comp_compgen: option -C no longer checks if OPTARG is empty

[YoruStar]

The -C option in _comp_compgen, similar to the -P option in bash's compgen, should support passing an empty string, for example: compgen -P ""-f

fix #1260

[akinomyoga]

I believe this should be fixed at the side of completions/_umount.linux. It is in general suspicious to specify an invalid directory name so I don't think it is a good idea to suppress the error message unconditionally. The directory name should be checked at the caller side. In the case of _umount.linux, it seems that the error still had occurred before 6b3dfa5, but the message was redirected to /dev/null as cd is performed as command cd -- "$dircur" 2>/dev/null. The caller should ensure the validity of the path. In the present case, the caller can be updated to _comp_compgen -aC "${dircur:--}" ...

[YoruStar]

I believe this should be fixed at the side of completions/_umount.linux. It is in general suspicious to specify an invalid directory name so I don't think it is a good idea to suppress the error message unconditionally. The directory name should be checked at the caller side. In the case of _umount.linux, it seems that the error still had occurred before 6b3dfa5, but the message was redirected to /dev/null as cd is performed as command cd -- "$dircur" 2>/dev/null. The caller should ensure the validity of the path. In the present case, the caller can be updated to _comp_compgen -aC "${dircur:--}"

I tried command cd -- "" and it did not fail.

$ command cd -- ""
$ echo $?
0

In the changes at 6b3dfa5, there are many modifications besides _umount.linux, and these may also potentially have the same issue.

Moreover, in the _comp_compgen__call_generator function, there is already a check to determine if the $_dir variable is empty, making the logic for checking if $OPTARG is empty in the -C option somewhat redundant.

Of course, aside from the umount command, I haven't found similar issues elsewhere, so modifying _umount.linux alone is also feasible.

[akinomyoga]

I tried command cd -- "" and it did not fail.

$ command cd -- ""
$ echo $?
0

OK, thanks.

Nevertheless, I don't think it would be right just to suppress the error message when one sees an error message.

In the changes at 6b3dfa5, there are many modifications besides _umount.linux, and these may also potentially have the same issue.

Yes, they should also be fixed if they would pass an invalid directory. Before my first reply, I actually had checked it after I saw your PR. The cases for feh and sbopkg should also be fixed properly. The other cases would probably be fine, but we need to double-check them.

Moreover, in the _comp_compgen__call_generator function, there is already a check to determine if the $_dir variable is empty,

No, it's used to check whether the -C option is specified or not. In the current master, exactly because we reject an empty directory, we can use [[ $_dir ]] to check whether the -C option is specified. If you break the assumption, it doesn't work as expected, so one needs to use another variable to store whether the -C option is specified.

making the logic for checking if $OPTARG is empty in the -C option somewhat redundant.

This is false. It's the opposite. It relies on the fact that the empty directory for -C is rejected.

Of course, aside from the umount command, I haven't found similar issues elsewhere, so modifying _umount.linux alone is also feasible.

Could you also look at feh and sbopkg? I haven't looked at them closely, but they might also need modifications. If you wouldn't have time, I'll later look at them someday.