As September begins, the perl5-porters, ignoring the changing weather, continue to work. This week, some small things, and a few bigger ones, are selected in the report. Read below.
Elizabeth Mattijsen found that, while executing a CLONE method (the hook
to support cloning of data when new threads are created), threads->tid
returns the thread id of the original thread. This is because the cloning
is actually performed (at the C level) in the parent thread.
However, to execute some routine when a new thread starts, and from this
new thread, a workaround is to hijack threads::new. Nick Ing-Simmons
also suggested to implement some sort of post-clone handler via another
special subroutine (for example CLONED).
John Peacock, the man who courageously continues to struggle with v-strings, posted a restatement of current objections to version objects -- version objects being what the v-strings should be compiled to internally by perl, in order to make them less unusable.
The thread is worth a read if you are annoyed by v-strings and if you want to have some idea on how they're likely to change.
http://groups.google.com/groups?threadm=...
To quote Hugo, Larry continues to like them [the v-strings] despite various attempts to convince him otherwise, so they are here to stay. I think they could become much less of a bugbear if John's sterling work culminates in a consistent and appropriately documented set of behaviours. Waiting for the v-string objects Apocalypse...
Now that pseudo-hashes are removed from bleadperl, Gurusamy Sarathy noticed that using them doesn't trigger any deprecation warning in perl 5.8.0. Jarkko is considering adding one in perl 5.8.1.
Benjamin Goldberg also started a thread on the implementation-to-be
of the fields pragma in perl 5.10.
http://groups.google.com/groups?threadm=...
Jarkko Hietaniemi found out that Perl 5.8.0 doesn't compile on Mac OS X
10.2 with the August update of the Developers Tools, due to a small
communication problem between the Configure script and Apple's modified
gcc 3. He also provided a fix for that, which you can grab here if you
need it
http://www.xray.mpe.mpg.de/...
Last minute update : has Jarkko dreamed ? have the sunspots moved ? He's apparently no longer able to reproduce this bug.
Abe Timmerman is now the official maintainer of Test::Smoke. See http://qa.perl.org/ for more info about smoke testing.
Steve Piner reported an (already known, IIRC) bug : my $x; our $x;
doesn't generate a warning, whereas our $x; my $x; does. (Bug #16922.)
H.Merijn Brand, doing some speed experiments, found that perlio has slowed down considerably compared to stdio from 17805 to 17821. No confirmation or comment was issued.
Allen Smith provided some patches to bleadperl and to perl 5.8.0 to
improve IRIX support, namely on long doubles, and their handling by
s?printf().
The perl5-changes mailing list now carries the changes committed by Hugo, and all previously dropped changes have been sent back to it, for the purpose of archiving.
Sadahiro Tomoyuki released a new version (0.23) of Unicode::Collate.
Elizabeth Mattijsen, after having uploaded a first version of
Thread::Exit on CPAN, found that POSIX::_exit() makes perl hang on
Linux when called from within a thread (bug #17057). This problem is
probably related to the Linux implementation of threads.
Elizabeth emitted the idea to start a new FAQ section on threads.
Hugo is leaving for Zurich (the Perl 6 mini-conference) and to Munich (YAPC::Europe).
This summary brought to you by Rafael Garcia-Suarez. It's also available via a mailing list, which subscription address is perl5-summary-subscribe@perl.org . As usual, comments, corrections and additions welcome.
Re: Jaguar and Camel
jhi on 2002-09-09T13:58:36
> reproducibly brokenIf I may point out a bit more...
:-) I wouldn't call the build "broken", just "noisy". It builds and tests fine (modulo the "standard" Berkeley DB failures). It's just that because the Configure as it was before my patch and the new gcc setup of Apple don't communicate that well, one gets a noisy (two-line) warning about -I/usr/local/include for each and every compiled C file. With the patch, one doesn't.
Re: Jaguar and Camel
rafael on 2002-09-09T14:05:50
OK, next time I won't start the weekly summary on sunday evening -- at least not too late on sunday evening.Re: Jaguar and Camel
Theory on 2002-09-09T17:19:56
It's just that because the Configure as it was before my patch and the new gcc setup of Apple don't communicate that well, one gets a noisy (two-line) warning about -I/usr/local/include for each and every compiled C file.Interesting. I found this problem on Jaguar with 5.8.0 and the original 10.2 build tools, not just the latest ones. It only happened when I actually had a
/usr/local/include directory. I was going to post a bug, but never got round to it. Thanks, Jarkko! --David