Bug#1072533: procps: flaky autopkgtest:
Hi, Sorry for being tense yesterday. On 19-07-2024 8:12 p.m., Paul Gevers wrote: On 19-07-2024 10:04 a.m., Craig Small wrote: Ideally, I'd like to push a pre-release into the Debian CI and see if it works now. I'm not sure there is a way of doing that. You can upload to experimental. The upload will be tested on amd64 and arm64 automatically (in unstable, like normal uploads to unstable are tested in testing), and you can use the self-service [1] to schedule more. I can give you access to a testbed on our infrastructure where you can play in a production environment exactly like the tests run. We'd need to align the date/time, as in the current way that I can provide this service, I need to be on-line at the same time. Paul OpenPGP_signature.asc Description: OpenPGP digital signature
Bug#1072533: procps: flaky autopkgtest:
Hi On 19-07-2024 10:04 a.m., Craig Small wrote: Ideally, I'd like to push a pre-release into the Debian CI and see if it works now. I'm not sure there is a way of doing that. You can upload to experimental. The upload will be tested on amd64 and arm64 automatically (in unstable, like normal uploads to unstable are tested in testing), and you can use the self-service [1] to schedule more. Paul [1] https://ci.debian.net/user/ OpenPGP_signature.asc Description: OpenPGP digital signature
Bug#1072533: procps: flaky autopkgtest:
On Tue, 4 Jun 2024 at 05:57, Paul Gevers wrote: > I looked at the results of the autopkgtest of your package, because it > showed up in the glibc regressions. I noticed that it regularly fails on > amd64, ppc64el and s390x. For your info, as it seems to correlate, those > are the architectures where multiple tests run in parallel, each in > their own lxc container. > > Because the unstable-to-testing migration software now blocks on > regressions in testing, flaky tests, i.e. tests that flip between > passing and failing without changes to the list of installed packages, > are causing people unrelated to your package to spend time on these > tests. > > Don't hesitate to reach out if you need help and some more information > from our infrastructure. I'll need some help here, the lxc environment is doing something odd/wrong with the proc filesystem. We've been bit before from lxc, but I don't know enough about the CI system to know what's going on. This *might* be fixed with a commit checking for odd ways LXC presents files[1] discussed at [2]. It also appears to be LXC version dependent. 1: https://gitlab.com/procps-ng/procps/-/commit/42dce4d9f4132360647c4dcae1fbbfa1171528b3 2: https://gitlab.com/libvirt/libvirt/-/issues/492 Ideally, I'd like to push a pre-release into the Debian CI and see if it works now. I'm not sure there is a way of doing that. - Craig
Bug#1072533: procps: flaky autopkgtest:
On Tue, 4 Jun 2024 at 05:57, Paul Gevers wrote:
> I looked at the results of the autopkgtest of your package, because it
> showed up in the glibc regressions. I noticed that it regularly fails on
> amd64, ppc64el and s390x. For your info, as it seems to correlate, those
> are the architectures where multiple tests run in parallel, each in
> their own lxc container.
So the lxc container is probably presenting something odd to the
procfs because that type of virtualisation in particular naively
overwrites files in /proc.
What is strange is this works every time on my (amd64) system. It also
works on salsa.
> 53s [1;31mASSERT: [0mvalue line
I see the prepare_testbed just before this failed too, not sure what that is.
The value line is running 'vmstat' and then for the third line
grep -E '^([[:space:]]+[[:digit:]]+){18}$' "${stdoutF}"
i.e. is there 18 fields.
- Craig
Bug#1072533: procps: flaky autopkgtest:
Source: procps Version: 2:4.0.4-4 Severity: serious User: debian...@lists.debian.org Usertags: flaky Dear maintainer(s), I looked at the results of the autopkgtest of your package, because it showed up in the glibc regressions. I noticed that it regularly fails on amd64, ppc64el and s390x. For your info, as it seems to correlate, those are the architectures where multiple tests run in parallel, each in their own lxc container. Because the unstable-to-testing migration software now blocks on regressions in testing, flaky tests, i.e. tests that flip between passing and failing without changes to the list of installed packages, are causing people unrelated to your package to spend time on these tests. Don't hesitate to reach out if you need help and some more information from our infrastructure. Paul E.g. https://ci.debian.net/packages/p/procps/testing/amd64/47064938/ 52s autopkgtest [06:08:26]: test vmstat: [--- 53s test_no_arguments 53s [1;31mASSERT:[0mvalue line 53s [1;31mshunit2:ERROR[0m test_no_arguments() returned non-zero return code. 53s test_forks 53s test_disk 53s 53s Ran [1;36m3[0m tests. 53s 53s [1;31mFAILED[0m ([1;31mfailures=2[0m) 53s autopkgtest [06:08:27]: test vmstat: ---] OpenPGP_signature.asc Description: OpenPGP digital signature
