jaeger changed the topic of #crux to: CRUX 3.7 | Homepage: https://crux.nu/ | Ports: https://crux.nu/portdb/ https://crux.ninja/portdb/ | Logs: https://libera.irclog.whitequark.org/crux/
joe9 has joined #crux
<darfo> i've found qt6-webengine with 2G/core compiles but swaps a lot. By hacking ninja to run on half the cores it actually compiles faster because of no swapping.
<darfo> 1G/core does with OOM
<darfo> s/does/dies/
joe9 has quit [Quit: leaving]
<ukky> jaeger: qtwebengine builds fine on my system, no need to limit CPU cores. I just remove one core to watch youtube during the build. My RAM ratio is 5G per core.
<jaeger> hrmm, ok
<ukky> There were a few processes that allocated 3 GiB of memory. I think the build used approximately 40-45 GiB of RAM at the peak.
<jaeger> I'm currently testing 2G/core (including swap) on the laptop
<ukky> 2G/core plus swap might be a minimum for successful build.
<jaeger> Just failed with 8G RAM, 8G swap, 8 cores
<ukky> Could you try 1.5G/core ratio?
<ukky> I mean physical RAM ratio
serpente has quit [Remote host closed the connection]
<cruxbot> [contrib/3.7]: hwloc: 2.10.0 -> 2.11.0
<cruxbot> [contrib/3.7]: python3-dkim: 1.1.6 -> 1.1.7
<cruxbot> [contrib/3.7]: svt-av1: 2.1.0 -> 2.1.1
_moth_ has quit [Ping timeout: 268 seconds]
lavaball has joined #crux
dlcusa_ has joined #crux
dlcusa has quit [Ping timeout: 268 seconds]
ppetrov^ has joined #crux
braewoods has quit [Remote host closed the connection]
SiFuh has quit [Remote host closed the connection]
SiFuh has joined #crux
<cruxbot> [xorg/3.7]: xorg-xf86-video-nouveau: updated to version 1.0.17-7
<cruxbot> [xorg/3.7]: xorg-xf86-video-intel: updated to version 2.99.917-931
<cruxbot> [xorg/3.7]: libdrm: updated to version 2.4.122
<jaeger> 2G/core doesn't work so I'm sure 1.5G/core won't either
_moth_ has joined #crux
Guest9655 has joined #crux
Guest9655 has quit [Client Quit]
Guest9655 has joined #crux
Guest9655 has quit [Remote host closed the connection]
dlcusa_ is now known as dlcusa
_moth_ has quit [Quit: _moth_]
_moth_ has joined #crux
CrashTestDummy3 has quit [Quit: Leaving]
CrashTestDummy has joined #crux
CrashTestDummy has quit [Quit: Leaving]
CrashTestDummy has joined #crux
CrashTestDummy has quit [Client Quit]
CrashTestDummy has joined #crux
lavaball has quit [Quit: lavaball]
<jaeger> SiFuh: starting to test 3.8 builds, heads up for the contrib kernel config
<ppetrov^> awesome
<ppetrov^> jaeger, which kernel version are you planning to use?
<jaeger> currently 6.6.35
farkuhar has joined #crux
lavaball has joined #crux
<ukky> also this discussion might me intresting: https://forums.gentoo.org/viewtopic-p-8786167.html
<jaeger> interesting... looking at that ebuild I guess it somehow respects NINJAFLAGS rather than MAKEFLAGS or similar? Will have to try that. opt/qtwebengine sets that but not opt/qt6-webengine
<cruxbridge> <tim> do you all use ccache/distcc by any chance or am I the only one? Because this is the biggest difference I can see unmentioned here by all the reports
<ukky> tim: I do not use ccache/distcc
<jaeger> I don't use either of them anymore. I used to but have run into problems just enough to stop relying on them.
<cruxbridge> <tim> r0ni is currently investigating a bug that prevents me from compiling it on the rpi4 with 4gb of ram, but it used to work on there too, with 8 jobs between two pies
<cruxbridge> <tim> I am just curious why I don't seem to trigger the same problem you all seem to have. anyway, if it helps to set NINJAFLAGS as well I can add that
<ukky> jaeger: I believe gentoo does not limit number of compile jobs for qtwebengine. But it will not even try to start building and exit with a warning that system does not have enough RAM.
<ukky> all requirements checks are made here: https://mirror.umd.edu/gentoo-portage/eclass/check-reqs.eclass
<ukky> yes, line 214 is how gentoo limits number of compiler jobs for ninja based on MAKEOPTS jobs. I am not sure if ninja takes MAKEOPTS into consideration.
<jaeger> It does not
<jaeger> I've tried MAKEOPTS and MAKEFLAGS before, didn't know about NINJAFLAGS
ppetrov^ has quit [Quit: Leaving]
<darfo> only one I've found that ninja respects is JOBS
<darfo> oh, NM.
<darfo> It is because I shim in a ninja shell script that runs the real ninja with -j $JOBS
<jaeger> ninja doesn't respect JOBS either, yeah
<jaeger> I plan to test with NINJAFLAGS but if it also ignores that I'll shim as well and forget about it
<darfo> No respect at all :) Shim works though but I have to move it out of the way for full speed with other builds
<jaeger> Maybe make the shim do something silly like "if [ "$name" = "qtwebengine" ]" :D
<darfo> I've had trouble with some non-qt6* builds if the shim is in place. This is the one I use: https://dpaste.com/HMH34J5M4
<jaeger> pretty straightforward
<darfo> yes. Only thing I've found that works. I should have built my desktop at 4G/core instead of 2G/core.
<jaeger> As a side note, ninja's default seems to be to use $(nproc)+2, so even when I was disabling 4 cores it was still using 6 jobs, heh
<cruxbridge> <tim> :D
lavaball has quit [Remote host closed the connection]
<jaeger> Looks like it does not respect NINJAFLAGS or NINJA_FLAGS (found references to both), so wrapper it is
DesRoin has quit [Ping timeout: 264 seconds]
DesRoin has joined #crux
<cruxbot> [contrib/3.7]: firefox: 127.0.1 -> 127.0.2
tilman has quit [Ping timeout: 264 seconds]
tilman has joined #crux