Ben Collins-Sussman wrote:
> The test was XFail() until yesterday; that explains why we didn't see
> the failure over http until today.
>
> Still, though... I can't remember why I put the +1 there in the code.
> It looks extremely deliberate, as if I were certain that while both
> paths were absolute fs paths, one always had a leading '/' and the
> other didn't. So it surprises me that this fix is required, or that
> it makes the test pass.
>
> Commit it, and see what buildbot says over all 4 RA layers.
>
Ok, so we noticed that the patch makes c-o-u work over ra_serf/neon, but
now it fails over ra_local & ra_svn.
What confuses me is that on add_file_smartly o_path - which is the full
path of the target entry - starts with a '/' when running over
ra_[neon/serf], but without '/' over ra_local and ra_svn.
Function svn_fs_closest_copy is perfectly capable to handle both
absolute and relative paths, as it converts any path to its absolute and
canonical equivalent first.
One way to fix this nicely would be to canonicalize our o_path before
comparing strings. This only requires moving
svn_fs__canonicalize_abspath to libsvn_subr/svn_path.c so we can access
it in the repos layer.
Thoughts?
Lieven
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 6 22:51:19 2007