> -----Original Message-----
> From: Bert Huijben [mailto:b...@qqmail.nl]
> Sent: zaterdag 15 maart 2014 10:03
> To: 'Stefan Sperling'; 'Danilo Piazzalunga'
> Cc: users@subversion.apache.org
> Subject: RE: Bug report: svn assert failure: svn:
> subversion/libsvn_client/merge.c:3128: merge_dir_deleted: Assertion `delb
> != ((void *)0)' failed.
>
>
>
> > -----Original Message-----
> > From: Stefan Sperling [mailto:s...@elego.de]
> > Sent: zaterdag 15 maart 2014 09:45
> > To: Danilo Piazzalunga
> > Cc: users@subversion.apache.org
> > Subject: Re: Bug report: svn assert failure: svn:
> > subversion/libsvn_client/merge.c:3128: merge_dir_deleted: Assertion
> `delb
> > != ((void *)0)' failed.
> >
> > On Fri, Mar 14, 2014 at 10:53:05PM +0100, Danilo Piazzalunga wrote:
> > > Hi all,
> > > I am asking for confirmation before reporting a bug
> > >
> > > Using Subversion 1.8.8 installed from Ubuntu 14.04 packages, svn
> > > crashed while performing a merge operation involving a replaced
> > > directory.
> > >
> > > I already reportted this bug to Ubuntu Launchpad as
> > > https://bugs.launchpad.net/ubuntu/+source/subversion/+bug/1288829
> > >
> > > I was able to reproduce the bug with Subversion 1.9.0-dev trunk
> > > (r1577730) with a self-contained test case (see attachment).
> > >
> > > Can I report this to the Issue Tracker?
>
> It would be nice if the thread (that should be referred from the issue)
> actually discussed the issue.
>
> The only thing discussed here now is that there is some way to trigger the
> assertion. Not how it can be triggered, or how common the problem is.
>
> And if it is perhaps the result three steps after a user error or
> something...
>
> I really hope the issue will be more informative than the simple text '
svn
> assert failure: svn: subversion/libsvn_client/merge.c:3128:
> merge_dir_deleted: Assertion `delb != ((void *)0)' failed.' as our
changelog
> would be 'very user friendly' if we would describe all issues like this.
The really interesting thing here is the 'svn merge --force'
Without the '--force' the merge just succeeds without a problem.
That would be really useful information to add to an issue...
When I tried to create a reproduction recipe as part of the testsuite that
was something I missed...
With Subversion 1.8 I would really recommend not using this argument. Before
1.8 the merge infrastructure couldn't really test if a directory delete was
exactly valid, while with 1.8 we resolved that problem.
We just don't have many tests that use this '--force' and that is what
triggers the segfault for you.
I would assume a bugfix will be in the next patch release as this problem is
most likely easy to fix. (I expect that I have a patch in the next half
hour)
Bert
