daniels changed the topic of #freedesktop to: https://www.freedesktop.org infrastructure and online services || for questions about freedesktop.org projects, please see each project's contact || for discussions about specifications, please use https://gitlab.freedesktop.org/xdg or xdg@lists.freedesktop.org
pixelcluster_ has quit [Ping timeout: 480 seconds]
zxrom_ has quit []
privacy has joined #freedesktop
vx has quit [Quit: G-Line: User has been permanently banned from this network.]
PuercoPop has joined #freedesktop
vx has joined #freedesktop
PuercoPop has quit [Ping timeout: 480 seconds]
ghishadow has quit [Ping timeout: 480 seconds]
bmodem has joined #freedesktop
ghishadow has joined #freedesktop
AbleBacon has quit [Read error: Connection reset by peer]
ximion has quit [Quit: Detached from the Matrix]
sima has joined #freedesktop
privacy has quit [Remote host closed the connection]
Haaninjo has joined #freedesktop
tzimmermann has joined #freedesktop
<pjakobsson>
daniels, trying to get dim working with committer access again after pointing drm-misc to gitlab
<pjakobsson>
daniels, I've been using the username patrik but my gitlab account is pjakobsson. Should these users be merged somehow?
<mupuf>
robclark: https urls shouldn't require authentication, but the ssh obviously will
<mupuf>
what url are you using?
lynxeye has joined #freedesktop
zxrom has joined #freedesktop
sahilister has joined #freedesktop
f_ has joined #freedesktop
bmodem has quit [Ping timeout: 480 seconds]
f_ has quit [Remote host closed the connection]
f_ has joined #freedesktop
Juest has quit [Read error: No route to host]
f_ has quit [Ping timeout: 480 seconds]
Juest has joined #freedesktop
mripard has joined #freedesktop
vkareh has joined #freedesktop
r00tobo has joined #freedesktop
guludo has joined #freedesktop
scrumplex has joined #freedesktop
<zmike>
what happened to the automatic piglit uprev jobs?
<zmike>
am I just not seeing them?
scrumplex_ has quit [Ping timeout: 480 seconds]
<mupuf>
zmike: sergi was on holiday :)
<zmike>
how does someone being on holiday affect automated jobs? did the jobs die?
<mupuf>
and AFAIK he is now focusing on automatic update of expectations based of nightly results... which should help with the piglit auto uprev process
<mupuf>
I doubt it was 100% automated
<zmike>
ah
<daniels>
the job itself is completely automated, as you can see
<daniels>
however depending on the results it may not be automerged
<daniels>
like if it regresses 40000 tests, you probably don't want to slam that straight in
<zmike>
I asked because I didn't see a new MR for a while
<zmike>
but maybe I didn't look hard enough
privacy has joined #freedesktop
bmodem has joined #freedesktop
ghishadow has quit []
ghishadow has joined #freedesktop
<robclark>
mupuf: it was the https one.. https for mesa/mesa works. I couldn't find any setting in robclark/mesa that would make it private, so idk what is going on
<robclark>
hmm.. it does for me in the host.. but didn't yesterday in a vm which didn't have any auth tokens or whatever else..
<mupuf>
I just tested on my server... whbich definitely has no gitlab token
<mupuf>
weird
jarthur has joined #freedesktop
ximion has joined #freedesktop
<robclark>
mupuf: hmm, it's possible I typo'd the URL .. and I vaguely remember that in that sort of case gitlab decides it's something that might exist if you were authenticated..
<daniels>
robclark: and yeah, GitLab (as with GitHub) will return 401 rather than 403/404, so you can't try to discover hidden repos
ximion has quit [Quit: Detached from the Matrix]
sergi has joined #freedesktop
<sergi>
zmike: mupuf: there is a scheduled pipeline with the uprevs. One of the jobs is to prepare a candidate to uprev Piglit in Mesa. So he can have a proposal to uprev to yesterday's latest piglit every day. But it is only a proposal, not defined yet when assign it to marge. This usually happens when someone asks, but I'm thinking in actively request to do so at least once a week.
