Re: svn commit: rev 5944 - in trunk/subversion: libsvn_client libsvn_wc

Greg Stein <gstein@lyra.org> writes:

> On Thu, May 15, 2003 at 09:41:29AM -0500, cmpilato@tigris.org wrote:
> >...
> > +++ trunk/subversion/libsvn_client/commit_util.c Thu May 15 09:41:16 2003
> >...
> > + err = svn_wc_adm_retrieve (&parent_access, parent_dir,
> > + parent, subpool);
> > + if (err && err->apr_err == SVN_ERR_WC_NOT_LOCKED)
> > + {
> > + svn_error_clear (err);
> > + SVN_ERR (svn_wc_adm_open (&parent_access, NULL, parent,
> > + FALSE, FALSE, subpool));
> > + }
> > + else if (err)
> > + {
> > + return err;
> > + }
>
> Could you leverage the new svn_wc_adm_probe_try() here?

As it turns out, no. I tried to re-add the _try() there again just
now, and commit-test 25 quickly reminded me of the error it gave the
first time I tried this:

   lt-svn: subversion/libsvn_wc/lock.c:654: do_close: Assertion `!
   adm_access->set_owner || apr_hash_count (adm_access->set) == 0'
   failed.

It took me a bit to see the crucial difference between the code
portion above and the implementation of svn_wc_adm_probe_try(), but I
finally did:

> > + SVN_ERR (svn_wc_adm_open (&parent_access, NULL, parent,
> > + FALSE, FALSE, subpool));

Here, I had to pass NULL as the "associated" access baton, where
_try() would have just passed 'parent_dir'.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri May 23 01:22:42 2003