#1150251 kill: non-zero exit despite kill syscall succeeding

Package:
procps
Source:
procps
Description:
/proc file system utilities
Submitter:
Johannes Schauer Marin Rodrigues
Date:
2026-10-14 11:11:01 UTC
Severity:
normal
Tags:
#1150251#5
Date:
2026-10-07 15:31:21 UTC
From:
To:
Hello procps maintainers,

my package mmdebstrap has its autopkgtest failing because when sending a signal
to a process group using a negative pid, kill will exit with a non-zero exit
status even if the system call succeeds.

Here is the failure of the autopkgtest in salsaci:

https://salsa.debian.org/debian/mmdebstrap/-/jobs/10435775

And here on ci.debian.net:

https://ci.debian.net/packages/m/mmdebstrap/testing/armhf/76490131/

In both cases, the sigint-during-customize-hook test of mmdebstrap fails at the
point where /bin/kill (so definitely not kill from the shell) is sending a
SIGINT to a negative process. Since the script is run with set -e, it fails at
that point.

I reported the issue upstream here 3 weeks ago:

https://gitlab.com/procps-ng/procps/-/work_items/438

Luckily, upstream was very quick with fixing this with this commit in the
master branch:

https://gitlab.com/procps-ng/procps/-/commit/a4632f5cc97815a56b1afb9c294e2038cfb35b68

Thus, I'm tagging this bug with "upstream" as well as "patch".

This bug blocks #1141078 which is about the mmdebstrap autopkgtest failing.

Please consider either asking upstream for another release containing the fix
or backporting the fix from upstream to the version in Debian until a new
upstream release happens.

Since this bug is blocking my package mmdebstrap from transitioning to testing
I also would offer an NMU with this change if you are currently short on time.

Thanks!

cheers, josch

#1150251#10
Date:
2026-10-14 11:08:22 UTC
From:
To:
Hi Craig,

Quoting Johannes Schauer Marin Rodrigues (2026-10-07 17:31:21)

I just noticed that you seem to be also the upstream contact for procps!

Could you give me an estimate about a rough timeframe in which either a new
procps release can happen or the patch you commited to procps git (thanks)
could be backported to procps in Debian?

I am wondering whether I should add a random patch to the test such that it
ignores the non-zero exit for now or whether I should wait for procps to get
fixed?

The problem is not only in the mmdebstrap autopkgtest job but also in the
jenkins job. The only test that fails is the one that calls kill with a
negative PID:
https://jenkins.debian.net/job/mmdebstrap-jenkins-worker/1554/console

Thanks!

cheers, josch