jaeger changed the topic of #crux-devel to: CRUX (https://crux.nu/) development channel | Logs: https://libera.irclog.whitequark.org/crux-devel/
<farkuhar> beerman: pulseaudio's pre-install script might be better as a README, as you did for openblas's post-install script. The pulse user and the pulse-access group are really only needed for a system-wide daemon (a configuration that is discouraged by upstream: https://www.freedesktop.org/wiki/Software/PulseAudio/Documentation/User/WhatIsWrongWithSystemWide/ )
<beerman> farkuhar: maybe, yeah
SiFuh has quit [Ping timeout: 252 seconds]
SiFuh has joined #crux-devel
<farkuhar> the pulseaudio port doesn't currently deploy a rc.d startup script, so any user who wants a system-wide daemon will have to write one by hand anyway. Everyone else can run a per-user process, without needing any new entries in /etc/group or /etc/passwd.
<farkuhar> I like how it's stated in the openblas README, "the decision to swap them out [openblas for lapack] is only yours to make! Please review the consequences of this script!" Similar language could be used in the pulseaudio port to introduce the changes needed for a system-wide daemon.
* SiFuh cuts the line
<beerman> i took a more simplistic approach, its very much discouraged, its only in there because i adopted it like that, i never cared much about pulseaudio and people should infact use pipewire by now, which does it the same way, so I don't think there is much to say on the topic
ppetrov^ has joined #crux-devel
<ppetrov^> hey guys, as I posted in #crux: building openblas on multiple cores always failed here, but adding -j1 after `make` seems to fix that, e.g. line 13 of the Pkgfile:
<ppetrov^> make -j1 NO_LAPACK=1 NO_LAPACKE=1 NO_STATIC=1 NO_AFFINITY=1 \
<beerman> doesn't fail for me. can you post your pkgmk.conf? Are there significant changes to the build process (e.g. further modifications to prtutils or other core ports..)? Do you use ccache/distcc/sccache? more infos please
<ppetrov^> my cpu is i7-9700kf
<ppetrov^> i do not use ccache/distcc/sccache, however I compile in RAM
<beerman> compiling in ram i wouldn't view as critical. no idea why it fails for you, I am running the build with -j32 over here and have no problems. Can you post your complete build log including the failure?
<ppetrov^> just a moment
<beerman> nevermind, I think I see what you mean, I'll try something out over here
<ppetrov^> there, check the very beginning
<jaeger> I see openblas running make -j32 in my test here on a 16-thread system... is it just doing cpus*2?
<jaeger> It did fail here as well on the first build, for reference
<beerman> setting openmp=1 might help, i am testing that right now
<beerman> the issue is old, i have no idea why it just now pops up over here
<beerman> i have no idea about the logic, I explicitly set 32 for the cores i have flying around in my distcc cluster
<beerman> seems like this is it though, builds for me again
<jaeger> USE_OPENMP=1 is already in the Pkgfile, guess that wasn't enough?
<beerman> actually no, its not openmp, I think its the shared flag parsed to make
<jaeger> ah
<beerman> USE_OPENMP=0 was what I was thinking
<beerman> today the server is still booted (rip my electrical bill) so give me a few more minutes
<jaeger> ah
<beerman> i always hated the openblas build system.. I see they also ship cmake, but I will try that out some other time
<beerman> i'll double check it here on my desktop and a recent R project
<beerman> jaeger: ppetrov^ you can try again please
<ppetrov^> sure
<ppetrov^> builds fine using all cores
<beerman> cheers
<ppetrov^> thanks for fixing this!
ppetrov^ has left #crux-devel [Leaving]
<jaeger> was away but I'll check
<jaeger> Seems to build fine now
<beerman> thanks
<jaeger> np
crash_ has joined #crux-devel