5.8.8 closer to working
Posted by mcummings
OK, so i think i identified the bug with makemake that was killing pdl, patch submitted - not sure of complete ramifications, so it won't be hitting gentoo anytime soon. the only other problems i've had with 5.8.8 have been with swig, svn, and wxperl - but before anyone gripes, there is a patch (hopefully destined for 5.8.8 rc2) that fixes the bug for swig, which in turn fixes the bug for svn, and with luck maybe wxperl will get pulled in there to 
other than that...nothing to complain about. did some reworking of the perl-cleaner script, trying to tone it up a bit so it has a smoother run through. ok, off to make dinner for the kids, later all
makemaker bug (i think)
Posted by mcummings
the battle with 5.8.8 continues. i think i've identified a bug in eu::mm 6.30* which results in perlrun being used when perlruninst should be invoked, whichbreaks things like pdl with its multilevel makefiles. i've submitted the bug upstream since this would need to be fixed upstream, but its the kind of thing that will delay getting 5.8.8 out of package mask. i suppose 5.8.8 will also be a good time to fix some collisions with perl-core ebuilds via pdepend...
perl 5.8.8 rc1 – oh the joy
Posted by mcummings
I added the rc1 for perl 5.8.8 to the tree today (package masked), chiefly to give folks a chance to play and see what's broken before the final release gets pushed out.
Imagine my unending joy to discover that the jump in ExtUtils::MakeMaker versions will again visit us with some snags. Not nearly as trecherous as the 5.6.x => 5.8.x perl releases, but I've already hit a few bugs. For starters, the new EU::MM is far more particular about stray spaces in the Makefile.PL's - not necessarily an unexpected tightening, but still a surprise gotcha in at least one module out there (gtk2-ex-printdialog iirc).
Right now I'm struggling with PDL, which built fine under 5.8.7's makemaker, but not anymore (weird lost blib issue in the source, only affects the Demos code...makes ya wonder if they should just be clobbered). So far PDL is my only unfixable, everything else has been pretty straightforward.
Bumps are grand...
Apology for tree breakage
Posted by mcummings
OK, this is absolutely the last place I'm going to post this, but since its been so long since I did anything to break people's boxes, I figured one more place couldn't hurt. The following is the email I posted to -dev on the subject this morning. If people are still yearning for the smell of blood, bah, I've got none left to give.
===============
Protocol doesn't demand it - but human decency sure as heck does. I just wanted to apologize to everyone for my erroneous commit to the 1Q-2006 file yesterday.
The boring background is that some upstream cpan authors use a numbering scheme that makes sense in a decimal fashion, where .29001 is less than .30 (expand the 0's, yada yada yada, ask your kid sibling about math at school today), but portage expects that numbering scheme to actually be a version identifier (so to speak, its the gist I'm going for here, don't flame me on the technicalities of how portage reads a version number). dev-perl/DateTime came out with a .2901 release, which I put in the tree as such many a moon ago (about 1.4 moons ago to be exact), and recently .30 came out. Because of the math business I just mentioned, .30 would never get picked up by anyone since (dropping the decimal point) 2901 > 30. I 'moved' .2901 to .29.01 to make this work better. All was good.
Then I had this brilliant idea - use the quarterly file to help people migrate that number, no bizarre downgrades when the emerge world. 'cept move in the quarterly isn't for moving a version, its for moving between categories, package names, etc., and as such I made a mistake.
No need to respond to this message (i'd prefer it if you didn't actually - between the 'stuff happens' and the 'your an idiot, revoke that bad boys right to a keyboard' messages, i think it would all even out).
~mcummings
353: My magic number for one day
Posted by mcummings
That's right - *blink* - 353 cvs commits all said and done today, and so far not one related bug submitted on it (must be a record somewhere for something). The first 100 or so were for bumping versions of packages, which means unless there's a slew released on cpan this weekend, Monday's weekly email about the number of packages out of date in dev-perl should be down to the handful of mistakes (I can't figure out why the report can't figure out they're not up to date, and really don't want to waste the energy on it). The other 250+ were package stable bumps - most of them for packages that have been sitting for over 100-600 days for no better reason than neglect. About time dev-perl wasn't a wasteland anymore, eh?