On Wed, 4 Jul 2007 22:56:48 +0000 [EMAIL PROTECTED] wrote: > > This is on gentoo linux x86. I ran Configure.pl with no arguments. > > Apparently this test passes on chromatic's linux box, and on > > particle's win32 box, but fails on my gentoo box and particle's > > ubuntu box. So its a bit sporadic. > I just tried this test case with a fresh parrot from svn on a sparc > machine running gentoo and got the first set of output. > > I don't know if this helps or confuses.
It helps, thanks. Glad to know I can't just blame gentoo. After doing the binary search I mentioned earlier and finding that it started breaking in svn r19441, there was some discussion in the IRC channel. The important bit is: [13:35] <@chromatic> Hm, I don't blame 19441 for the shootout test. [13:35] <@chromatic> I think it just changed the memory characteristics of the program sufficiently to trigger a lurking GC bug. So now I'll try to see if there was a DOD run that freed up something it shouldn't have, or something like that. Thanks for testing! Mark
