vc-svn-dir-status

vc-svn-dir-status

[Nick Roberts]

2009-01-31  Dan Nicolaescu  <dann@ics.uci.edu>

	* vc-svn.el (vc-svn-dir-status): Disable the vc-stay-local-p
	logic, it makes the makes the normal case too slow.

You reported this slowdown as bug#2117.  It probably manifested itself now
as I recently fixed some problems with vc-stay-local here:

2009-01-30  Nick Roberts  <nickrob@snap.net.nz>

	* vc-svn.el (vc-svn-dir-status): Correct logic for remote
	connection.  (Bug#2093)

In response to bug#2117 I suggested that I suggested that vc-stay-local-p need
not be slow if vc-svn-registered need parsed the SVN controlled text-base
directory instead of executing "svn status -v DIR".

Why did you not reply to this suggestion?

We could undo each other's changes but it probably wouldn't be very
constructive.  I spent a fair amount of time going through the stay-local logic
trying to preserve functionality that vc-dired had and that Stefan has agreed
is desirable.

Frankly, I don't think that adding a line like:

	 (remote (or t (not local) (eq local 'only-file))))

can really be considered a contribution to Emacs development.

-- 
Nick                                           http://www.inet.net.nz/~nickrob

Re: vc-svn-dir-status

[Stefan Monnier]

> In response to bug#2117 I suggested that I suggested that vc-stay-local-p need
> not be slow if vc-svn-registered need parsed the SVN controlled text-base
> directory instead of executing "svn status -v DIR".

That would be a good change.  Many VC backends do similar things to
speed up this common operation.  Tho, admittedly, this is usually done
for files rather than for dirs, so there might be unexpected problems.


        Stefan

Re: vc-svn-dir-status

[Nick Roberts]

 > > In response to bug#2117 I suggested that I suggested that vc-stay-local-p
 > > need not be slow if vc-svn-registered need parsed the SVN controlled
 > > text-base directory instead of executing "svn status -v DIR".
 > 
 > That would be a good change.  Many VC backends do similar things to
 > speed up this common operation.  Tho, admittedly, this is usually done
 > for files rather than for dirs, so there might be unexpected problems.

Maybe the text-base directory is irrelevant but I've tried to understand why
"svn status" gets called twice in this case.  I think in this first case,
vc-local-p -> vc-registered -> vc-svn-registered it just needs the status of
one file, namely the directory, and it needn't recurse.  If that's so "svn
status -v -N DIR" would presumably be quicker and only the first line need be
parsed.  vc-cvs-registered does nothing in the case of a directory argument,
presumably because, in CVS, directories aren't under version control.

To add to my confusion I don't understand the documentation string of
vc-svn-parse-status:

   ...Unless FILENAME is non-nil, parse only information about FILENAME and
   return its status.

It always appears to be non-nil but if it was nil how could you parse
information about it?

-- 
Nick                                           http://www.inet.net.nz/~nickrob