Trent Buck wondered on IRC tonight of "optimize --relink" was broken in Darcs2.
I observed that we don't have an automated test for "--relink", although there
is a "tests/optimize.pl" already. It would be nice to have one. 

[21:15] <twb> Is darcs optimize --relink broken in 2.0.0~pre2, or am I using it 
[21:17] <twb> du -s and du -sl report the same number, even after relinking

Trent just wrote a test for optimize --relink, which fails with Darcs2 and
1.0.9.  I'll be sending that patch to darcs-devel shortly. 

I'm unassigning the ticket now, since it seems to need the attention of a 
Haskell programmer now, not a test programmer.

On Mon, Jan 21, 2008 at 03:47:18AM -0000, Mark Stosberg wrote:
> Trent just wrote a test for optimize --relink, which fails with Darcs2 and
> 1.0.9.  I'll be sending that patch to darcs-devel shortly. 
> 
> I'm unassigning the ticket now, since it seems to need the attention of a 
> Haskell programmer now, not a test programmer.

I just tested with both 1.0.9 and darcs2, and both of them does
relink the files in _darcs/patches. Darcs2 even created a link
when I did a pull. I'm not sure I understand what the test tries
to do. What does Trent expect --relink to do?

On Mon, Jan 21, 2008 at 04:22:40AM -0000, Tommy Pettersson wrote:
> I'm not sure I understand what the test tries to do.

Ah, the second patch made it a little clearer.

There must be at least one patch in the repo for --relink to
have something to relink!

Thanks for the comment, Tommy. 

I'll turn it back into a "testing" issue and take it myself to get the test
updated to test with some actual patches.

On Mon, Jan 21, 2008 at 06:38:54PM -0000, Mark Stosberg wrote:
> I'll turn it back into a "testing" issue and take it myself to get the test
> updated to test with some actual patches.

I've already made a few changes, so don't work on this just yet.
-- 
David Roundy
Department of Physics
Oregon State University

Incidentally, this turns out to be a bug in the handling of the flag --sibling.
 If we cd into x and then run with --sibling=../y, the test passes.

David

I believe this is fixed by a patch I'm about to push to unstable.