On Fri, 7 Nov 2008, Ganesh Sittampalam wrote: > On Fri, 7 Nov 2008, Eric Kow wrote: > >> This appears related to issue1017 because the foreverness is happening >> on the the test for that issue. >> http://allmydata.org:13010/builders/jules%20apollo%20Darwin-9.2.2%20Mac%20OS-10.5.3%20i386/builds/197/steps/test/logs/stdio >> > I can't see any reason for it to loop :-( It should either terminate > reasonably fast, or if something isn't fixed properly, then stack > overflow. > > Not sure what to suggest beyond sticking in some traces or using the ghci > debugger on the machine where it fails. If I can get access to it I > could do that.
Jules has kindly given me an account so I can investigate further. Unfortunately, when I try with what I think is the version of darcs that things went wrong, I can't reproduce the problem - the test passes fine, and quickly, as I would expect. There are a couple of areas of slight confusion in my attempt to reproduce the test circumstances, referring to the main page for the test at: http://allmydata.org:13010/builders/jules%20apollo%20Darwin-9.2.2%20Mac%20OS-10.5.3%20i386/builds/197 1) Why do steps 11 and 12 mention ghc 6.6 on that main page? configure seems to be finding ghc 6.8.3, and that's all that I can find on the machine. Also, build 200 on the same machine doesn't say anything about this. 2) Is there a hidden clean-up between steps 8 and 9? I can't see one, but without it the second build won't work, as it tries to reuse the type witness .o and .hi files. All-in-all, it's still rather puzzling :-( Ganesh _______________________________________________ darcs-users mailing list darcs-users@darcs.net http://lists.osuosl.org/mailman/listinfo/darcs-users
