This was never an issue for me on 24.11 (or any previous release), but since updating to 25.05 something like 10% of my local builds randomly fail with a segfault from GCC. It seems to be nondeterministic. I know the GCC major version bumped from 13 to 14 between releases, but before I go crazy with downgrading my stdenv for all of my locally-built packages, I wanted to see if anyone else is observing the same thing (or maybe my hardware is at fault… I sure hope that isn’t it).
Depends on the error. Is it OOMing?
I don’t think so; plenty of system RAM still available, and I keep the number of parallel jobs I build low.
I also experienced random GCC crashes (usually an internal compiler error in a random subroutine). Eventually I ran a memtest and found one (!) faulty bit of RAM. The crashes went away after I replaced that RAM module.
2 Likes
Thanks, that’s a good idea. I’ll try it on my next reboot.
I remember this very well, it was on this review: z3: 4.8.17 -> 4.13.4; z3_4_12: 4.12.5 -> 4.12.6 by numinit · Pull Request #327052 · NixOS/nixpkgs · GitHub
It wasn’t enough that rebuilding everything downstream of z3 takes up hundreds of gigabytes of RAM, all that RAM had better be good. Thank you for your efforts on that PR. 