problems.

About this summary

This week's summary was written by Rafael Garcia-Suarez. Summaries are published weekly on and on a mailing list, which subscription address is Corrections and comments are welcome.