Re: Issue #3119 - File '...' already exists" (Regression since 1.5.0)

Bert Huijben wrote:
> Hi,
>
> Restarting this topic in a new thread:
> (This topic original hijacked a few others)
>
> http://subversion.tigris.org/issues/show_bug.cgi?id=3119
>
> Reproduction script on:
> http://subversion.tigris.org/nonav/issues/showattachment.cgi/958/issue-3
> 119-1.sh
> == <snippet>==
> mkdir Data
> svn co ${URL} Data/wc
> mkdir Data/wc/tags
> mkdir Data/wc/trunk
> mkdir Data/wc/trunk/dir
> cp /dev/null Data/wc/trunk/dir/file.cpp
> ${SVN} add Data/wc/t*
> ${SVN} ci -m"test" Data/wc
>
> echo "ABC" >> Data/wc/trunk/dir/file.cpp
>
> ${SVN} ci -m"test" Data/wc
>
> # Using http:// only, the next command fails with this error:
> #
> # subversion/libsvn_client/copy.c:1319: (apr_err=175005)
> # svn: Commit failed (details follow):
> # subversion/libsvn_ra_neon/commit.c:1036: (apr_err=175005)
> # svn: File '/ms/repos/tags/V1.0.0/dir/file.cpp' already exists
> #
> ${SVN} copy -m"tagging" Data/wc/trunk ${URL}/tags/V1.0.0
> == </snippet> ==
>
> Karl performed a diagnosis in:
> http://svn.haxx.se/users/archive-2008-10/0829.shtml
>
> And I added some addition research in
> http://svn.haxx.se/dev/archive-2008-12/0309.shtml
>
> Other references to this issue include
> http://svn.haxx.se/users/archive-2008-11/0581.shtml
> and http://svn.haxx.se/users/archive-2008-11/0603.shtml
>
>
> It looks like r31692 introduced the regression described in issue #3119,
> when fixing issue #2939 by moving a check on whether a node exists from
> the HEAD url to the work url. (It was merged to 1.5.x before 1.5.1)
>
> http://subversion.tigris.org/issues/show_bug.cgi?id=2939
>
> "Basically, it's possible for tools like gvn and svnmucc to create
> revisions which *cannot* be svnsync'd over http -- unless we apply this
> fix."
>
>
> The serf fix in r31695 looks like a much better fix of this issue (as it
> really checks for the deleted case described in #2939), but this doesn't
> map to the neon way of doing things.
>
>
> Could somebody with more neon knowledge please look at this issue and
> help me determine whether it would be a wise decision to revert r31692
> to fix the normal client?
>
> It would reintroduce the issue that is only triggerable with gvn,
> svnmucc or another custom application, which should then be fixed in
> another way.

I've been looking at this a bit today, but haven't gotten very far. I think
this is related to the issue maven users have been having and described here:
http://jira.codehaus.org/browse/SCM-406

I'm attempting to reproduce using the jira bug description. It'd be good to fix
this, of course, and other eyes are welcomed to look at it.

Thanks,
-Hyrum

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=985149