Random selection of fixes and cleanups#171
Conversation
Based on the documentation dkms can do a simple action at a time. Although the code seems to disagree. Considering the complexity such multi-action might involve, to work properly, simply remove that feature. After staring at dozens of scripts using dkms, I'm yet to see one which use this undocumented feature. Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Move all but the "add_module" check_module_args() call earlier. As seen in the removal actions - we do not need to be root, only to check the user has provided sane input params. The add_module remains as-is since the args are optional. Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Function is used in a single, place only to call a variant of module_is_added_or_die, followed up with an essentially duplicate "is_added" check. Just kill it all off, adding module_is_added_or_die() into the main switch, like many other actions. Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reuse the existing maybe_{build,install}_module to simplify the command.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Just use the trap and kill off the manual rm instances. We've been doing rather poor job with them anyway. Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Bit of a no-brainer - use local read-only variables so the codebase is actually readable. Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Tar (GNU or BSD) has been able to detect the format for a while. Drop the local auto-detection code. In addition, consistently use -C instead of open-coding it. Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Currently we check for write permissions only if --archive is provided and is pointing to an existing folder. In any other case we don't bother, so we get cryptic error messages. Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Current codebase is mixing indentation levels - be that tabs or spaces, to outright omitting a level. This patch addresses a few cases of the latter, by breaking/returning early. This way the developer does not need to read/parse through the remaining code, since the codeflow is clearer. Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
|
Seems like we're trying to sign the modules twice atm - once in I'm inclined that we should nuke |
Fine to me, would you do it in a separate merge request? I also created this issue as a reminder for myself, maybe you can attach to it with a separate merge request: Also, I don't have project access to change members, otherwise I would give you direct commit access. Thanks! |
|
Yup, will do separate MR maybe tomorrow. Mind you I don't know much about module signing, so I won't be tackling #162 anytime soon. Thanks for the vote of confidence - wrt commit access. I think it's better to use PRs, regardless if one has direct commit access. |
2 participants
Grouping the lot in a single MR since they fit no particular topic: