bug#34949: 27.0.50; Docstring of `vc-deduce-fileset' incomplete

[Dmitry Gutov]

On 24.02.2020 1:25, Juri Linkov wrote:

> > Um, not really. Just only adding the 'V' binding to VC-Dir buffers, where
> > it would use the existing known fileset. But no 'C-x v V' binding.
>
> Should then 'V' typed in VC-Dir operate on edited files only,
> or try to use all files including unregistered?

Which option would you choose?

Do you know why we generally try to only operate on the files in the 
same status?

> > If that would both satisfy you and simplify the implementation, of course.
>
> What about invoking state-changing VC-operations from Dired?
> Should typing 'v' in Dired open the VC-Dir buffer?

No, I just wouldn't do that.

> BTW, why 'C-x v d RET' requires a confirmation?
> This additional 'RET' is too annoying.

You participated in this discussion: 
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=12492

Since then, we've had at least another one on the same subject. The 
consensus was the current behavior. If you can find that discussion, 
please go ahead and revive it if you like. Not in this bug report, though.

> Maybe like a proposal in bug#39704 to use a prefix arg for
> 'C-u M-x vc-print-branch-log' to ask for a directory, then
> 'C-u C-x v d' could do the same and ask for a root directory,
> otherwise without prefix arg 'C-x v d' could use the default root
> without using a prompt.

I would like that, personally. But see above.