This week, the P5P report features the Great Module Renaming ; the Second Great Module Renaming ; a Couple Of Bugs, old and new ; and, finally, Perl 5.8.0 RC 3 released to the world.
On monday morning, the version of MIME::QuotedPrint included with Perl still
provided a few methods (namely PUSHED, FILL, WRITE and FLUSH) that added
to it the ability to be used as a PerlIO layer. Elizabeth Mattijsen,
hunting bloat, submitted a patch to remove these methods. Noticing that
these were used in the test for PerlIO::Via, and in some documentation as
code examples, Jarkko decided to borgify Elizabeth CPAN's module,
PerlIO::Via::QuotedPrint, which is essentially the above PerlIO methods
put in a separate namespace.
On the path of doing this namespace migration, a PerlIO bug reappeared :
open($fh, "<:Via(foo)", "file") not returning false on failure. It
turned out that this was a bug in the:stdio pseudo-layer. It was
promptly corrected by Nick Ing-Simmons.
Continuing on the PerlIO::Via topic, Elizabeth Mattijsen asked why the
second part of PerlIO::encoding was lowercase, whereas PerlIO::Via and
PerlIO::Scalar where uppercase.
This was obviously a good question. A consensus was found quite rapidly :
all core IO layers should have all-lowercase names. Thus, PerlIO::Via (and
its submodules PerlIO::Via::*) moved to PerlIO::via (note the lowercase
'v'), and PerlIO::Scalar to PerlIO::scalar.
During the previous discussion, Nick Ing-Simmons noticed that is was
impossible to pop an IO layer from within Perl. Thus he added a:pop
experimental pseudo-layer, that does just that.
Ilya Martynov tried to backport RC1's Data::Dumper to perl 5.6.1.
He found out that some tests were failing, because the internal hashing
algorithm used by perl has changed, and thus hash keys were not returned
in the same order. The trick should be to have version-specific tests, or
to use some more clever way to verify the results. Anyway, Ilya is now
with Gurusamy Sarathy co-maintainer of Data::Dumper, and will be able to
upload new versions to CPAN.
Simon Cozens found that the DESTROY method was not called when blessed
code references go out of scope. Nobody replied, and it's still not clear
whether this is a bug.
defined $$foo Elizabeth Mattijsen noticed that
use strict 'refs';
our $foo = 'bar';
if (defined $$foo) { ... }
was valid perl code, and wondered whether this was an undocumented feature. After some discussion, it was decided that this was actually a bug, but the kind of bug that can wait until after 5.8.0 -- it's non-critical, and it's present in 5.6.x.
Abigail reported a couple of bugs. The first was a complaint about/(?{...})?/ not being allowed. This triggered a terse comment by Hugo
about how he wants to enhance the regexp engine during the 5.9
development track.
The second bug was about local()-izing an $#array variable not working
correctly (for some value of "correctly"). (I think that this bug was
already reported in the past, but I didn't find it.) Anyway, if we want to make
this work, the semantics of this construct should be precisely defined.
This goes in the to-do list for 5.9.
Elizabeth Mattijsen was also busy on the ithreads front. She uploaded to
CPAN a thread-pool module. The first version of it was called
threads::farm (now it's Thread::Pool). Johan Vromans noted the
all-lowercase name, quite inappropriate for a module. It's worth noting that
this triggered the Second Great Renaming of the week : the thread utility
modules now live in the Thread:: namespace. The thread pragmatic modules
(threads and threads::shared) keep their lowercase name, in
conformance with the general naming conventions; but
threads::shared::queue and threads::shared::semaphore now become
respectively Thread::Queue and Thread::Semaphore.
Jarkko released RC3 to the world, but you already know that. He said also
The post-RC3-plan is no damn code patches (or additions), the possible exceptions being firefighting by me (Unicode), NI-S (perlio), Arthur (ithreads), or Hugo (regexen).
and
Timeframe for the 5.8.0 final: Thursday.
(this last word being of course subject to Murphy's law.) (It also means that you have to quickly download and test RC3 on your favorite platforms before it's too late...)