You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@subversion.apache.org by Ryan Schmidt <su...@ryandesign.com> on 2009/09/16 00:55:13 UTC

Re: annoying "Attempt to add tree conflict that already exists"

On Sep 15, 2009, at 19:28, Paul Hammant wrote:

> Nope, fails as well in the same way.
>
> ldd still reports it as "Not an dynamic executable"

Out of curiosity, what is it then, if not a dynamic executable? What  
does "file" say about it?

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=2395311

To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].

Re: annoying "Attempt to add tree conflict that already exists"

Posted by Paul Hammant <pa...@hammant.org>.
Stefan,

> Is that really all you are seeing?
> There should be messages like "local delete, incoming add upon merge"
> in the output of 'svn status' which tell you what kind of tree  
> conflict
> has happened. This is important important information to figure out
> what's going on.
>
> Can you copy/paste the output or redirect it to a file and send that
> file as an attachment or paste it into your mail?
>

The result of the merge (sys-out, last few lines) ...

     UU   foo/one
      G   foo/two
     --- Merging r10889 through r11137 into '.':
     GU   foo/three
        C foo/four

The sys-err from the same merge command ...

     svn: Attempt to add tree conflict that already exists at 'foo/nine'
     svn: Error reading spooled REPORT request response

The results of svn st ....

     [ many other lines ]
     MM      foo/seven
     A  +    foo/eight
     !     C foo/nine
           >   local missing, incoming edit upon merge
     !     C foo/ten
           >   local missing, incoming edit upon merge
     !     C foo/eleven
           >   local missing, incoming edit upon merge
     !     C foo/twelve
           >   local missing, incoming edit upon merge
     !     C foo/thirteen
           >   local missing, incoming edit upon merge
     !     C foo/fourteen
           >   local missing, incoming edit upon merge
     M       foo/fifteen
     M       foo/sixteen
     [ many other lines ]

... so yes you're right my grepping had omitted the lines beginning  
with ">"


>> What I'd rather see, is :
>>
>>> svn merge -accept theirs-full https://svn/repo/trunk/web .
>>       [ many successful ]
>>       G    foo/Bar
>>       svn: Attempt to add tree conflict that already exists at apple/
>> banana.xml, apple/grape.xml, apple/orange.xml, apple/kiwi.xml
>
> That's not possible, the message is generated for one item at a time.
> But yes, it could/should print more of those, possibly.
>
> What I'd like to know now is how exactly do you end up in that state?
> I want to know exactly what I need to do to end up in the same state
> as you, given an empty repository.

We're contemplating the larger effort of trying to make a reproduction.

- Paul

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=2395695

To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].

Re: annoying "Attempt to add tree conflict that already exists"

Posted by Stefan Sperling <st...@elego.de>.
On Wed, Sep 16, 2009 at 11:50:11AM -0500, Paul Hammant wrote:
> >> When trying to 'resolved' that one, and optimistically resume the
> >> merge, we're discovering that svn had made multiple items marked as
> >> "!     C", not just the first one.
> >
> > Can you please provide more of the actual output you are seeing?
> 
>        > svn merge -accept theirs-full https://svn/repo/trunk/web .
>        [ many successful ]
>        G    foo/Bar
>        svn: Attempt to add tree conflict that already exists at apple/ 
> banana.xml
>        svn: Error reading spooled REPORT request response
> 
>         > svt st
> 
>        !    C   apple/banana.xml
>        !    C   apple/grape.xml
>        !    C   apple/orange.xml
>        !    C   apple/kiwi.xml
> 
> That's more or less what we're seeing.

Is that really all you are seeing?
There should be messages like "local delete, incoming add upon merge"
in the output of 'svn status' which tell you what kind of tree conflict
has happened. This is important important information to figure out
what's going on.

Can you copy/paste the output or redirect it to a file and send that
file as an attachment or paste it into your mail?

> What I'd rather see, is :
> 
>        > svn merge -accept theirs-full https://svn/repo/trunk/web .
>        [ many successful ]
>        G    foo/Bar
>        svn: Attempt to add tree conflict that already exists at apple/ 
> banana.xml, apple/grape.xml, apple/orange.xml, apple/kiwi.xml

That's not possible, the message is generated for one item at a time.
But yes, it could/should print more of those, possibly.

What I'd like to know now is how exactly do you end up in that state?
I want to know exactly what I need to do to end up in the same state
as you, given an empty repository.

