[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: svn commit: r1339892 - /subversion/trunk/subversion/libsvn_fs/editor.c

From: Hyrum K Wright <hyrum.wright_at_wandisco.com>
Date: Thu, 17 May 2012 21:38:58 -0500

Thanks for taking a look. It feels like all of these are issues with
the complex copies and inheriting the correct replacement revision.
The problem then lies back in the commit driver, but I don't think it
will be too hard to track down.

The blame test failure is an unrelated EOL repairing issue.

-Hyrum

On Thu, May 17, 2012 at 7:48 PM, Greg Stein <gstein_at_gmail.com> wrote:
> Notes:
>
> copy 29: a mixed-rev child of a copied parent that is re-copied to
> pick up a different source revision needs to specify that parent
> revision for REPLACES_REV. Since it passes SVN_INVALID_REVNUM, the
> existence check comes into play, and the node exists (implicitly as a
> result of copying the parent).
>
> merge 18: same. I_at_2 and I/theta_at_6 are copied to L. It copies I_at_2, then
> it (re)copies I_at_6 without specifying REPLACES_REV=2.
>
> switch 33: a little bit different, but copied children again. In this
> case, R/G is switched to a different subtree than the R/G_at_1 subtree.
> So the commit process copies R, and then it must replace R/G with a
> different subtree. Again: it needs to specify REPLACES_REV.
>
> Hope that helps,
> -g
>
> On Thu, May 17, 2012 at 7:32 PM, Greg Stein <gstein_at_gmail.com> wrote:
>> Hyrum,
>>
>> This fixes the two tree_conflict tests you mentioned, and blame 7
>> continues to fail.
>>
>> However, it introduced failures in copy 29, merge 18, and switch 33.
>> Each one has the error that I just added. I haven't investigated yet:
>> it could be that a different error was expected.
>>
>> (all this on the ev2-export branch, of course, and on ra_local)
>>
>> Cheers,
>> -g
>>
>> On Thu, May 17, 2012 at 6:27 PM,  <gstein_at_apache.org> wrote:
>>> Author: gstein
>>> Date: Thu May 17 22:27:03 2012
>>> New Revision: 1339892
>>>
>>> URL: http://svn.apache.org/viewvc?rev=1339892&view=rev
>>> Log:
>>> Generate out-of-date errors for paths which already exist (rather than
>>> waiting for an error when we try to write over existing content).
>>>
>>> * subversion/libsvn_fs/editor.c:
>>>  (can_create): new helper function
>>>  (add_directory_cb, add_file_cb, add_symlink_cb, copy_cb, move_cb):
>>>    call the new can_create() helper
>>>
>>> Modified:
>>>    subversion/trunk/subversion/libsvn_fs/editor.c
>>>
>>> Modified: subversion/trunk/subversion/libsvn_fs/editor.c
>>> URL: http://svn.apache.org/viewvc/subversion/trunk/subversion/libsvn_fs/editor.c?rev=1339892&r1=1339891&r2=1339892&view=diff
>>> ==============================================================================
>>> --- subversion/trunk/subversion/libsvn_fs/editor.c (original)
>>> +++ subversion/trunk/subversion/libsvn_fs/editor.c Thu May 17 22:27:03 2012
>>> @@ -268,6 +268,25 @@ can_modify(svn_fs_root_t *txn_root,
>>>  }
>>>
>>>
>>> +/* Can we create a node at FSPATH in TXN_ROOT? If something already exists
>>> +   at that path, then the client is out of date.  */
>>> +static svn_error_t *
>>> +can_create(svn_fs_root_t *txn_root,
>>> +           const char *fspath,
>>> +           apr_pool_t *scratch_pool)
>>> +{
>>> +  svn_node_kind_t kind;
>>> +
>>> +  SVN_ERR(svn_fs_check_path(&kind, txn_root, fspath, scratch_pool));
>>> +  if (kind != svn_node_none)
>>> +    return svn_error_createf(SVN_ERR_FS_OUT_OF_DATE, NULL,
>>> +                             _("'%s' already exists, so may be out"
>>> +                               " of date; try updating"),
>>> +                             fspath);
>>> +  return SVN_NO_ERROR;
>>> +}
>>> +
>>> +
>>>  /* This implements svn_editor_cb_add_directory_t */
>>>  static svn_error_t *
>>>  add_directory_cb(void *baton,
>>> @@ -291,7 +310,10 @@ add_directory_cb(void *baton,
>>>       SVN_ERR(can_modify(root, fspath, replaces_rev, scratch_pool));
>>>       SVN_ERR(svn_fs_delete(root, fspath, scratch_pool));
>>>     }
>>> -  /* else better not be there!  */
>>> +  else
>>> +    {
>>> +      SVN_ERR(can_create(root, fspath, scratch_pool));
>>> +    }
>>>
>>>   SVN_ERR(svn_fs_make_dir(root, fspath, scratch_pool));
>>>   SVN_ERR(add_new_props(root, fspath, props, scratch_pool));
>>> @@ -321,7 +343,10 @@ add_file_cb(void *baton,
>>>       SVN_ERR(can_modify(root, fspath, replaces_rev, scratch_pool));
>>>       SVN_ERR(svn_fs_delete(root, fspath, scratch_pool));
>>>     }
>>> -  /* else better not be there!  */
>>> +  else
>>> +    {
>>> +      SVN_ERR(can_create(root, fspath, scratch_pool));
>>> +    }
>>>
>>>   SVN_ERR(svn_fs_make_file(root, fspath, scratch_pool));
>>>
>>> @@ -353,7 +378,10 @@ add_symlink_cb(void *baton,
>>>       SVN_ERR(can_modify(root, fspath, replaces_rev, scratch_pool));
>>>       SVN_ERR(svn_fs_delete(root, fspath, scratch_pool));
>>>     }
>>> -  /* else better not be there!  */
>>> +  else
>>> +    {
>>> +      SVN_ERR(can_create(root, fspath, scratch_pool));
>>> +    }
>>>
>>>   /* ### we probably need to construct a file with specific contents
>>>      ### (until the FS grows some symlink APIs)  */
>>> @@ -506,6 +534,10 @@ copy_cb(void *baton,
>>>       SVN_ERR(can_modify(root, dst_fspath, replaces_rev, scratch_pool));
>>>       SVN_ERR(svn_fs_delete(root, dst_fspath, scratch_pool));
>>>     }
>>> +  else
>>> +    {
>>> +      SVN_ERR(can_create(root, dst_fspath, scratch_pool));
>>> +    }
>>>
>>>   SVN_ERR(svn_fs_revision_root(&src_root, svn_fs_root_fs(root), src_revision,
>>>                                scratch_pool));
>>> @@ -541,6 +573,10 @@ move_cb(void *baton,
>>>       SVN_ERR(can_modify(root, dst_fspath, replaces_rev, scratch_pool));
>>>       SVN_ERR(svn_fs_delete(root, dst_fspath, scratch_pool));
>>>     }
>>> +  else
>>> +    {
>>> +      SVN_ERR(can_create(root, dst_fspath, scratch_pool));
>>> +    }
>>>
>>>   /* ### would be nice to have svn_fs_move()  */
>>>
>>>
>>>

-- 
uberSVN: Apache Subversion Made Easy
http://www.uberSVN.com/
Received on 2012-05-18 04:39:37 CEST

This is an archived mail posted to the Subversion Dev mailing list.