This Week on perl5-porters - 17-23 October 2005
Perl 5 was released this week, eleven years ago.
Defining "sub _ {...}" breaks "File::Find"
Jeff "japhy" Pinyan mentioned the Issue That Will Not Die, that if you
define "sub _ {...}", then "_" as a shorthand for *the file most
recently statted* stops working. The problem at hand for japhy being
that "File::Find" fails, and wondered what workarounds were possible.
Mark Jason Dominus was against the idea of teaching "File::Find" to
work around such damage, because it would probably not be the only
module that breaks under this scenario. The best fix would be to
adjust the parser to prefer string context for "_" where it makes
sense (and thus not call "sub _"). And Rafael Garcia-Suarez did just
that with a quick patch to put more DWIMery into the tokeniser in
change #25799.
It turns out that "_" is the usual idiom in the "gettext" world for
dealing with internationalisation (i18n). Yitzchak Scott-Thoennes
thought that a new warning should be generated for attempting to
define "sub _". Abigail thought not. The thread then evolved into a
discussion about how and when new warning can, or should, be added to
the interpreter.
Earlier thread, summarised here
http://dev.perl.org/perl5/list-summaries/2005/20050909.html
This time around
http://xrl.us/h5ye
Musing on warnings
http://xrl.us/h5yf
Loading multiple versions of modules
Nicholas Clark wants to know how much tweaking to the internals would
be required in order to let perl load different versions of the same
module. Something as mundane as package Alpha using version 1.0 of
module Charlie, and package Bravo using version 2.0 of module Charlie
would come in quite handy. Without getting into the brain-melting
complexity of an @array containing heterogeneously versioned Charlie
objects...
The idea sounds quite exciting, sadly, after a few responses the
thread flickered out.
http://xrl.us/h5yg
Sub return values are inaccurately documented
The monster thread of the week, with 44 replies. It starts with a
sentence from "perlsub", noted by Xavier Noria: "The return value of a
subroutine is the value of the last expression evaluated by that sub".
Pop quiz: what does the following return?
C<sub f { 1 for 1 }>
Answer: ''. Adriano Ferreira supplied a number of variations on the
theme, with "if" or "while" in the place of "for", and Abigail pointed
out that "do {...}" has the same sorts of problems, so it's not only
apparent in subroutines. The problem is that "for" is a control
structure. And the value of a control structure is unspecified. In
fact, it doesn't have one.
Yves Orton noticed that "my @b=(0,1); print $b[0+do {1 for 1}]" dumps
core. He also wanted to specify the last expression produced by a
control structure. Xavier pointed out that control structure are not
expressions. And Rafael concurred, saying that you can't say "my $x =
while (...) { ... }". Yves countered with "my $x = do {while (...) {
... }}". Rafael put his foot down and said that loops are evaluated in
void context. To change them now to be evaluated in scalar or list
context (which is what Yves wanted) would break lots of code.
But the core dump problem remained unresolved.
The beginning
http://xrl.us/h5yh
Dave explains exactly what's going on
http://xrl.us/h5yi
Rafael explains from the other end
http://xrl.us/h5yj
Xavier summarises the current state of play
http://xrl.us/h5yk
"do { EXPR for EXPR }"
Robin Houston picked up the core dump noticed by Yves in the above
thread and analysed it. And patched the source to fix it.
http://xrl.us/h5ym
List slice subscripting
Nicholas Clark said that he didn't want to see the new slicing syntax
appearing in "maint", and related a horror story of battling with
syntax variants in 5.004, when only 5.004_05 compiles easily on modern
platforms due to compiler evolution (specifically, Configure-back-then
gets confused by what compiler-right-here-right-now says in response
to its probes).
If you like this syntax, wait for 5.10. If you can't wait, try and
help to make 5.10 get here faster.
http://xrl.us/h5yn
Initially summarised here
http://dev.perl.org/perl5/list-summaries/2005/20050915.html
Recursive comparison routine can cause segfault in sort
Robin Houston found that using a sort comparison routine that calls
itself can cause perl to dump core. Dave Mitchell confirmed that the
situation is deeply unsatisfactory. Even if the code base was patched
to die gracefully when a recursive sort comparison was found, there
are problems with threads that remain in any event. Hugo van der
Sanden suggested an approach using attributes. Robin and Dave started
to debate a way forward.
The bug report
http://xrl.us/h5yo
The trouble with threads
http://xrl.us/h5yp
Hugo's suggestion
http://xrl.us/h5yq
Robin's way forward
http://xrl.us/h5yr
What John E. Malmberg did this week
John E. Malmberg got "open(FOO, "child.pl foo|")" working on VMS,
applied by Craig Berry,
http://xrl.us/h5ys
and synced "exit"'s behaviour with the documentation
http://xrl.us/h5yt
and posted his VMS TODO list
http://xrl.us/h5yu
There's a lot there. And it probably may not be complete.
What Jarkko Hietaniemi did this week
Jarkko Hietaniemi posted many patches to bring Perl on Symbian up to
speed.
http://xrl.us/h5yv
http://xrl.us/h5yw
http://xrl.us/h5yx
http://xrl.us/h5yy
http://xrl.us/h5yz
In brief
I have the Perl5 bug summary in my spool, but for some reason I cannot
find the message on "xray". There were 1507 open tickets as of
2005-10-17.
Nicholas Clark thinks he shook out a reference counting bug with a
ponie build:
http://xrl.us/h5y2
Ilya Zakharevich cc'ed p5p about a bug in "if" and appended a patch.
Some doubts were expressed about the error message (not everyone may
catch the cultural context of "cryptocontext"). Others noted
tangentially that testing for Windows platforms with "if ( $^O =~
/Win/i )" will come to grief on "cygwin" and "Darwin".
http://xrl.us/h5y3
Andy Lester, continuing on his quest to conts, hoisted some repeated
code out in av.c and thus shaved a few bytes of the resulting object
code.
http://xrl.us/h5y4
Following on from the discussion of manipulating the environment last
week, H.Merijn Brand added a configure probe for "clearenv".
http://xrl.us/h5y5
Sébastien Aperghis-Tramoni wrote patch #25802 for autodoc.pl so that
it generates the index entries for perlapi.pod and perlintern.pod.
http://xrl.us/h5y6
Gisle Aas fixed up "pp_sselect"
http://xrl.us/h5y7
Nicholas upgraded "ExtUtils::MakeMaker" to 6.30 in "maint". And wanted
to know what (is)? broke(n)?.
http://xrl.us/h5y8
About this summary
This summary was written by David Landgren, for once without Immediate
Realtime Corrections (IRC), so any typos or wordos are truly my own.
I'm offline as of tomorrow morning for a week. Yay!
Information concerning bugs referenced in this summary (as #nnnnn) may
be viewed at http://rt.perl.org/rt3/Ticket/Display.html?id=nnnnn
Information concerning patches to maint or blead referenced in this
summary (as #nnnnn) may be viewed at
http://public.activestate.com/cgi-bin/perlbrowse?patch=nnnnn
Weekly summaries are published on http://use.perl.org/ and posted on a
mailing list, (subscription: [EMAIL PROTECTED]). The
archive is at http://dev.perl.org/perl5/list-summaries/. Corrections
and comments are welcome.
If you found this summary useful or enjoyable, please consider
contributing to the Perl Foundation to help support the development of
Perl.