Thanks,
Stefan

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=2395658

To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].

Re: annoying "Attempt to add tree conflict that already exists"

Posted by Paul Hammant <pa...@hammant.org>.
Stefan,

>> As suggested earlier, svn-1.6.x-latest  is better at communicating
>> which file is the actual tree conflict. Like so
>>
>>      G    foo/Bar
>>       svn: Attempt to add tree conflict that already exists at apple/
>> banana.xml
>>       svn: Error reading spooled REPORT request response
>>
>> When trying to 'resolved' that one, and optimistically resume the
>> merge, we're discovering that svn had made multiple items marked as
>> "!     C", not just the first one.
>
> Can you please provide more of the actual output you are seeing?

       > svn merge -accept theirs-full https://svn/repo/trunk/web .
       [ many successful ]
       G    foo/Bar
       svn: Attempt to add tree conflict that already exists at apple/ 
banana.xml
       svn: Error reading spooled REPORT request response

        > svt st

       !    C   apple/banana.xml
       !    C   apple/grape.xml
       !    C   apple/orange.xml
       !    C   apple/kiwi.xml

That's more or less what we're seeing.

What I'd rather see, is :

       > svn merge -accept theirs-full https://svn/repo/trunk/web .
       [ many successful ]
       G    foo/Bar
       svn: Attempt to add tree conflict that already exists at apple/ 
banana.xml, apple/grape.xml, apple/orange.xml, apple/kiwi.xml
       svn: Error reading spooled REPORT request response

>
>> Could svn be changed to report all of the items that are tree
>> conflicts, not just the first one (or not just a single one) ?
>
> It should report all of them, during update, merge, and status.
>
> Have you run 'svn status' on the working copy?

Yes, results as about. I can triple confirm that there were zero  
outstanding commits before beginning the merge; the working copy was  
pristine.

> It should list all tree conflicts.

It does list all the tree conflicts, by the message issues by the  
aborting svn-merge does not.

> If you want help with analysing
> the conflicts which are happening, please post the entire output
> of merges you are running, with the output of 'svn status -v' before
> and after any merges.

Actually, all of this is symptomatic of my root issue : subversion  
1.6.6(dev) has issues with three conflicts that I would hope it could  
sail past.

>
>> I appreciate that resuming a merge is somewhat non-standard too, but
>> we have no choice here if we're trying to get past this issue.
>
> Note that any existing tree-conflict victim will be skipped by a
> subsequent merge. svn assumes it's not safe to do anything to them.
> You should probably resolve all tree conflicts before running  
> another merge.

We witnessed that the only way to get a merge to carry on with such  
tree conflicts (marked with "!     C" in a svn st) was to mark them as  
resolved.

>
>> What we'd really love is the ability to respond via std-in with a "P"
>> for pend, in respect of tree conflicts, and deal with them after the
>> merge invocation.
>
> http://subversion.tigris.org/tasks.html#tree-conflicts-ui

Thanks.

- Paul

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=2395642

To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].

Re: annoying "Attempt to add tree conflict that already exists"

Posted by Stefan Sperling <st...@elego.de>.
On Wed, Sep 16, 2009 at 09:52:15AM -0500, Paul Hammant wrote:
> > What we've learned though is that there are tree conflicts all over
> > the place.  Thus, we're going to try some other strategy & report  
> > back.
> 
> As suggested earlier, svn-1.6.x-latest  is better at communicating  
> which file is the actual tree conflict. Like so
> 
>       G    foo/Bar
>        svn: Attempt to add tree conflict that already exists at apple/ 
> banana.xml
>        svn: Error reading spooled REPORT request response
> 
> When trying to 'resolved' that one, and optimistically resume the  
> merge, we're discovering that svn had made multiple items marked as  
> "!     C", not just the first one.

Can you please provide more of the actual output you are seeing?

> Could svn be changed to report all of the items that are tree  
> conflicts, not just the first one (or not just a single one) ?

It should report all of them, during update, merge, and status.

Have you run 'svn status' on the working copy?
It should list all tree conflicts. If you want help with analysing
the conflicts which are happening, please post the entire output
of merges you are running, with the output of 'svn status -v' before
and after any merges.

> I appreciate that resuming a merge is somewhat non-standard too, but  
> we have no choice here if we're trying to get past this issue.

Note that any existing tree-conflict victim will be skipped by a
subsequent merge. svn assumes it's not safe to do anything to them.
You should probably resolve all tree conflicts before running another merge.

