bug#14242: vc-next-action with registered file [VC-mode with subversion 1.7.8]

bug#14242: vc-next-action with registered file [VC-mode with subversion 1.7.8]

[David Reitter]

X-Debbugs-CC: Robert Goldman <rpgoldman@sift.info>

Begin forwarded message:

> From: Robert Goldman <rpgoldman@sift.info>
> Subject: [aquamacs-devel] I'm not sure this is a bug.... [VC-mode with subversion 1.7.8]
> Date: April 21, 2013 6:40:08 PM EDT
> To: Development of Aquamacs Emacs <aquamacs-devel@aquamacs.org>
> Reply-To: aquamacs-devel@googlegroups.com
> 
> ... but using vc-mode on nightlies with subversion repositories, I am
> getting an odd behavior where ^X-V-V (vc-next-action) on a registered
> file gives me an error message that the file is already registered.
> 
> I have looked into this, and it seems like what's happening is that
> emacs mistakenly thinks the file is not registered, so decides the next
> action is to register the file.  Then, when it attempts to do so
> (presumably with "svn add"), it gets an error message.
> 
> If I do vc-next-action again, immediately following the one that yielded
> an error, then I can correctly make a commit.
> 
> As an example, here's what happens if I trace vc-deduce-fileset:
> 
> 1 -> (vc-deduce-fileset nil t state-model-only-files)
> 1 <- vc-deduce-fileset: (SVN (#1="/Users/rpg/org/notes.org") (#1#)
> unregistered implicit)
> 
> You will see that it believes the file to be unregistered (at least
> that's what I guess this means).
> 
> When I invoke vc-next-action again on the same file, I see this:
> 
> 1 -> (vc-deduce-fileset nil t state-model-only-files)
> 1 <- vc-deduce-fileset: (SVN (#1="/Users/rpg/org/notes.org") (#1#)
> edited implicit)
> 
> Anyone else seeing this?
> 
> svn version 1.7.8 from MacPorts.
> 
> It looks to me like "svn status -v" with this version of subversion may
> generate something that vc-svn doesn't expect:
> 
> [rpgoldman-3:~/org] rpg% svn status -v notes.org
>             21567    21567 rpgoldman    notes.org
> [rpgoldman-3:~/org] rpg% svn status notes.org
> [rpgoldman-3:~/org]
> 
> GNU Emacs 24.3.50.3 (x86_64-apple-darwin10.8.0, NS apple-appkit-1038.36)
> of 2013-04-04 on asura.local - Aquamacs Distribution 3.xdev
> 
> [yeah, time to pull a newer nightly...]
> 
> Best,
> r
> 
> 
> -- 
> You received this message because you are subscribed to the Google Groups "aquamacs-devel" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to aquamacs-devel+unsubscribe@googlegroups.com.
> 
> 

bug#14242: vc-next-action with registered file [VC-mode with subversion 1.7.8]

[Glenn Morris]

David Reitter wrote:

>> ... but using vc-mode on nightlies with subversion repositories, I am
>> getting an odd behavior where ^X-V-V (vc-next-action) on a registered
>> file gives me an error message that the file is already registered.

Not reproducible on GNU/Linux with Emacs 24.3 and svn 1.7.9.

>> It looks to me like "svn status -v" with this version of subversion may
>> generate something that vc-svn doesn't expect:
>> 
>> [rpgoldman-3:~/org] rpg% svn status -v notes.org
>>             21567    21567 rpgoldman    notes.org
>> [rpgoldman-3:~/org] rpg% svn status notes.org

Looks totally normal to me. What's supposed to be odd?

>> GNU Emacs 24.3.50.3 (x86_64-apple-darwin10.8.0, NS apple-appkit-1038.36)
>> of 2013-04-04 on asura.local - Aquamacs Distribution 3.xdev

bug#14242: vc-next-action with registered file [VC-mode with subversion 1.7.8]

[Glenn Morris]

Glenn Morris wrote:

> Not reproducible on GNU/Linux with Emacs 24.3 and svn 1.7.9.

Oh, but maybe with trunk.

bug#14242: vc-next-action with registered file [VC-mode with subversion 1.7.8]

[Glenn Morris]

Glenn Morris wrote:

> Glenn Morris wrote:
>
>> Not reproducible on GNU/Linux with Emacs 24.3 and svn 1.7.9.
>
> Oh, but maybe with trunk.

