Hello,
On 2025-06-05 20:41, Paul Gevers wrote:
I was debugging performance issues on ci.debian.net today, and noticed
that s390x was running into memory problems. openbabel quickly became a
suspect and indeed it turns out to use an insane amount of it during the
test.
I have spotted at least this (not the worse) in top:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+
COMMAND
653405 xxxxx 20 0 90.1g 26.7g 9908 D 6.3 85.0 10:04.02
python3
And the peak I spotted while looping free -h was 72 GB added by the test
and then I killed the test run.
I'll add openbabel to the reject_list and will remove it again when this
bug report is closed.
Certain modes of operation of openbabel are known to leak memory, for
example [1][2][3]. I did not investigate further, but running tests
under valgrind might be the first step to that direction.
Best,
Andrius
[1] https://github.com/openbabel/openbabel/issues/2370
[2] https://github.com/openbabel/openbabel/pull/2772
[3] https://github.com/openbabel/openbabel/pull/2735