> What we'd really love is the ability to respond via std-in with a "P"  
> for pend, in respect of tree conflicts, and deal with them after the  
> merge invocation.

http://subversion.tigris.org/tasks.html#tree-conflicts-ui

Stefan

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=2395607

To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].

Re: annoying "Attempt to add tree conflict that already exists"

Posted by Paul Hammant <pa...@hammant.org>.
> What we've learned though is that there are tree conflicts all over
> the place.  Thus, we're going to try some other strategy & report  
> back.

As suggested earlier, svn-1.6.x-latest  is better at communicating  
which file is the actual tree conflict. Like so

      G    foo/Bar
       svn: Attempt to add tree conflict that already exists at apple/ 
banana.xml
       svn: Error reading spooled REPORT request response

When trying to 'resolved' that one, and optimistically resume the  
merge, we're discovering that svn had made multiple items marked as  
"!     C", not just the first one.

Could svn be changed to report all of the items that are tree  
conflicts, not just the first one (or not just a single one) ?

I appreciate that resuming a merge is somewhat non-standard too, but  
we have no choice here if we're trying to get past this issue.

What we'd really love is the ability to respond via std-in with a "P"  
for pend, in respect of tree conflicts, and deal with them after the  
merge invocation.

Thanks,

- Paul

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=2395597

To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].

Re: annoying "Attempt to add tree conflict that already exists"

Posted by Paul Hammant <pa...@hammant.org>.
> Great work folks, we might be able to make progress. Will report  
> back later.

OK, we really like the 1.6.x-latest build, its much more  
communicative.  However its exposing multiple places where tree  
conflicts are happening.  We were hoping we could do ..

    cd branch
    cd web
    svn merge -accept theirs-full http://svn/repo/trunk/web .
    cd ../common
    svn merge -accept theirs-full http://svn/repo/trunk/common .
    cd ../data
    svn merge -accept theirs-full http://svn/repo/trunk/data .
    cd ..

.. given we knew that the relatively unimportant file services/foo.xml  
was the initial tree conflict.

What we've learned though is that there are tree conflicts all over  
the place.  Thus, we're going to try some other strategy & report back.

- Paul

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=2395589

To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].

Re: annoying "Attempt to add tree conflict that already exists"

Posted by Paul Hammant <pa...@hammant.org>.
> Paul, can you provide a script which reproduces the problem for
> you with 1.6.5 + patch, starting from an empty repository and
> leading up to the point where the double-tc error is produced?
> That would really, really, help! If it turns out that our fix is
> incomplete, we may have to revise it again. The earlier we can
> reproduce the failure the higher the chances of a proper fix ending
> up in 1.6.6.
>

Maybe.  We've made progress.  Whereas we previously though foo/Bar was  
the 'area' where merge conflicts were happening:

      G    foo/Bar
       svn: Attempt to add tree conflict that already exists
       svn: Error reading spooled REPORT request response

We see that 1.6.x@39362 is reporting more information - YAY  !

      G    foo/Bar
       svn: Attempt to add tree conflict that already exists at apple/ 
banana.xml
       svn: Error reading spooled REPORT request response

Great work folks, we might be able to make progress. Will report back  
later.

- Paul

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=2395579

To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].

Re: annoying "Attempt to add tree conflict that already exists"

Posted by Stefan Sperling <st...@elego.de>.
On Tue, Sep 15, 2009 at 07:55:13PM -0500, Ryan Schmidt wrote:
> On Sep 15, 2009, at 19:28, Paul Hammant wrote:
> 
> > Nope, fails as well in the same way.
> >
> > ldd still reports it as "Not an dynamic executable"
> 
> Out of curiosity, what is it then, if not a dynamic executable? What  
> does "file" say about it?

It should be a shell script wrapper generated by libtool.
This wrapper sets up LD_LIBRARY_PATH correctly, so if Paul was
running the wrapper script it's unlikely that there was a problem
with shared libraries.

Paul, can you provide a script which reproduces the problem for
you with 1.6.5 + patch, starting from an empty repository and
leading up to the point where the double-tc error is produced?
That would really, really, help! If it turns out that our fix is
incomplete, we may have to revise it again. The earlier we can
reproduce the failure the higher the chances of a proper fix ending
up in 1.6.6.

Thank you,
Stefan

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=2395422

To unsubscribe from this discussion, e-mail: [users-unsubscribe@subversion.tigris.org].