mwhudson | RAOF: i seem to recall that i was talking to you when i added what's now part 4 of this bugs test case: https://bugs.launchpad.net/ubuntu/+source/glibc/+bug/1951033 | 02:45 |
---|---|---|
ubottu | Launchpad bug 1951033 in glibc (Ubuntu Focal) "20.04 SRU" [Undecided, Fix Committed] | 02:45 |
mwhudson | RAOF: do you remember what this was about? | 02:45 |
RAOF | <mwhudson> "RAOF (he/his): do you remember..." <- I believe that was about graphics drivers in snaps linking to the core glibc and sometimes getting broken by it? | 02:52 |
mwhudson | RAOF: hm ok, do you know a snap like that off the top of your head? | 02:53 |
RAOF | The test case would be something like “run some GL-using snaps, on each of the supported graphics drivers” | 02:53 |
RAOF | Any flutter snap should suffice, I think? | 02:54 |
RAOF | That's probably the simplest option. | 02:54 |
mwhudson | ok | 02:54 |
mwhudson | so we still need a core20 snap with the new glibc in first? | 02:54 |
RAOF | Maybe? | 02:55 |
RAOF | Yes. | 02:55 |
RAOF | The host graphics drivers get copied in, but not the glibc, right? | 02:56 |
mwhudson | i don't know! | 02:56 |
mwhudson | i'm still an ubuntu server developer at heart :) | 02:57 |
RAOF | I think that's how it works 😬 | 03:00 |
Unit193 | I still think of you as the golang guy. | 03:00 |
RAOF | mwhudson: are you familiar with using the testflinger infrastructure? That might be the easiest way to smoke-test the {all supported releases} × {all supported drivers} matrix. | 03:06 |
mwhudson | Unit193: that's mostly William (Wilson) now! | 03:06 |
mwhudson | RAOF: no, mabye i should become familiar with that | 03:07 |
mwhudson | well more familiar, i have heard of it | 03:07 |
mwhudson | but first it's time to go into the big blue room for a bit | 03:07 |
serkan_ | hi. any ideas how to preseed Ubuntu 21.04 and 22.0 desktop! ISO for autoinstall. with older ISOs my solution works since those ISOs still have isolinux folder. I pass preseed file to kernel command line in isolinux config. it works. doing the same in grub config file in 22.x and 21.x does not seem to work. | 08:40 |
=== sem2peie- is now known as sem2peie | ||
RikMills | \o/ this now exists https://lists.ubuntu.com/mailman/listinfo/kinetic-changes | 09:44 |
Unit193 | Last I looked it didn't exist on LP/the mirrors. :3 | 09:45 |
RikMills | one step at a time | 09:45 |
Unit193 | (I'm waiting for the repo so I can add it to ubottu.) | 09:46 |
mwhudson | RikMills: not sure my subscription attempt went anywhere when i tried to subscribe earlier though | 09:56 |
cjwatson | Unit193: initializing it at the moment | 10:14 |
Unit193 | Nice, will add it my tomorrow then. Thanks. | 10:14 |
RikMills | mwhudson: it subscribe me ok just a little while ago. or at least appeared to. I got the right responses back | 10:14 |
mwhudson | RikMills: hmm tried again and it worked. maybe i typoed the email address earlier :) | 10:46 |
RikMills | and kinetic is here http://archive.ubuntu.com/ubuntu/dists/ | 11:31 |
ricotz | you stalking a kudu ;P | 11:33 |
RikMills | ppa builds also seem to work.... | 11:34 |
RikMills | ricotz: it seems so | 11:34 |
ricotz | so switching the sources.list already works :) | 11:37 |
laney | but who will be the first to dist-upgrade..? | 11:53 |
ricotz | just keep it rolling ... | 11:54 |
ahasenack | hi, the mariadb build in launchpad is failing, and this looks like is the problem: | 12:24 |
ahasenack | 2022-04-14 8:11:49 0 [Warning] mariadbd: io_uring_queue_init() failed with ENOMEM: try larger memory locked limit, ulimit -l, or https://mariadb.com/kb/en/systemd/#configuring-limitmemlock under systemd (262144 bytes required) | 12:24 |
ahasenack | it builds fine in a lxd container in a bare metal machine I have access to | 12:24 |
ahasenack | so the question is, does it need a "bigger" builder vm? How can that be done? | 12:24 |
ahasenack | https://launchpad.net/ubuntu/+source/mariadb-10.6/1:10.6.7-3/+build/23488405 | 12:24 |
ahasenack | looks like it needs a slightly higher limit for memlock | 12:32 |
