18:55 <janneg> caveat! firefox started using jemalloc in 120 and if that's not disabled like on fedora it will result in unuseable binaries when built on a 4k host machine

18:56 <janneg> at least debian and alarm are affected

18:56 <kidplayer666> janneg: is any other distro affected?

18:56 <kidplayer666> cause I'm on fedora and its fine

18:56 <janneg> fedora is fine, everything else is probably affected

19:00 <mps> hm, I cannot build firefox 120.0 on alpine because some problems with python and I don't know how to fix python

19:01 <mps> on aarch64 I mean. on other arches it builds

19:02 <mps> https://gitlab.alpinelinux.org/alpine/aports/-/jobs/1190685#L3419

19:04 <leio> I suppose if you build it yourself, you are likely to have a 16k page on the host doing the compilation and then not be affected, as I understood it

19:05 <psykose> mps: that 'Firefox exited with code -11 during profile initialization' means it segfaults when doing profiling

19:05 <mps> leio: no, it is on 4K host

19:06 <mps> psykose: hm, it works on other arches

19:07 <mps> but you are right, it doesn't failed 'on python' but later

19:07 <mps> I didn't read log carefully

19:12 <psykose> janneg: i'm pretty sure firefox has used jemalloc for multiple years, not just 120

19:13 <psykose> that --disable-jemalloc has been a thing for a very long time :D

19:15 <janneg> psykose: yes, --disable-jemalloc replaced a patch in the fedora spec 5 years ago

19:16 <mps> I remember that for long in alpine

19:18 <mps> and arguing with some people to not allow jemalloc for firefox (and not only firefox)

19:29 <janneg> oh, it's not actually jemalloc fault. https://searchfox.org/mozilla-central/source/memory/build/PHC.cpp#1076

19:30 <j`ey> but that causes a segfault??

19:35 <janneg> no, the SetPHCState() call segfaults since gMut is NULL, at the end of the file

19:38 <j`ey> right..

19:46 <janneg> lol: https://searchfox.org/mozilla-central/source/memory/build/PHC.cpp#295

19:47 <j`ey> :'(

22:14 <mps> just tested libreoffice on alpine aarch64, it works on machine with 4K page size but hangs on start on 16K page size kernel (m1pro and m1). same build, same binary package

22:14 <mps> and mesa is different