autopkgtest fails in impish (0.22)

Bug #1936820 reported by Christian Ehrhardt 
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
xtensor (Ubuntu)
Fix Released
Undecided
Unassigned

Bug Description

There are (at least) two issues why the tests fail in Impish right now.

It blocks xtensor itself, but also src:xtensor-blas.

The test fails on all architectures, but for different reasons.
The last good test was more than two years ago.

s390x / arm64 / i386 fail on bad-package.
  - xtensor-dev became a transitional and is available on all arch.
  - But the actual dependency libxtensor-dev only exists on a few
    libxtensor-dev | 0.22.0-5 | impish-proposed | amd64, armhf, ppc64el, riscv64
  - That explains the badpkg issue, and I guess we can just reset it
    instead of spending too much time fixing the package.
  - TODO test reset on those architectures

- Furthermore amd64 / ppc64 tests fail while armhf works.
    https://autopkgtest.ubuntu.com/packages/x/xtensor/impish/amd64
  - In debici this passes on amd64/arm64/armel/armhf/ppc64 but fails i386/s390x.
    https://ci.debian.net/packages/x/xtensor/
  - It appears to be a build error at first, but it really is an OOM (log below).
  - The problem that this isn't just "big packages and being ok" I tried with 3G/4G/8G
    memory. 4G still fails, 8G works so it is not an infinite-memory bug.
  - TODO mark as big package once 8G is confirmed

The test also has some control about its concurrency, but I tried to reduce it to -j 1 and it still fails with an OOM. But then it works with 3G which is an improvement.
TODO: submit to Debian to reduce the nproc in this test case.

[ 326.699865] Swap cache stats: add 0, delete 0, find 0/0
[ 326.699866] Free swap = 0kB
[ 326.699867] Total swap = 0kB
[ 326.699868] 1048443 pages RAM
[ 326.699869] 0 pages HighMem/MovableOnly
[ 326.699870] 43233 pages reserved
[ 326.699870] 0 pages hwpoisoned
[ 326.699871] Tasks state (memory values in pages):
[ 326.699872] [ pid ] uid tgid total_vm rss pgtables_bytes swapents oom_score_adj name
[ 326.699877] [ 280] 0 280 16183 668 110592 0 -250 systemd-journal
[ 326.699881] [ 305] 0 305 5604 875 61440 0 -1000 systemd-udevd
[ 326.699886] [ 317] 100 317 4003 490 77824 0 0 systemd-network
[ 326.699889] [ 398] 0 398 53869 4659 90112 0 -1000 multipathd
[ 326.699891] [ 426] 101 426 6025 1470 86016 0 0 systemd-resolve
[ 326.699894] [ 427] 102 427 22107 708 73728 0 0 systemd-timesyn
[ 326.699896] [ 432] 0 432 2079 1126 53248 0 0 haveged
[ 326.699898] [ 435] 0 435 657 375 40960 0 0 sh
[ 326.699900] [ 437] 0 437 1721 527 53248 0 0 cron
[ 326.699902] [ 439] 103 439 2125 941 53248 0 -900 dbus-daemon
[ 326.699906] [ 444] 0 444 20730 685 57344 0 0 irqbalance
[ 326.699908] [ 445] 0 445 7821 2956 102400 0 0 networkd-dispat
[ 326.699910] [ 446] 0 446 976 245 49152 0 0 rngd
[ 326.699912] [ 447] 104 447 55304 972 77824 0 0 rsyslogd
[ 326.699914] [ 448] 0 448 3450 608 61440 0 0 systemd-logind
[ 326.699915] [ 449] 0 449 98458 1307 118784 0 0 udisksd
[ 326.699917] [ 485] 0 485 3324 816 65536 0 -1000 sshd
[ 326.699919] [ 486] 0 486 1406 442 53248 0 0 agetty
[ 326.699921] [ 487] 0 487 1406 441 53248 0 0 agetty
[ 326.699923] [ 489] 0 489 1498 391 53248 0 0 agetty
[ 326.699924] [ 494] 0 494 58746 916 90112 0 0 polkitd
[ 326.699926] [ 644] 0 644 73985 1378 167936 0 0 packagekitd
[ 326.699928] [ 1668] 0 1668 657 33 40960 0 0 sh
[ 326.699931] [ 1669] 0 1669 2119 533 57344 0 0 su
[ 326.699932] [ 1671] 1000 1671 3839 957 73728 0 0 systemd
[ 326.699934] [ 1672] 1000 1672 25024 748 94208 0 0 (sd-pam)
[ 326.699936] [ 1677] 1000 1677 2280 492 61440 0 0 bash
[ 326.699938] [ 1684] 1000 1684 2009 516 57344 0 0 make
[ 326.699940] [ 1685] 1000 1685 2280 235 53248 0 0 bash
[ 326.699942] [ 1686] 1000 1686 2280 230 53248 0 0 bash
[ 326.699943] [ 1687] 1000 1687 1850 77 49152 0 0 tee
[ 326.699945] [ 1688] 1000 1688 1850 84 49152 0 0 tee
[ 326.699947] [ 1887] 1000 1887 2009 521 45056 0 0 make
[ 326.699949] [ 1890] 1000 1890 2009 543 49152 0 0 make
[ 326.699951] [ 1934] 1000 1934 2070 560 57344 0 0 make
[ 326.699953] [ 2179] 1000 2179 1701 23 45056 0 0 g++
[ 326.699954] [ 2180] 1000 2180 379282 366807 3063808 0 0 cc1plus
[ 326.699956] [ 2184] 1000 2184 1701 22 49152 0 0 g++
[ 326.699958] [ 2185] 1000 2185 437147 423981 3530752 0 0 cc1plus
[ 326.699960] [ 2199] 1000 2199 1701 22 57344 0 0 g++
[ 326.699962] [ 2200] 1000 2200 109270 97183 901120 0 0 cc1plus
[ 326.699964] [ 2204] 1000 2204 1701 22 49152 0 0 g++
[ 326.699965] [ 2205] 1000 2205 68071 56459 573440 0 0 cc1plus
[ 326.699967] oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=/,mems_allowed=0,global_oom,task_memcg=/user.slice/user-1000.slice/session-c2.scope,task=cc1plus,pid=2185,uid=1000
[ 326.699988] Out of memory: Killed process 2185 (cc1plus) total-vm:1748588kB, anon-rss:1695328kB, file-rss:596kB, shmem-rss:0kB, UID:1000 pgtables:3448kB oom_score_adj:0

tags: added: update-excuse
description: updated
Revision history for this message
Christian Ehrhardt  (paelzer) wrote :

The three TODOs that were identified are submitted.

Test reset on architectures that miss libxtensor-dev:
=> https://code.launchpad.net/~paelzer/britney/+git/hints-ubuntu/+merge/405894

Mark as big_package to provide the test more memory:
=> https://code.launchpad.net/~paelzer/autopkgtest-cloud/+git/autopkgtest-package-configs/+merge/405895

Submit to Debian to reduce the nproc in this test case:
=> https://salsa.debian.org/science-team/xtensor/-/merge_requests/1

Once at least the first two completed - and if big packages is only 4G (I was unable to find details today) then also the MR to Debian - are completed then these tests should work fine again (at least if the result of the local repro will work out the same in autopkgtest infrastructure).

Revision history for this message
Christian Ehrhardt  (paelzer) wrote :
Revision history for this message
Christian Ehrhardt  (paelzer) wrote :

It completed 7 minutes ago:
2021-07-20 11:53:50 CEST Published Impish release universe misc 0.22.0-5

Thereby this bug can be marked fixed.

Changed in xtensor (Ubuntu):
status: New → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.