Tim Bunce wrote: > On Thu, Jul 16, 2009 at 07:32:26AM -0500, Elliot Shank wrote: >> Tim Bunce wrote: >>> If you grep the log for ^New you'll see that it doesn't load in >>> many modules before it crashes. Looks like it only got to line 9 >>> of Build.PL and line 12 of Custom/Module/Build.pm, before finally >>> dieing in strict::unimport called from Fatal::_make_fatal. So it >>> seems likely that autodie is involved.
Line 12 is, unsurprisingly "use autodie qw< :all >;". >>> Could you try to cut that down to an as-small-as-possible >>> testcase? >> >> This was the simplest example in our codebase that I could >> immediately think of and could reproduce the problem with. There's >> no Moose involved. :] > > :) Lo, and behold, with r832, everything works with this Build.PL. Unfortunately, the things I'm really interested in still die. (More below.) >>> cpantesters shows passes for Strawberry perl on 5.8.9 and 5.10.0 >>> http://matrix.cpantesters.org/?dist=Devel-NYTProf%202.10;reports=1;os=MSWin32 >>> Does make test pass for you? >> Yes, the entire test suite works, both 2.10 and r812. The only >> failure that I was getting was test16 with r813-r828. > > So the clock must be working. Odd. Perhaps the printf warnings have > fixed a problem that was causing 0's to be shown in the log. Everything is clean with r832. > Perhaps you could persuade the Strawberry folk to make a debugging > build available. So I'm now building from the stock 5.10.0 tarball using the MinGW from Strawberry with -DDEBUGGING set. We'll see whether I can reproduce it. ====== Once "perl Build.PL" worked, I tried something closer to what I want to look at for real. There's a simple program that exercises a small part of the functionality. I tried profiling "foo --usage"; this obviously doesn't do much processing, but it does cause all the code to load and compile. It worked, with a bunch of noise: xsub/builtin exited via an exception (which isn't handled yet) xsub/builtin exited via an exception (which isn't handled yet) Attempt to free unreferenced scalar: SV 0xdd8394, Perl interpreter: 0x3e254c at c:/.../lib/perl5/MSWin32-x86-multi-thread/Class/MOP/Class.pm line 625. xsub/builtin exited via an exception (which isn't handled yet) xsub/builtin exited via an exception (which isn't handled yet) Attempt to free unreferenced scalar: SV 0xdfa2d4, Perl interpreter: 0x3e254c at c:/.../lib/perl5/MSWin32-x86-multi-thread/Class/MOP/Attribute.pm line 405. xsub/builtin exited via an exception (which isn't handled yet) Attempt to free unreferenced scalar: SV 0x1830c54, Perl interpreter: 0x3e254c at c:/.../lib/perl5/Some/Module.pm line 741. xsub/builtin exited via an exception (which isn't handled yet) Attempt to free unreferenced scalar: SV 0xdfd57c, Perl interpreter: 0x3e254c at c:/.../lib/perl5/Some/Module.pm line 872. Usage: foo --option ... Unable to determine line number in generated method (unknown origin) Unable to determine line number in generated method (unknown origin) So, now to attempt a test that does a small subset of what I'm looking at, i.e. "perl -d:NYTProf t/foo.t". We get Attempt to free unreferenced scalar: SV 0x120eb44, Perl interpreter: 0x3e27b4 at c:/.../lib/perl5/MSWin32-x86-multi-thread/Class/MOP/Class.pm line 625. xsub/builtin exited via an exception (which isn't handled yet) xsub/builtin exited via an exception (which isn't handled yet) Attempt to free unreferenced scalar: SV 0x1233d54, Perl interpreter: 0x3e27b4 at c:/.../lib/perl5/MSWin32-x86-multi-thread/Class/MOP/Attribute.pm line 405. And then perl chokes. ====== I'll have a new perl binary waiting for me in the morning and we'll see whether I can reproduce this with that. I'm going to be very unhappy if I can't. --~--~---------~--~----~------------~-------~--~----~ You've received this message because you are subscribed to the Devel::NYTProf Development User group. Group hosted at: http://groups.google.com/group/develnytprof-dev Project hosted at: http://perl-devel-nytprof.googlecode.com CPAN distribution: http://search.cpan.org/dist/Devel-NYTProf To post, email: [email protected] To unsubscribe, email: [email protected] -~----------~----~----~----~------~----~------~--~---
