Re: Resend: Issue #1297: `svn diff' after `svn merge' that adds a file fails

kfogel@collab.net writes:

> > There is a comment in libsvn_wc/diff.c:file_diff
> >
> > /* If the item is schedule-add *with history*, then we don't want to
> > see a comparison to the empty file; we want the usual working
> > vs. text-base comparision. */
> > if (copied)
> > schedule = svn_wc_schedule_normal;
> >
> > I don't remember writing that bit, perhaps someone else put it in.
>
> Hmm, that still begs the question, then, because _some_ text base
> ought to be there, no?

Yeah, that's the core problem with this bug. The main question is:
"how did this file get into a schedule-add-with-history state?"

When that code was written (possibly by me, long ago??) the author was
*only* thinking about the 'svn cp wc wc' case. In that scenario, the
text-base *does* exist, I believe. But the other possibility is 'svn
cp URL wc', where there's no text-base. Whoops. This is the command
that 'svn merge' runs.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Jul 28 19:23:12 2003