Most probable source of the problem:
https://github.com/dell/dkms/blob/218cd4ee8b115a542f8c2ff98701bf92e39a8908/dkms#L1064
Impact:
During CI, we are not able to catch errors in the post-install script because dkms simply assumes such scripts work without any error and continues, leaving a possibly undefined and broken system state behind. In the vast amount of logging generated by package installers, such an error just disappears in the noise. The user has no clue that something failed, neither they can get an idea of how to fix it.
Suggestion:
dkms should die if an error happens in the run_build_script phase, probably also cleaning up as good as possible. This may require to bubble the error up to the caller instead of handling it right in this function.
Most probable source of the problem:
https://github.com/dell/dkms/blob/218cd4ee8b115a542f8c2ff98701bf92e39a8908/dkms#L1064
Impact:
During CI, we are not able to catch errors in the post-install script because dkms simply assumes such scripts work without any error and continues, leaving a possibly undefined and broken system state behind. In the vast amount of logging generated by package installers, such an error just disappears in the noise. The user has no clue that something failed, neither they can get an idea of how to fix it.
Suggestion:
dkms should die if an error happens in the run_build_script phase, probably also cleaning up as good as possible. This may require to bubble the error up to the caller instead of handling it right in this function.