some vc-dir notes

some vc-dir notes

[Tom Tromey]

After a moderate hiatus, today I tried Emacs CVS head and vc-dir
again.  I ran it on various parts of my GCC tree, checked out from svn
and modified in places.

First, nice!  It is usably fast.  This feature is probably what will
drive me to running CVS Emacs daily...

Now on to the bugs.

Based on the help for "=" I expected an easy way to make a diff of the
entire directory.  I couldn't figure this out.  I can mark files and
do it, but I was hoping for something simpler.

If I type "=" outside the content area of the buffer, I still get a
result.  Maybe I even wrote this part this way on purpose, I forget.
Right now I'm not a big fan of it though.  Maybe an operation with
point before the content should work on the whole tree.

I ran vc-dir on a directory with no changes and no unrecognized files.
The display is just empty, which feels odd.  I actually do miss the
more explicit running/not running status patch I wrote... maybe just a
footer would be nice.

In any case, when something is running the mode line says
"Waiting...".  I think that is a bit passive and "Working..." would be
a nicer.

If I type '=' on an unregistered file I get:

  require(vc-nil)
  vc-find-backend-function(nil diff)
  vc-call-backend(nil diff ("/home/tromey/gnu/Trunk/trunk/Old.jar") nil nil "*v\
c-diff*")
  vc-diff-internal(t ("/home/tromey/gnu/Trunk/trunk/Old.jar") nil nil t)
  vc-diff(nil t)
  call-interactively(vc-diff nil nil)

IMO it should just use whatever backend is displayed at the top of the
buffer.  In fact, using a different backend for any operation would
seem weird.

Tom

Re: some vc-dir notes

[Stefan Monnier]

> Based on the help for "=" I expected an easy way to make a diff of the
> entire directory.  I couldn't figure this out.  I can mark files and
> do it, but I was hoping for something simpler.

Adding directory entries in the display is being worked on.
The "." entry will allow applying a command to the whole tree.

> If I type "=" outside the content area of the buffer, I still get a
> result.

If it does, it's a hack.  An explicit "." entry sounds much better.

> If I type '=' on an unregistered file I get:
>   require(vc-nil)
[...]
> IMO it should just use whatever backend is displayed at the top of the
> buffer.  In fact, using a different backend for any operation would
> seem weird.

Yes, it's a bug I already pointed out: the backend to use should not be
associated with the file-name (as is currently the case) but should just
be a buffer-local variable.

This might require a fairly large change, tho.


        Stefan