bug#2092: 23.0.60; vc-svn-diff

bug#2092: 23.0.60; vc-svn-diff

[Nick Roberts]

vc-svn-diff fails when oldvers equals (vc-working-revision f).
In that case "svn diff" executes with no -r argument and only gives
a diff if the file is locally modified.

To see this bug, do vc-print-log on a file under Subversion control that
needs an update (newer revisions have been committed by someone else).
Place the cursor over the revision after (in time) the working-revision (the
revison with the number in the modeline) and press d (log-viw-diff).

I think this clause needs to be removed:

  (and oldvers
       files
       (catch 'no
	 (dolist (f files)
	   (or (equal oldvers (vc-working-revision f))
	       (throw 'no nil)))
	 t)
       ;; Use nil rather than the current revision because svn handles
       ;; it better (i.e. locally).  Note that if _any_ of the files
       ;; has a different revision, we fetch the lot, which is
       ;; obviously sub-optimal.
       (setq oldvers nil))

I don't see how it could ever work (please note that I'm not saying that
it could never work just that I don't see how it could).

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

bug#2092: 23.0.60; vc-svn-diff

[Stefan Monnier]

> vc-svn-diff fails when oldvers equals (vc-working-revision f).
> In that case "svn diff" executes with no -r argument and only gives
> a diff if the file is locally modified.

That's the right behavior when newvers is nil, isn't it?

> I don't see how it could ever work (please note that I'm not saying that
> it could never work just that I don't see how it could).

I think it was written assuming (incorrectly) that "newvers == nil".
Is the bug fixed if you add (null newvers) to the conjunction?


        Stefan


PS: I see you've already removed this code, but I remember we added
specifically upon request from some users, so it would be better to fix
it than to remove it.

bug#2092: 23.0.60; vc-svn-diff

[Nick Roberts]

 > > vc-svn-diff fails when oldvers equals (vc-working-revision f).
 > > In that case "svn diff" executes with no -r argument and only gives
 > > a diff if the file is locally modified.
 > 
 > That's the right behavior when newvers is nil, isn't it?
 > 
 > > I don't see how it could ever work (please note that I'm not saying that
 > > it could never work just that I don't see how it could).
 > 
 > I think it was written assuming (incorrectly) that "newvers == nil".
 > Is the bug fixed if you add (null newvers) to the conjunction?
 > 
 > 
 >         Stefan
 > 
 > 
 > PS: I see you've already removed this code, but I remember we added
 > specifically upon request from some users, so it would be better to fix
 > it than to remove it.

OK, I've looked through the archives.  I find it odd because:

1) I think that svn should be able to work out whether it can handle a
   request locally (a bug in svn?).
2) If you do "C-x v =" there is no problem.
3) I can only see a problem with "C-u C-x v =" with the default values which
   is eqivalent to 2).

So it looks very much like a corner case.

Anyway I've done as you say.

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

bug#2092: 23.0.60; vc-svn-diff

[Stefan Monnier]

> OK, I've looked through the archives.  I find it odd because:

> 1) I think that svn should be able to work out whether it can handle a
>    request locally (a bug in svn?).

Agreed.

> 2) If you do "C-x v =" there is no problem.

Yes, in that case OLDVERS is nil.

> 3) I can only see a problem with "C-u C-x v =" with the default values which
>    is eqivalent to 2).

Indeed.  Basically, `svn diff foo' works locally, whereas `svn diff -r
REV foo' never works locally, even if VERS is the current working
revision (in which case the result should be identical to `svn diff
foo').  It's arguably a misfeature of `svn'.

> So it looks very much like a corner case.

It is.

> Anyway I've done as you say.

Thank you.


        Stefan

bug#2092: marked as done (23.0.60; vc-svn-diff)

[Emacs bug Tracking System]

[-- Attachment #1: Type: text/plain, Size: 856 bytes --]


Your message dated Thu, 5 Feb 2009 00:03:01 +1300
with message-id <18825.30181.243954.921791@kahikatea.snap.net.nz>
and subject line Re: bug#2092: 23.0.60; vc-svn-diff
has caused the Emacs bug report #2092,
regarding 23.0.60; vc-svn-diff
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner@emacsbugs.donarmstrong.com
immediately.)


-- 
2092: http://emacsbugs.donarmstrong.com/cgi-bin/bugreport.cgi?bug=2092
Emacs Bug Tracking System
Contact owner@emacsbugs.donarmstrong.com with problems

[-- Attachment #2: Type: message/rfc822, Size: 3027 bytes --]

From: Nick Roberts <nickrob@snap.net.nz>
To: emacs-pretest-bug@gnu.org
Subject: 23.0.60; vc-svn-diff
Date: Wed, 28 Jan 2009 19:34:34 +1300 (NZDT)
Message-ID: <20090128063434.74C018FC6D@kahikatea.snap.net.nz>


vc-svn-diff fails when oldvers equals (vc-working-revision f).
In that case "svn diff" executes with no -r argument and only gives
a diff if the file is locally modified.

To see this bug, do vc-print-log on a file under Subversion control that
needs an update (newer revisions have been committed by someone else).
Place the cursor over the revision after (in time) the working-revision (the
revison with the number in the modeline) and press d (log-viw-diff).

I think this clause needs to be removed:

  (and oldvers
       files
       (catch 'no
	 (dolist (f files)
	   (or (equal oldvers (vc-working-revision f))
	       (throw 'no nil)))
	 t)
       ;; Use nil rather than the current revision because svn handles
       ;; it better (i.e. locally).  Note that if _any_ of the files
       ;; has a different revision, we fetch the lot, which is
       ;; obviously sub-optimal.
       (setq oldvers nil))

I don't see how it could ever work (please note that I'm not saying that
it could never work just that I don't see how it could).

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



[-- Attachment #3: Type: message/rfc822, Size: 2771 bytes --]

From: Nick Roberts <nickrob@snap.net.nz>
To: Stefan Monnier <monnier@iro.umontreal.ca>
Cc: 2092-done@emacsbugs.donarmstrong.com, emacs-pretest-bug@gnu.org
Subject: Re: bug#2092: 23.0.60; vc-svn-diff
Date: Thu, 5 Feb 2009 00:03:01 +1300
Message-ID: <18825.30181.243954.921791@kahikatea.snap.net.nz>

 > > vc-svn-diff fails when oldvers equals (vc-working-revision f).
 > > In that case "svn diff" executes with no -r argument and only gives
 > > a diff if the file is locally modified.
 > 
 > That's the right behavior when newvers is nil, isn't it?
 > 
 > > I don't see how it could ever work (please note that I'm not saying that
 > > it could never work just that I don't see how it could).
 > 
 > I think it was written assuming (incorrectly) that "newvers == nil".
 > Is the bug fixed if you add (null newvers) to the conjunction?
 > 
 > 
 >         Stefan
 > 
 > 
 > PS: I see you've already removed this code, but I remember we added
 > specifically upon request from some users, so it would be better to fix
 > it than to remove it.

OK, I've looked through the archives.  I find it odd because:

1) I think that svn should be able to work out whether it can handle a
   request locally (a bug in svn?).
2) If you do "C-x v =" there is no problem.
3) I can only see a problem with "C-u C-x v =" with the default values which
   is eqivalent to 2).

So it looks very much like a corner case.

Anyway I've done as you say.

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