14:10fdobridge: <gfxstrand> I build with clang on Linux all the time (or at least I used to)
14:10fdobridge: <gfxstrand> I think we have a clang build in CI
15:34ice9: nouveau is causing soft and hard CPU lockups only after locking the screen, any idea?
15:46KungFuJesus: Hmm, I'll have to see what's going on. It seems like clang on powerpc64 + linux is confusing meson. Also looks like I may need to jump through some hoops to get the clang runtime going with msan
15:53RSpliet: ice9: I recently retired a GK107 because of frequent pseudo-random lock-ups. Never got nouveau quite as stable as it should be.
15:54RSpliet: You may wish to add additional details, and report your issue on... hmm... kernel bugzilla or gitlab
15:54RSpliet: karolherbst: preferences?
15:54KungFuJesus: I really wish gcc would implement msan, what they have in asan isn't quite up to what clang can do with msan
15:59ice9: RSpliet, I filed a bug on ubuntu launchpad, should I file another directly to nouveau too?
16:00RSpliet: Hmm... that depends on whether you're running a 6.1 kernel or not
16:00ice9: 5.19 RSpliet
16:02RSpliet: Bugs against upstream require an upstream software stack. I don't know too much about Ubuntu, but historically they've been behind a little bit on updates (per their policy).
16:02RSpliet: If you file against upstream, which is where most/all devs look, the first question you'll get is "are you running the upstream version of kernel/mesa/...
16:04ice9: RSpliet, so if bug is filed on the distro and the bug is confirmed, who should fix it? the distro developers or the upstream devs?
16:06RSpliet: ice9: the distro devs. Who may/should triage with the upstream devs.
16:06RSpliet: Upstream devs (I think all of the people here, defo most of them) generally aren't interested in chasing bugs in older versions of the software, as they may well have already fixed the issue and it just hadn't been backported. They'd end up barking up wrong trees endlessly.
18:29karolherbst: RSpliet: there is an annoying error, where the GPU crashes due to low memory bandwidth
18:30karolherbst: and icnreasing the pstate fixes it :)
18:30karolherbst: ice[m]: what error are you getting btw?
19:11HdkR: GPU crashes from low memory bandwidth? How does that work with the low-end DDR3/4 devices on the market?
19:13karolherbst: nvidia can upclock the GPU
19:13karolherbst: but it's more of a timeout thing and we don't handle it very well
19:13karolherbst: so if you are on low bandwidth GPU and the driver only runs your gpu at 10% you are kind of in a bad situation :)
19:17HdkR: Oh, so a timeout that results in a crash rather than logic exploding
19:26karolherbst: well.. firmware stuff (tm)
19:26karolherbst: I think
19:26karolherbst: or _something_
19:27karolherbst: but I think the context switching firmware breaks if it has to much stuff to do and stuff times out or something
19:27karolherbst: we had this werid ctx sw bug and I think it's exactly that
19:27karolherbst: which apparently only happens on lower end GPUs