Re: master 4803fba487 1/2: 'C-x v v' on a diff buffer commits it as a patch (bug#52349)

[Dmitry Gutov]

On 30.08.2022 19:37, Eli Zaretskii wrote:
> >   As long as we fundamentally agree on the behavior.
> I think the "C-x v D" step should be part of the command's execution.
> It makes little sense to me to ask the user to prepare the diffs in
> advance, since conceptually the changes are already on disk.  Starting
> from diffs makes sense when you want to apply them first, and only
> after that commit.  Which is not the case here.

The user can choose the files first (e.g. choose a directory from 
VC-Dir), or just pick one file they are currently editing. Then see the 
diff, and make the choice of making a commit from it.

What you're suggesting could possibly improve discoverability (if the 
user manages to read the doc and find out about the meaning of C-0 
prefix), but WRT flexibility, the effect can be negative. That is, 
unless we keep the new meaning of 'C-x v v' in diff buffers, the ability 
to commit an arbitrary diff (not applied yet) will go away. And it's 
what the author of the top message in this tree was particularly happy 
about.

And as for the length of key sequences, in VC-Dir, 'D C-x v v' is about 
the same as 'C-0 C-x v v'.

> I understand that typing "C-x v v" in a diffs buffer is used here as
> an indication that this particular variant of "committing" is
> requested, but that's a weak justification, IMO.  It could even
> backfire: the user could be in the diffs by sheer luck (or lack
> thereof), and type "C-x v v" without any intention to select parts of
> the changes.

If they have no intention of creating a commit from that diff and press 
'C-x v v' by accident, they can simply 'C-c C-k' out of the resulting 
Log-Edit buffer. That will bring them back to the diff, restoring the 
previous window configuration.