> -----Original Message-----
> From: Greg Stein [mailto:gst...@gmail.com]
> Sent: donderdag 24 maart 2011 18:37
> To: dev@subversion.apache.org
> Subject: Re: svn commit: r1084971 -
> /subversion/trunk/subversion/libsvn_wc/update_editor.c
>
> On Thu, Mar 24, 2011 at 13:33, Greg Stein <gst...@gmail.com> wrote:
> > On Thu, Mar 24, 2011 at 10:39, <rhuij...@apache.org> wrote:
> >>...
> >> +++ subversion/trunk/subversion/libsvn_wc/update_editor.c Thu Mar 24
> 14:39:55 2011
> >>...
> >> @@ -3615,10 +3609,10 @@ merge_file(svn_skel_t **work_items,
> >> *install_pristine = FALSE;
> >> *install_from = NULL;
> >>
> >> - if (new_text_base_sha1_checksum != NULL)
> >> + if (fb->new_text_base_sha1_checksum != NULL)
> >> SVN_ERR(svn_wc__db_pristine_get_path(&new_text_base_tmp_abs
> path,
> >> - eb->db, fb->local_abspath,
> >> - new_text_base_sha1_checksum,
> >> + eb->db, pb->local_abspath,
> >> +
fb->new_text_base_sha1_checksum,
> >> pool, scratch_pool));
> >
> > Euh... I don't think you meant to replace fb->local_abspath with pb->
> > ... did you? That sounds quite incorrect. IIRC, that would be asking
> > for the pristine path of the containing directory.
>
> Ah... never mind. That is the WRI_ABSPATH. The checksum is the key input.
Yes, this was intended.
And the parent directory is in the wc_db hashtable, while the files aren't.
So this could avoid a disk stat in the current implementation, while it
should still be safe in the future.
Bert
