http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51727
Tobias Schlüter <tobi at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #28410|0 |1 is obsolete| | --- Comment #13 from Tobias Schlüter <tobi at gcc dot gnu.org> 2012-10-11 13:03:25 UTC --- Created attachment 28424 --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28424 Patch for testing This patch should make output deterministic. Joost, can you please try it out? I would also be interested in the compile-time impact on module-heavy code. Because the pointer_info tree can change during writing, I sometimes have to sort the tree several times, and I have no idea if this can become important in files with large modules. Actually, now that I wrote this, I can already think of a trivial speedup, which I will implement before the final submission. Nevertheless, I would still be interested if the change disappears in the noise.