Closed lazka closed 3 months ago
It's not clear if this helps, so closing for now.
If anyone ends up here and thinks otherwise, please speak up!
I tried to do a little investigation, but keep in mind I don't really deal with docker. What I saw is that in the current normal case, when using podman run
, PID 1 is filled by the instance of bash
running the msys2
script. It seems bash is reaping child processes. When using this PR, PID 1 is filled by tini
, and PID 2 is the bash
running the msys2
script.
However, if I attempt to reproduce what GHA does by doing
podman run -it --entrypoint /usr/bin/tail ghcr.io/msys2/msys2-docker-experimental -f /dev/null
, PID 1 is still tail
and I get zombie processes even with this change.
See #5