00:03 DavidHeidelberg[m]: jenatali: Hey! I guess you don't know about any possibility how to implement caching on Windows runners? I saw https://gitlab.com/gitlab-org/gitlab-runner/-/issues/4091 and it doesn't look well. I always hit this when I testing piglit and have to wait 10 minutes comparing to 1 minute for Debian to finish
00:04 jenatali: David Heidelberg: Nope, no idea. I almost never touch docker
00:16 HdkR: From what I've been told by others, Docker on Windows is a hellscape
00:19 jenatali: ¯⁠\⁠_⁠(⁠ツ⁠)⁠_⁠/⁠¯
00:21 HdkR: Maybe it was specific to something about mapped volumes being a nightmare
02:08 karolherbst: uhhh.. found a CTS bug :'(
02:10 karolherbst: so... on intel if you want to launch 1024 threads, you can only do so with 32 subgroups (because SIMD 32, 16 won't work as this would require more groups than available: 56). So the CTS queries for 1024 and gets 32 as the max. Then it tries with to check what local size works for 33 sub groups and assumes it fails. Of course I return 528 (== 16 * 33), because that works 🙃
02:10 karolherbst: this API is so silly sometimes
03:59 kode54: if someone posts a reviewed-by, do I need to edit my MR to include it?
04:04 jenatali: kode54: Generally, yes. There are some components that are not requiring reviewed-by tags on their commits anymore (e.g. Microsoft components and zink, and IIRC panfrost?) but common code still currently does
04:04 jenatali: Note that the tag goes in the commit message, editing the merge request description doesn't do anything, you need to git commit --amend / git rebase -i, and then git push -f
13:04 Hi-Angel: kode54: I crafted up some command that batch-adds R-bs to a bunch of commits long time ago, you can use it to make that faster https://stackoverflow.com/a/54903478/2388257
15:19 hch12907: <Hi-Angel> "kode54: I crafted up some..." <- ooh that's nice! I always used `git rebase -i` and reworded every commit with it
15:32 Hi-Angel: ^_^
15:37 Hi-Angel: hch12907: well, technically speaking, in absence of the command, another way, which lengthy but still easier than manually editing a dozen of commits, could be to do `git format-patch -20 --stdout > /tmp/1.patch` (where 20 it the number of commits), then `sed -i 's/---$/Reviewed-by: User<email>\n---/g' /tmp/1.patch`, and then `git reset --hard HEAD~20 && git am -3 /tmp/1.patch`
15:39 Hi-Angel: One could wrap it to a zsh/bash command as well
16:09 apteryx: hello! is it known that the TLS cert at https://archive.mesa3d.org has problems? e.g. wget https://archive.mesa3d.org/mesa-23.0.3.tar.xz fails with "The certificate of 'archive.mesa3d.org' doesn't have a known issuer."
16:12 Hi-Angel: apteryx: works for me, starts downloading
16:12 Hi-Angel: What IP does it show if you do ping archive.mesa3d.org?
16:13 apteryx: 131.252.210.176
16:13 Hi-Angel: Odd, same as me
16:13 apteryx: I use Mozilla's nss-certs
16:14 kisak: looks fine here in a browser and with wget. Issue seems local to you?
16:15 apteryx: it'd points to my certs then yes
16:15 apteryx: is this supposed to work: https://www.ssllabs.com/ssltest/analyze.html?d=archive.mesa3d.org&latest
16:15 apteryx: keeps reloading the page for me
18:10 karolherbst: yeah, it reloads until it's done.. I think it might also just refresh via some JS async stuff or whatever
22:02 kode54: appears to be Let's Encrypt