-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi Michael,
thanks a lot for looking at this. it sounds like a fairly safe option, given how awkward and ancient this code is. i will give it a go and upload it in a few days. best, Paul On 22/11/2014 15:14, Michael Banck wrote: > tags 769264 +patch thanks > > Hi, > > On Wed, Nov 12, 2014 at 11:45:28AM +0100, Lucas Nussbaum wrote: >>> ./fusebirth > fused_loop.c 2>/dev/null make[1]: *** >>> [fused_loop.c] Error 139 > > So what happens here is that fusebirth segfaults on i386 in > topo_sort(), while trying to sort whatever in order to generate > fused_loop.c. > > Tracing through topo_sort() with a debugger, it seems that after a > dozen or so recursions into it, get_children(node) returns an > invalid (but not NULL) pointer, and on the next iteration we get a > segfault in it. > > I stared at the code for a few hours, but it doesn't look like this > is how GLib is supposed to be used nowadays, so I went for an > alternative solution: I just included the auto-generated .c code > into the source package and changed the build system to not > generate/delete that .c file. > > I made sure the generated source is the same on amd64 and s390x and > will test on a couple more architectures to make sure. > > Proposed Debdiff attached. > > > Michael > -----BEGIN PGP SIGNATURE----- iKYEARECAGYFAlRw8ApfFIAAAAAALgAoaXNzdWVyLWZwckBub3RhdGlvbnMub3Bl bnBncC5maWZ0aGhvcnNlbWFuLm5ldEJEMkE4NDdFMkMzMUJDNzg4NjMxQ0RFNTky RTBCREU3QzYwMDJDQkQACgkQkuC958YALL0SnACfaD7nuZdOTT3OUuAG8j6bQF40 IoMAn19WyR25AK5KP1Mr1cHI9TlWIPPd =1Iy/ -----END PGP SIGNATURE----- -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
