Bug#1001661: notcurses: flaky autopkgtests on armhf
Hi bunk, On 20-08-2022 21:38, Adrian Bunk wrote: A package with a flaky autopkgtest is worse than a package without an autopkgtest. Not having working and used software in the next stable is also worse than disabling the autopkgtest. One possible mitigation would be to disable the autopkgtest until the issue has been fixed. For avoidance of doubt, I totally agree with you (I have already NMUed packages to only turn of tests). Maybe I should suggest that more often explicitly? Paul OpenPGP_signature Description: OpenPGP digital signature
Bug#1001661: notcurses: flaky autopkgtests on armhf
On Wed, Mar 16, 2022 at 10:48:23PM +0100, Paul Gevers wrote: >... > All timeout after 2:47 hours. There are no reports of runtime issues. There do not seem to be any FTBFS issues due to it: https://tests.reproducible-builds.org/debian/history/notcurses.html A package with a flaky autopkgtest is worse than a package without an autopkgtest. Not having working and used software in the next stable is also worse than disabling the autopkgtest. One possible mitigation would be to disable the autopkgtest until the issue has been fixed. > Paul cu Adrian
Bug#1001661: notcurses: flaky autopkgtests on armhf
Version 3.0.7 seems to be be passing most of the time on armhf, though there were a few failures duing attempts to migrate gcc-12. The failures appear to be timeouts. Unfortunately 3.0.7 seems to be pretty consistently failing on s390x :(
Bug#1001661: notcurses: flaky autopkgtests on armhf
On Mon, 4 Apr 2022 02:05:15 -0400 nick black wrote: > i'm pretty sure raspberry pi is armhf, so i ought be able to dig > that one RPi4 i've got up and explore this. if we can reproduce > the problem interactively, it ought fall pretty quickly. Note that on Debian, the RPi 0/0w/1 use armel, the RPi 2 uses armhf and RPi 3 and 4 use arm64. I know RPi 0/0w/1/2/3 can (also) run in 32-bit 'mode' and I think the RPi 4 does too (but I don't have that version). On Mon, 4 Apr 2022 02:20:40 -0400 nick black wrote: > wait...even the most recent of these logs shows that it is > testing notcurses 3.0.4, which indeed had this timing problem on > input, which was fixed in notcurses 3.0.5: I guess you checked the wrong log? The highest version is: https://ci.debian.net/data/autopkgtest/testing/armhf/n/notcurses/20022214/log.gz starting date: 2022-03-15 ... autopkgtest [22:12:59]: testbed dpkg architecture: armhf autopkgtest [22:13:00]: testbed running kernel: Linux 5.10.0-12-arm64 #1 SMP Debian 5.10.103-1 (2022-03-07) autopkgtest [22:13:00]: apt-source notcurses Get:1 http://deb.debian.org/debian testing/main notcurses 3.0.7+dfsg.1-1 (dsc) [3,148 B] Get:2 http://deb.debian.org/debian testing/main notcurses 3.0.7+dfsg.1-1 (tar) [7,399 kB] Get:3 http://deb.debian.org/debian testing/main notcurses 3.0.7+dfsg.1-1 (asc) [833 B] Get:4 http://deb.debian.org/debian testing/main notcurses 3.0.7+dfsg.1-1 (diff) [17.3 kB] > oh, i guess "No testing version" indicates it's gone from > testing. so why isn't 3.0.7 in unstable the candidate for > testing? 3.0.4 is definitely known to be bad. https://tracker.debian.org/pkg/notcurses shows there is a version in testing. root@bagend:~# aptitude versions notcurses-bin Package notcurses-bin: p 2.2.3+dfsg.1-1stable 500 i 3.0.7+dfsg.1-1 testing,unstable500 root@bagend:~# apt-show-versions notcurses-bin notcurses-bin:amd64/testing 3.0.7+dfsg.1-1 uptodate notcurses-bin:arm64 not installed And both aptitude and apt-show-versions confirm that on my machine. HTH, Diederik signature.asc Description: This is a digitally signed message part.
Bug#1001661: notcurses: flaky autopkgtests on armhf
i'm pretty sure raspberry pi is armhf, so i ought be able to dig that one RPi4 i've got up and explore this. if we can reproduce the problem interactively, it ought fall pretty quickly.
Bug#1001661: notcurses: flaky autopkgtests on armhf
Control: severity -1 serious Control: found -1 3.0.7+dfsg.1-1 On Mon, 13 Dec 2021 22:26:10 +0100 Sebastian Ramacher wrote: For a recent failure see https://ci.debian.net/data/autopkgtest/testing/armhf/n/notcurses/17505950/log.gz https://ci.debian.net/data/autopkgtest/testing/armhf/n/notcurses/20022214/log.gz https://ci.debian.net/data/autopkgtest/testing/armhf/n/notcurses/19977381/log.gz https://ci.debian.net/data/autopkgtest/testing/armhf/n/notcurses/19573842/log.gz https://ci.debian.net/data/autopkgtest/testing/armhf/n/notcurses/18318908/log.gz https://ci.debian.net/data/autopkgtest/testing/armhf/n/notcurses/18235559/log.gz https://ci.debian.net/data/autopkgtest/testing/armhf/n/notcurses/18231285/log.gz All timeout after 2:47 hours. Paul OpenPGP_signature Description: OpenPGP digital signature
Bug#1001661: Info received (Bug#1001661: notcurses: flaky autopkgtests on armhf)
well, we don't yet know whether this "took", since 3.0.2 shows regressions across the board in autopkgtests. i've set up the upstream bug to track this, and am on it: https://github.com/dankamongmen/notcurses/issues/2505 thus far i've been able to run down that it's the "box" demo that's failing in the "demo" autopkgtest. i expect to have it resolved tonight. i'll either cut a new debian package with the necessary patch, or cut a 3.0.3, we'll see.
Bug#1001661: notcurses: flaky autopkgtests on armhf
Thanks for the heads-up. I believe/hope that this is the same issue that affected our ARM MacOS builds and Alpine i686/ARMHF builds through the 3.0.0 release, and which has been fixed in the 3.0.1 release: https://github.com/dankamongmen/notcurses/issues/2420 3.0.0 is currently in the NEW queue, uploaded to experimental, pending approval from the FTPteam (new SONAME -> libnotcurses3). Once it is approved, I will upload 3.0.1 (or a newer release, if one exists) to unstable, and I expect that this will address this problem. In the meantime, since the fix is small and precise, I could easily prepare a patch for 2.4.9. I think I will go ahead and do so, both to support/disabuse the notion that this is the same issue, and since FTPteam + TransitionTeam could take significant time. signature.asc Description: PGP signature
Bug#1001661: notcurses: flaky autopkgtests on armhf
Source: notcurses Version: 2.4.9+dfsg.1-1 Severity: important X-Debbugs-Cc: sramac...@debian.org notcurses' autopkgtest fail from time to time on armhf: | bash: line 1: 5188 Segmentation fault bash -ec 'TERM=xterm-256color notcurses-demo -d0 -c' 2> >(tee -a /tmp/autopkgtest-lxc.w6ikw2j4/downtmp/demo-stderr >&2) > >(tee -a /tmp/autopkgtest-lxc.w6ikw2j4/downtmp/demo-stdout) | autopkgtest [23:46:21]: test demo: ---] | autopkgtest [23:46:21]: test demo: - - - - - - - - - - results - - - - - - - - - - | demo FAIL non-zero exit status 139 For a recent failure see https://ci.debian.net/data/autopkgtest/testing/armhf/n/notcurses/17505950/log.gz Cheers -- Sebastian Ramacher signature.asc Description: PGP signature