ddstreet_away | ahasenack or maybe it should be fixed to handle failing to memlock, that's a problem with a few 'special' applications that think they should be exempt from ever swapping any of their memory, e.g. bug: #1911904 | 13:34 |
ubottu | Bug 1911904 in corosync (Ubuntu Hirsute) "corosync locks all its current and future memory" [Medium, Fix Released] https://launchpad.net/bugs/1911904 | 13:34 |
ahasenack | I'm trying a build with some measurements, to see if it's even trying to raise the limit | 13:34 |
cjwatson | ahasenack: there are no bigger builder vms, that's not a thing | 13:34 |
cjwatson | memlock limit may be tweakable with some effort, but also what ddstreet_away said | 13:35 |
=== ddstreet_away is now known as ddstreet | ||
ddstreet | ahasenack raising the limit will likely only delay the crash, and probably is the wrong approach | 13:36 |
ahasenack | this is for a test it runs during build | 13:37 |
ddstreet | if you read that bug, i laid out the issue in some detail | 13:37 |
ahasenack | I will, thanks | 13:37 |
ddstreet | but i am pretty sure simply raising its limit is the wrong thing to do | 13:37 |
ahasenack | hilariously, it builds on the problematic architectures, like armhf and rsync :) | 13:38 |
ahasenack | er, s/rsync/riscv64/ | 13:38 |
ahasenack | armhf I know is in lxd, risc I don't know | 13:38 |
ddstreet | ahasenack it looks like there is a --memlock param so i assume that's what is failing the test case, and from a quick 60 second scan of the code it seems like it only locks current memory (not future) in which case simply raising the RLIMIT_MEMLOCK would work, however that's not always possible inside (unprivileged) containers | 13:46 |
ddstreet | in any case, good luck with it :) | 13:46 |
ahasenack | I think it's a change that came with uring support | 13:46 |
ahasenack | in any case, `262144 bytes required` doesn't seem like a lot | 13:47 |
ddstreet | mariadb seems to try to provide a systemd service override too, to increase the LimitMEMLOCK, maybe that just is defaulting too low | 13:52 |
ddstreet | so maybe bumping that is all that's needed, at least until it hits the new limit later ;-) | 13:53 |
ahasenack | during the test I doubt it uses systemd | 13:56 |
ahasenack | it's launched directly I bet | 13:56 |
ahasenack | huh, ulimit -l shows 64 just before the test starts | 13:58 |
ahasenack | that's lower than even the ubuntu default, on a fresh jammy vm | 13:58 |
seb128 | bdmurray, hey, the http://reqorts.qa.ubuntu.com/reports/rls-mgr/rls-<nn>-incoming-bug-tasks.html for bb, ff, ii have no desktop section which seems weird, is there any known issue with those? | 14:12 |
schopin | Dumb question: is the kk archive opened for uploads? | 14:19 |
ahasenack | check the #topic in #ubuntu-release, but no, not yet open | 14:22 |
cjwatson | ahasenack: "armhf I know is in lxd"> that's for autopkgtest, not for Launchpad builds | 15:35 |
cjwatson | ahasenack: all .deb package builds in Launchpad run in chroots of the appropriate Ubuntu series inside focal VMs | 15:35 |
cjwatson | (at present) | 15:36 |
ahasenack | ah, right | 15:38 |
ahasenack | (armhf lxd) | 15:38 |
bdmurray | enr0n: there is an #ubuntu-mirrors channel | 16:42 |
enr0n | bdmurray: cool, thanks | 17:46 |
bdmurray | seb128: I don't see desktop-packages in rls-ii but then I don't see foundations-bugs there either because they are no bugs which are tagged rls-ii-incoming for those teams | 18:13 |
bdmurray | I do see desktop-packages in rls-ff-incoming | 18:13 |
ahasenack | schopin: hi, during the openssl3 transition, did you encounter sendmail patches by any chance? | 18:46 |
ahasenack | or a git repo from upstream? | 18:47 |
schopin | That doesn't sound familiar, no. | 18:47 |
ahasenack | schopin: did you have a tag for openssl3 related build failures? | 19:35 |
vorlon | hallyn: heya - so this was not on my radar, but I can take care of merging the new upstream version early in kinetic cycle. I'll let the current version hit the release pocket first though, to get rid of krb5 dependencies | 19:57 |
hallyn | vorlon: great, thanks | 19:58 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!