[08:14] could someone retry this https://launchpad.net/ubuntu/+source/golang-1.22/1.22.10-1/+build/29365721 it's flaky test, build succesfully in debian [08:43] if a package is deleted on an arch and its tests expectedly starts failing, do I need to trigger a migration-reference/0 test or is the removal enough? [08:43] hmm, no, forget that question, I misread something [09:14] zhsj: retried [09:14] sudip: thanks === voonij is now known as cokpolbb [14:07] vtk9 build took 1 day, 3 hours, 58 minutes, 32.5 seconds and then failed without a log :( [18:32] looking at the gpaw autopkgtest failures, it needs a machine with 4 core. how do I trigger a test to use the large machine? [20:18] sudip: I'm preparing a patch instead [20:18] 4 cores could be obtained through big_packages [20:19] see https://wiki.ubuntu.com/ProposedMigration#Openstack_Flavor_Details_for_Architectures [20:19] adrien: yes, seen that, but then how did the migration-reference job triggered by tsimonq2_ landed in that type of machine? [20:20] so I assumed there will be some undocumented option :) [20:21] big_packages list does not have gpaw [20:22] that's an interesting question, and I don't have the answer; one possibility is that the behavior is different with the new openmpi and I think I may have seen something like that in some log [20:23] basically, does one slot require one CPU, or can you put several slots on the same CPU? (maybe I just wrote something heretical to anyone versed in MPI :) ) [20:23] I have an MR almost ready [20:28] sudip: https://code.launchpad.net/~adrien/ubuntu/+source/gpaw/+git/gpaw/+merge/477858 === tsimonq2_ is now known as tsimonq2 [20:30] sudip: I think there's a hardcoded list in addition to big_packages. I *think*. That's probably a vorlon (or similar) question. [20:30] tsimonq2: I think https://git.launchpad.net/~ubuntu-release/autopkgtest-cloud/+git/autopkgtest-package-configs/tree/big_packages?h=main [20:31] another list ? [20:31] sudip: My guess is that we're talking hardcoding. [20:31] Oh, and if it's autopkgtests and not Launchpad, paride might know [20:33] imho it's debatable that it's a big package; there are quotas for big_packages sadly, that's why I decided to lower the core count rather than try adding it to big_packages [20:33] I triggered another migration-reference, if it again goes to large then there is something [21:07] ok, so the previous migration_reference did not go to a large machine. if you see the testinfo.json in artifacts, it says "nproc": "2", so something else has changed in openmpi [21:34] adrien: parmetis autopkgtest also has the same failure [21:44] sudip: good catch, and there's indeed a very suspicious 4 in the tests; I've started testing a patch but I'll go to bed before the results are in [21:44] and this time I used nproc actually; I thought it was in a package that may not be installed but it's actually in coreutils (nowadays?) [22:28] adrien, sudip, tsimonq2: that is strange. If I look at https://autopkgtest.ubuntu.com/results/autopkgtest-plucky/plucky/amd64/g/gpaw/20241123_192254_230c6@/log.gz I see it ran using the 'autopkgtest' flavor on the lcy02 cloud (you can see this info in the runner/autopkgtest command line, at the very beginning of the log) [22:28] I tried `autopkgtest flavor show autopkgtest` on that cloud, and: [22:28] | vcpus | 2 [22:28] I also tried `nproc` in a VM created using that flavor, and it's 2 again [22:30] in this PASS run triggered by sudip: https://autopkgtest.ubuntu.com/results/autopkgtest-plucky/plucky/amd64/g/gpaw/20241205_205031_28042@/log.gz [22:31] openstack failed and made some extra logs visible, including the fact that the flavor has 2 vcpus [22:32] (the test run was auto-retried after the failure, which is why it eventually passed) [22:32] so: I don't know, but looks like "gpaw now requires >= 4 cores" is not the whole story === JanC is now known as Guest3359