https://bugzilla.suse.com/show_bug.cgi?id=1202027 https://bugzilla.suse.com/show_bug.cgi?id=1202027#c8 --- Comment #8 from Richard Biener <rguenther@suse.com> --- (In reply to Martin Li��ka from comment #7)
Ok, I've just made 2 builds with profiledbootstrap and --with-build-config=bootstrap-lto on a machine where ASLR is disabled but still we face so arc counter differences in hash table related functions like htab_find_slot_with_hash, htab_mod, ... Well, very many hash tables are pointer-based on based on the address hash is different and so table collisions are not stable. I think it's an unrealistic assumption that the memory layout of 2 running compiler invocations will be always the same, right?!
The gcc package also disables ASLR. But yes, I think that with ASLR disabled pointer hash table behavior should reproduce exactly. At least I can't find a good reason why it should not? Btw, I wanted to check whether the issue reproduces also without LTO ... -- You are receiving this mail because: You are on the CC list for the bug.