bug#18475: 24.4.50; Wdired: cannot use C-k to delete a dir name if -F switch used

[Basil L. Contovounesios]

Stephen Berman <stephen.berman@gmx.net> writes:

> This happens not only with `/' but also with the other ls file indicator
> characters appended when using the -F switch.  Making C-k work as
> expected for these cases is a small fix.  However, the same issue also
> arises with symlinks, whether or not -F is used, and it does not seem as
> straightforward to deal with this case.  The attached patch (against
> master) tries to ensure the following behavior:
>
> - Typing `C-k' with point just before the first character of a file name
>   ending with an indicator character (using -F), or of link name,
>   deletes the file name/symlink in WDired, but it is restored but marked
>   for deletion on returning to Dired.  The same also happens when just
>   deleting the link name in WDired.
>
> - Typing `C-k' with point on such a file or link name, but after the
>   first character, deletes the rest, resulting in renaming on returning
>   to Dired.
>
> - Deleting the indicator character is possible in WDired but a noop: the
>   character is restored on returning to Dired.  (In the current code
>   without the patch, the indicator characters are read-only, but I had
>   to change that to make C-k work.)
>
> With symlinks:
>
> - The patch preserves the current behavior that an edit of the target
>   name (possible when wdired-allow-to-redirect-links is non-nil, as it
>   is by default) is saved on returning to Dired, and if the target name
>   is deleted, then the new target on returning to Dired is "/dev/null".
>
> - Editing (changing or deleting) the string " -> " between the link and
>   target names is possible in WDired but a noop: the symlink is
>   unaltered on returning to Dired.  (In the current code without the
>   patch, " -> " is read-only, but I had to change that to make C-k work.)
>
> I've tested these cases, but it is quite possible that I overlooked some
> variants or other cases, so I'd appreciate testing and feedback from
> others.  (Also, the code still needs more commenting and probably
> cleaning up.)

I can confirm your patch fixes the issue in the OP, but I haven't tested
it extensively.

Thanks for working on this,

-- 
Basil