06:32 <whot> well, we're now down to only hotmail and gmail spam users, at the rate of ... 2-3 per hour or so

06:32 <whot> that's managable and they should in theory get cleaned up at some point by the cron jobs

08:01 <karolherbst> I wished we could do something about gmail (like blocking them :P)

09:07 <DragoonAethis> whot: I had a slightly different idea, with automated external verification for users

09:07 <DragoonAethis> GitLab can make all new users require approval (manual or automated) and send you a webhook that a new user was registered

09:08 <DragoonAethis> So an external app could send an email to the new user that extra steps are required, something like "message this bot account on IRC" or "run a command and enter its output here"

09:08 <DragoonAethis> Nothing too hard or super annoying, just enough to make most bots bail out

09:45 <bentiss> DragoonAethis: (just passing by) last time we checked, webhooks on new accounts are not 100% reliable -> you get them from manual signup, but social accounts don't trigger a webhook :(

09:46 <alatiera[m]> jenatali: hi, was on vacation till no but we should be good to enable them again

09:47 <alatiera[m]> already did everything needed last week but didn't make an MR in mesa to turn them on

09:47 <alatiera[m]> will do that now

10:28 <whot> fgfg

10:29 <DragoonAethis> bentiss: then maybe a cronjob that collects new pending users every 10mins or so would work, if the app does not receive a webhook?

10:29 <DragoonAethis> OmniAuth can be configured to put new SSO users into pending as well

10:29 <DragoonAethis> GitLab has a ticket about that webhook not triggering for OmniAuth signups ._.

10:58 <whot> https://paste.centos.org/view/58ba4776 <- email domains of external users created since 1 April this year

10:58 <whot> I think I've now blocked all the ones in the top 50, together with *.click and *.top

10:59 <whot> realistically we'll need this in source control somewhere so we can just update and re-upload the file instead of keeping a long list

10:59 <whot> daniels, bentiss: ^^ (post will expire in 24h)

11:02 <whot> i'm also quite tempted to just wipe any username with 88 in the name and personally apologise to the handful of 36 year olds that want to legitimately sign up

11:05 <daniels> haha, fair

11:10 <mupuf> gmail.com and outlook.com is... problematic :s

11:13 <whot> the two together are only about 10 users per day and those should get wiped after 2 months of never logging in, so I don't think that's much of a problem

11:13 <whot> the actual spam domains creating 200k new users in 5 weeks is serious though

11:14 <whot> anyway, i'm off for today, happy weekend everyone

11:20 <mupuf> whot: enjoy your weekend, and thanks a lot for your work!

13:07 <slomo> wasn't there a way to directly look at html artifacts of CI jobs in the browser without downloading the whole thing first? now it just tells me "The source could not be displayed because it is stored as a job artifact. You can download it instead." when trying to look at a html file

13:36 <pq> slomo, needing a few clicks, yes. Seems to still work in Weston.

13:36 <pq> artifacts: expose_as:

13:41 <slomo> pq: thanks :)

14:30 <jenatali> alatiera: thank you!

14:30 <jenatali> Unfortunately there was already a break that snuck in. This happens every time

14:30 <jenatali> I'm so frustrated that our runner isn't online yet to prevent this kind of downtime

14:32 <__tim> this only happened because the second runner didn't have the mesa tag, but now it does have it

14:35 <jenatali> Fair. Having a third hosted elsewhere won't hurt though in case something happens to both of those

22:43 <zmike> https://gitlab.freedesktop.org/mesa/mesa/-/jobs/58594276

22:43 <zmike> 🤔