<zmike>
it would be nice to do this weekly for sure
<mupuf>
Weekly?
<mupuf>
That's a lot of artifacts :o
<zmike>
I think an uprev now would eliminate a couple hundred piglit fails across the tree
<sergi>
the merge request started a month ago, it got merged a week ago
<mupuf>
Well, good that I have a proposal to reduce their cost!
<zmike>
sergi: what caused it to take so long?
lsd|2 has joined #freedesktop
<sergi>
what is "it"? Why not get it merged earlier that this month? If so, just because noone assigned it to marge
<daniels>
zmike: someone needed to review all the expectation changes, no-one did
<zmike>
ah, is that all?
<zmike>
okay, so we need to train people to start expecting/reviewing these
<sergi>
I'm thinking to ping people when I see expectation changes on certain families. And then, try to see if there is consensus to assign to marge
<zmike>
I think the better workflow would be to ping when the MR goes up, and if people don't nak within the week then it merges
<zmike>
it's just ci expectations
<zmike>
if there are new fails added then they can be examined whenever
<sergi>
and we have a time window of a day to study. So, if these expectations changed, they are because these changes in piglit and in mesa.
f_ has joined #freedesktop
<daniels>
yeah I think that's pretty reasonable, we just didn't automate the 'land if it's not NAKed' bit (probably because trying to parse what is and isn't a NAK from a comment is rough), and no-one manually swept it up because we've been in Australia / having children / etc
<daniels>
mupuf: you wanted to do ML for automated updates right? maybe you could start with comment sentiment analysis to try to figure out wtf constitutes a nak :P
lsd|2 has joined #freedesktop
<zmike>
just require that someone post case-insensitive 'nak'
<mupuf>
Lol, a human assigning to Marge is fine by me
<mupuf>
Even a bot that ping is fine
<zmike>
realistically
<zmike>
like
<zmike>
I'm not sure why we should allow naks at all?
<zmike>
is there a real purpose to it?
<zmike>
even if an uprev adds a thousand new xfails and they're wrong, the tests can just be fixed and the next uprev will address it
<daniels>
if a piglit rev is completely broken, then realistically we're dropping a huge amount of coverage until it's fixed
<daniels>
and that's fine if someone does fix it, but I think we've already demonstrated a tragedy of the commons above
<sergi>
Nowadays the ci-uprev tool is working well, but this wasn't like that in earlier stages of this development.
<zmike>
yeah, but in theory SOMEONE is looking at these uprev MRs
<zmike>
so if this does happen even someone who doesn't work on drivers can notice that something is wrong
<zmike>
if we have some catastrophic failure case then we can use that as an example to improve automation
<zmike>
but it seems like that would be extremely rare and probably just hypothetical
<sergi>
I'll be checking the gfx-ci-bot daily proposals. Then, ping people based on expectation changes. If by the end of the week, no-one scream, I'll assign to marge.
<sergi>
At any time (if I'm not around) anyone with right to assign to marge, can do this action
<sergi>
If any developer needs a piglit uprev, can ping me or assign the MR to marge. (besides the fact that it is currently having trouble with a build job)
<sergi>
zmike: I didn't read all the messages before. The automated job work while I wasn't here. As you can see in the issue it creates, the automated process posted every day a note in an issue saying that it couldn't prepare a candidate. No-one react to it, neither me.
<zmike>
aha
<zmike>
that explains it
<zmike>
thanks
funderscore has joined #freedesktop
f_ has quit [Ping timeout: 480 seconds]
funderscore is now known as f_
<__tim>
daniels, mupuf, fwiw, in GStreamer have labels like 'Merge in 5 days', and a script that checks tagged MRs regularly to see if they're due, and assigns them to Marge when time has passed (unless there are open discussions or the MR can't be merged for other reasons, then the script will post a comment and remove the label). So someone opening a
<__tim>
discussion thread would block the automatic assignment to marge
<mupuf>
Smart
<mupuf>
Sounds ideal
<daniels>
__tim: nice, where's the script?
<__tim>
I'll put it somewhere in a bit
<__tim>
I should warn you that I write C code in any language though :P