You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This is triggered by the specific way Debian's autopkgtest framework sets up the sources. To illustrate, I was able to workaround this with this patch:
> Enter [It] [email protected] - /tmp/autopkgtest.lL1okL/build.rYm/real-tree/_build/src/github.com/containers/podman/test/e2e/quadlet_test.go:921 @ 12/06/24 15:12:56.818
Warning: Failed to chown /tmp/podman-e2e-2554937726/subtest-376925938/quadlet/[email protected]/10-env.conf to uid=1000, gid=0: lchown /tmp/podman-e2e-2554937726/subtest-376925938/quadlet/[email protected]/10-env.conf: operation not permitted
Running: /usr/libexec/podman/quadlet --user --no-kmsg-log /tmp/podman-e2e-2554937726/subtest-376925938/generated with QUADLET_UNIT_DIRS=/tmp/podman-e2e-2554937726/subtest-376925938/quadlet
< Exit [It] [email protected] - /tmp/autopkgtest.lL1okL/build.rYm/real-tree/_build/src/github.com/containers/podman/test/e2e/quadlet_test.go:921 @ 12/06/24 15:12:56.831 (12ms)
Apparently this is because the sources are installed into the build environment with uid=1000 but gid=0, and the copy operation is failing at setting the gid.
Is that chown really necessary?
Podman in a container
No
Privileged Or Rootless
Rootless
Upstream Latest Release
Yes
Additional environment details
Additional environment details
Additional information
Additional information like issue happens only occasionally or issue happens with a particular architecture or on a particular setting
The text was updated successfully, but these errors were encountered:
It may not be necessary for the purpose of the test. It looks like this function is only used by quadlet tests so we should be able to change it without worrying about effecting other tests.
Given quadlet only reads the source files it doesn't need any particular permissions as long as the test user can read the files.
Apparently this is because the sources are installed into the build environment with uid=1000 but gid=0, and the copy operation is failing at setting the gid.
Any reason you cannot chown to $UID:$GID for the entire source tree? That is what we are doing in CI.
Apparently this is because the sources are installed into the build environment with uid=1000 but gid=0, and the copy operation is >> failing at setting the gid.
Any reason you cannot chown to $UID:$GID for the entire source tree? That is what we are doing in CI.
Interesting.
To be frank, I haven't seriously considered that, but I am a bit pessimistic that this is a good idea. Let me think a bit more where exactly to implement this.
Issue Description
This is triggered by the specific way Debian's autopkgtest framework sets up the sources. To illustrate, I was able to workaround this with this patch:
Now I'm getting this output in the log:
Apparently this is because the sources are installed into the build environment with uid=1000 but gid=0, and the copy operation is failing at setting the gid.
Is that chown really necessary?
Podman in a container
No
Privileged Or Rootless
Rootless
Upstream Latest Release
Yes
Additional environment details
Additional environment details
Additional information
Additional information like issue happens only occasionally or issue happens with a particular architecture or on a particular setting
The text was updated successfully, but these errors were encountered: