Re: *vc-diff* buffers are read-only, but M-k works

Re: *vc-diff* buffers are read-only, but M-k works

[Chong Yidong]

David Reitter <david.reitter@gmail.com> wrote:

> M-k works just fine in read-only *vc-diff* buffers.
>
> However, undoing an M-k command is impossible, as is cutting, pasting,  
> or all the other things that I did all the time in *vc-diff* buffers  
> in Emacs 22.
>
> The M-k thing is a bug, the read-only (default?) setting is annoying.

The purpose of making the vc-diff buffer read-only (2007-11-10 change to
vc.el) was so that people can use the read-only diff-mode bindings like
"k" and "p".  However, it's counter-intuitive that undo does not work.
Probably the easiest fix is to define a diff-undo command that lifts the
read-only restriction temporarily.  Any thoughts?

Re: *vc-diff* buffers are read-only, but M-k works

[Juanma Barranquero]

On Wed, Mar 4, 2009 at 16:41, Chong Yidong <cyd@stupidchicken.com> wrote:

> However, it's counter-intuitive that undo does not work.
> Probably the easiest fix is to define a diff-undo command that lifts the
> read-only restriction temporarily.  Any thoughts?

Most operations on diff-mode buffers do not change the buffer or are
easily reversible (diff-reverse-direction, diff-unified->context,
etc.). Making the buffer read-write is a C-x C-q away, if the user
needs to use undo or normal editing commands.

What is a bit of a pain is that the user cannot easily setup vc-diff
to default to a read-write buffer, because there's no vc-diff-hook and
`(setq buffer-read-only t)' is done after calling diff-mode, so
diff-mode-hook isn't useful either. The only way right now (other than
modifying the source) is to advise vc-diff.

My 0.02€

    Juanma

Re: *vc-diff* buffers are read-only, but M-k works

[Lars Magne Ingebrigtsen]

Chong Yidong <cyd@stupidchicken.com> writes:

> The purpose of making the vc-diff buffer read-only (2007-11-10 change to
> vc.el) was so that people can use the read-only diff-mode bindings like
> "k" and "p".  However, it's counter-intuitive that undo does not work.
> Probably the easiest fix is to define a diff-undo command that lifts the
> read-only restriction temporarily.  Any thoughts?

Perhaps a more general solution would be possible here.  I think there
probably are other "mostly-read-only" modes where you have one or two
special commands that alter the buffer, anyway.  Like with `M-k'.
Perhaps there could be a buffer-local variable like
`undo-read-only-buffer' that if set, would allow `undo' to work in these
cases.

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/

Re: *vc-diff* buffers are read-only, but M-k works

[Stefan Monnier]

>> The purpose of making the vc-diff buffer read-only (2007-11-10 change to
>> vc.el) was so that people can use the read-only diff-mode bindings like
>> "k" and "p".  However, it's counter-intuitive that undo does not work.
>> Probably the easiest fix is to define a diff-undo command that lifts the
>> read-only restriction temporarily.  Any thoughts?

> Perhaps a more general solution would be possible here.  I think there
> probably are other "mostly-read-only" modes where you have one or two
> special commands that alter the buffer, anyway.  Like with `M-k'.
> Perhaps there could be a buffer-local variable like
> `undo-read-only-buffer' that if set, would allow `undo' to work in these
> cases.

Or we could record in the buffer-undo-list the fact the the modification
was done with inhibit-read-only bound to non-nil, and bind it similarly
during the undo.


        Stefan

Re: bug#2553: *vc-diff* buffers are read-only, but M-k works

[Lars Magne Ingebrigtsen]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> Perhaps a more general solution would be possible here.  I think there
>> probably are other "mostly-read-only" modes where you have one or two
>> special commands that alter the buffer, anyway.  Like with `M-k'.
>> Perhaps there could be a buffer-local variable like
>> `undo-read-only-buffer' that if set, would allow `undo' to work in these
>> cases.
>
> Or we could record in the buffer-undo-list the fact the the modification
> was done with inhibit-read-only bound to non-nil, and bind it similarly
> during the undo.

That would be even better, and is even more general.

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/