Thought I saw it once while testing, can't reproduce it now.

bug#14242: vc-next-action with registered file [VC-mode with subversion 1.7.8]

[Robert Goldman]

Glenn Morris wrote:
> David Reitter wrote:
> 
>>> ... but using vc-mode on nightlies with subversion repositories, I am
>>> getting an odd behavior where ^X-V-V (vc-next-action) on a registered
>>> file gives me an error message that the file is already registered.
> 
> Not reproducible on GNU/Linux with Emacs 24.3 and svn 1.7.9.
> 
>>> It looks to me like "svn status -v" with this version of subversion may
>>> generate something that vc-svn doesn't expect:
>>>
>>> [rpgoldman-3:~/org] rpg% svn status -v notes.org
>>>             21567    21567 rpgoldman    notes.org
>>> [rpgoldman-3:~/org] rpg% svn status notes.org
> 
> Looks totally normal to me. What's supposed to be odd?
> 
>>> GNU Emacs 24.3.50.3 (x86_64-apple-darwin10.8.0, NS apple-appkit-1038.36)
>>> of 2013-04-04 on asura.local - Aquamacs Distribution 3.xdev

The former shows what happens with -v which is, IIUC, how the latest
vc-svn invokes it.  The latter (empty) is what happens without -v.  I
thought I was seeing vc-svn failing to parse the verbose output with the
cryptic numbers, and mistakenly concluding that the file was not registered.

I am not seeing this 100% of the time, either.

But I just saw a related thing happening:  VC-mode accepted a commit
from me where there was no actual change in the file (the commit did not
go through to the svn repo).  Here before trying vc-next-action, I
checked svn status:

[rpgoldman-3:~/org] rpg% svn status -v todo.org
             21586    21586 rpgoldman    todo.org
[rpgoldman-3:~/org] rpg% svn status todo.org
[rpgoldman-3:~/org] rpg%

I *conjecture* that the fact that there was output from svn status is
what led vc-svn to think there was a modification and present me with
the log window to make a commit.

I confess to not understanding the innards of vc-svn that well...

bug#14242: vc-next-action with registered file [VC-mode with subversion 1.7.8]

[Glenn Morris]

Robert Goldman wrote:

> [rpgoldman-3:~/org] rpg% svn status -v todo.org
>              21586    21586 rpgoldman    todo.org
> [rpgoldman-3:~/org] rpg% svn status todo.org
> [rpgoldman-3:~/org] rpg%

Again, this is totally normally svn status output that vc-svn.el has
always handled.

bug#14242: vc-next-action with registered file [VC-mode with subversion 1.7.8]

[Robert Goldman]

Glenn Morris wrote:
> Robert Goldman wrote:
>
>> [rpgoldman-3:~/org] rpg% svn status -v todo.org
>>              21586    21586 rpgoldman    todo.org
>> [rpgoldman-3:~/org] rpg% svn status todo.org
>> [rpgoldman-3:~/org] rpg%
>
> Again, this is totally normally svn status output that vc-svn.el has
> always handled.

Thanks.  Question: isn't the stuff from my trace of vc-svn wrong?  I.e., do we agree that vc-svn is incorrectly seeing a file that's unregistered?  If that's the case, I can try to zero in on why I'm getting these bad results, and provide you some more useful feedback.

Also, is there some way I can examine the copy of vc-svn.el that I have to make sure that you and I are talking about the same thing?

Thanks for your time.

Best,
Robert

bug#14242: vc-next-action with registered file [VC-mode with subversion 1.7.8]

[Glenn Morris]

(Please keep the debbugs address cc'd)

Robert Goldman wrote:

> Thanks. Question: isn't the stuff from my trace of vc-svn wrong? I.e.,
> do we agree that vc-svn is incorrectly seeing a file that's
> unregistered? If that's the case, I can try to zero in on why I'm
> getting these bad results, and provide you some more useful feedback.

It certainly sounds like Emacs is getting the svn status of the file
wrong. This may or may not be due to a problem parsing the output of the
"svn status" external command. Eg it could be due to us not calling "svn
status"; http://debbugs.gnu.org/7850 comes to mind.

The most helpful thing is of course a recipe to reliably reproduce the problem.
Failing that, you might try edebugging vc-next-action to see why it gets
"state" wrong.

> Also, is there some way I can examine the copy of vc-svn.el that I
> have to make sure that you and I are talking about the same thing?

I'm using the one in the current Emacs trunk. You can use locate-library
or somesuch to find yours.

bug#14242: vc-next-action with registered file [VC-mode with subversion 1.7.8]

[Lars Ingebrigtsen]

Glenn Morris <rgm@gnu.org> writes:

> (Please keep the debbugs address cc'd)
>
> Robert Goldman wrote:
>
>> Thanks. Question: isn't the stuff from my trace of vc-svn wrong? I.e.,
>> do we agree that vc-svn is incorrectly seeing a file that's
>> unregistered? If that's the case, I can try to zero in on why I'm
>> getting these bad results, and provide you some more useful feedback.
>
> It certainly sounds like Emacs is getting the svn status of the file
> wrong. This may or may not be due to a problem parsing the output of the
> "svn status" external command. Eg it could be due to us not calling "svn
> status"; http://debbugs.gnu.org/7850 comes to mind.
>
> The most helpful thing is of course a recipe to reliably reproduce the problem.
> Failing that, you might try edebugging vc-next-action to see why it gets
> "state" wrong.

More information was requested, but no response was given within a few
months, so I'm closing this bug report.  If the problem still exists,
please reopen this bug report.

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