Hello Seb, Le mercredi 11 janvier 2012 à 02:16 +0100, Sébastien Wilmet a écrit : > On Wed, Jan 11, 2012 at 12:16:41AM +0100, Matthieu Baerts (matttbe) wrote: > > @Sebastien, any idea? :) > > I looked at the Vala code and at the C code generated, and I don't > understand what goes wrong. > > Is the bug always reproducible? Or the bug occurs only randomly? > > Is the package in the PPA compiled with the C code generated from the > tarball, or the Vala code is recompiled? If the C code is regenerated > from Vala, which valac version exactly? This package is compiled only with the C code available in the tarball. > So a possibility is that there is a dangling reference (the variable > "self" is still reachable but has been reclaimed). In this case, it's a > bug in Vala. But when reading the C code, I don't see why the variable > would be reclaimed… Also, with a dangling ref, the bug should occur > randomly, and should occur with the version compiled from the tarball > too. But it's strange because we use the same C code. > Maybe the compilation options of GCC? Which version of GCC? But the > chance that the problem comes from the "-O2" optimization is very very > small. Yes, maybe... Matt
Attachment:
signature.asc
Description: This is a digitally signed message part