12:00:47 <jki> #startmeeting CIP IRC weekly meeting 12:00:47 <collab-meetbot> Meeting started Thu Apr 23 12:00:47 2026 UTC and is due to finish in 60 minutes. The chair is jki. Information about MeetBot at http://wiki.debian.org/MeetBot. 12:00:47 <collab-meetbot> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 12:00:47 <collab-meetbot> The meeting name has been set to 'cip_irc_weekly_meeting' 12:00:58 <jki> #topic AI review 12:01:15 <jki> no Action Items (we will come to other AI later ;)) 12:01:27 <jki> 5 12:01:29 <jki> 4 12:01:31 <jki> 3 12:01:33 <jki> 2 12:01:35 <jki> 1 12:01:39 <jki> #topic Kernel maintenance updates 12:01:52 <uli_> i've been working on 4.4 12:01:54 <masami> This week reported 102 new CVEs and 132 updated CVEs. 12:02:06 <iwamatsu> I reviewed 6.12.80. 12:02:28 <pave1> I am reviewing 6.12.82 and 83 12:03:35 <jki> anything to add? 12:03:49 <jki> 5 12:03:52 <jki> 4 12:03:54 <jki> 3 12:03:56 <jki> 2 12:03:58 <jki> 1 12:04:01 <jki> #topic Kernel release status 12:04:20 <jki> 4.4 is now quite late but almost done, right? 12:04:31 <uli_> reviews are in, so i expect to be able to push it tomorrow 12:04:57 <jki> was there a particular reason for the delay? 12:05:19 <uli_> more backports than expected. many things went into 4.19 cleanly but had to be backported for 4.4 12:05:31 <uli_> usually that is not the case, or at least to a much lower extent 12:05:54 <jki> ok, please speak up early if it repeats 12:06:11 <uli_> ok 12:06:23 <jki> if workload grows, we need to flag that 12:06:37 <jki> before being boiled alive ;) 12:06:57 <uli_> it varies between releases, sometimes dramatically 12:07:00 <uli_> can't really predict that 12:07:51 <jki> sure, but in that case one question would be if there is chance to scale out the work 12:07:56 <jki> have more hands when needed 12:08:28 <uli_> not sure. for the last release, 4.4 was very little effort, iirc i released it on the same day as 4.19 or something like that 12:09:36 <jki> anyway, let's just monitor carefully 12:09:43 <jki> anything else? 12:10:26 <jki> 5 12:10:27 <jki> 4 12:10:29 <jki> 3 12:10:30 <jki> 2 12:10:32 <jki> 1 12:10:34 <jki> #topic Kernel testing 12:11:02 <patersonc> Not much from me. 12:11:14 <arisut> not much from me, applied kernel testing and dependability MC for LPC 2026 12:11:20 <patersonc> iwamatsu: your lab is currently offline, could you take a look when you get a chance? 12:11:47 <iwamatsu> patersonc: I will check after this meeting. 12:11:58 <patersonc> Thanks 12:12:50 <jki> just reminded Pasquale of the prague lab state... 12:13:17 <patersonc> Thanks 12:14:22 <jki> more testing topics? 12:14:58 <jki> 5 12:15:00 <jki> 4 12:15:02 <jki> 3 12:15:04 <jki> 2 12:15:06 <jki> 1 12:15:09 <jki> #topic AOB 12:15:43 <jki> Glasswing topics from kernel and testing perspective - just a reminder for my email 12:16:05 <jki> if you have anything to discuss here already, go ahead 12:16:26 <jki> timeline is still unclear, though 12:16:55 <pave1> Just be sure thete is human sending issues to us 12:17:09 <pave1> so we can tell him to stop. 12:17:40 <pave1> We already have source of low quality issues 12:17:43 <jki> might be me in the end - you may even shout then ;) 12:17:50 <uli_> +1. i use llms extensively, but that's me controlling the ai. if somebody else sent me false positives, i'd become very annoyed very quickly. 12:18:02 <pave1> - CVE. 12:18:19 <pave1> Yeah, I seen you volunteering. 12:18:41 <pave1> Just make sure there's human on 12:18:43 <jki> I would expect to 2-phase process: learn based on few examples what it can do, then scale up, or down 12:18:52 <pave1> the other end, too. 12:19:44 <jki> Glasswing access is apparently always operated by one individual, human, per participating LF project 12:20:09 <jki> they told us this is by design, to have many projects benefiting, rather than few with more bandwidth 12:20:39 <pave1> So no human oversight at the other end? 12:21:26 <jki> the humans in there are LF project representatives 12:22:01 <jki> the give prompts and (pre-)process the results, then decide what to do with them (simplified) 12:22:38 <jki> I haven't seen any written conditions yet, remains a bit speculation 12:23:25 <pave1> I'd ask someone on AI side to do at least pre-processing. 12:24:11 <uli_> that would be a massive manual effort; i'm sure they will leave that to us... 12:24:11 <jki> they (Anthropic) just to the operation and safety supervision here - again only my current understanding 12:24:21 <jki> yep 12:25:14 <jki> but it will remain fully up to us as project to which degree we make use of it, or can do that if we are limited in time/hands 12:26:02 <jki> so, representative depth first, then discuss if/how to go wide 12:26:46 <jki> if you have done past experiments with moderate success, those could be retried e.g. 12:27:16 <jki> if you have harder questions to ask that were not possible so far (or too expensive), let's try one of them 12:27:19 <jki> etc... 12:29:17 <jki> other topics for today? 12:30:06 <jki> 5 12:30:08 <jki> 4 12:30:10 <jki> 3 12:30:12 <jki> 2 12:30:14 <jki> 1 12:30:16 <jki> #endmeeting