On Mon, Jul 8, 2019 at 5:30 PM Lars Ingebrigtsen wrote: > I'm still seeing this... but not after every recompilation. If I say > "make bootstrap", then Emacs usually works. If I then recompile parts, > I sometimes get a non-working Emacs. Are you on master? As far as I know, no fix has been applied there yet. Also, if I understand correctly, the bug depends on ASLR both at the time of dump and in the final Emacs image. Also also, can you please stash copies of the emacs binaries and emacs.pdmp that fail to work? I did something very silly and overwrote my compiler with a very slightly different version (after carefully backing up /usr/local/bin/gcc, too. Too bad that's just the driver program), and now can no longer reproduce the issue easily. > Perhaps the original change should be reverted until you've figured out > how to fix this completely? It's very annoying. It is. I think it's worth the risk to push my original fix (reattached to this message) now, but I don't have commit privileges so it's not my decision to make. Alternatively, we could simply disable purify-flag hash consing for now and everything should work, I think. But I don't think there's a clear candidate for "the original change". The bug's been with us for a while, possibly since pdumper was introduced, but it got significantly worse when the recent byte compilation changes were committed; however, those changes are actually okay. Is there anything I can do to help get this bug fixed on master? Thanks!