[Carpet] [Carpet-darcs] Carpet: Correct regridding error during recovery

[Thomas Radke]

Thomas Radke wrote:
> I think this patch broke the scheduling somehow. A parfile with 
> 'cactus::cctk_itlast = 19' is running for 20 iterations now.

Sorry, I take that back. It was a mistake in my parfile testing 
recovery: after turning off IOHDF5::use_grid_structure_from_checkpoint I 
was missing all finer reflevels. The reason for that was that I still 
had set Carpet::regrid_during_recovery = "no". Because 
Carpet::max_refinement_levels was still 2, the coarse level was evolved 
at even iterations, and the simulation stopped at iteration 20, not at 
19 as it would have with 2 reflevels.

See my next email about problems with setting up the weight function 
after recovery with IOHDF5::use_grid_structure_from_checkpoint turned on.

> I also noticed in the Cactus user portal that since today 3 fewer 
> testsuites are passing during the nightly integration tests on Peyote. I 
> haven't checked which ones, probably from Carpet.

This is still true though.

-- 
Cheers, Thomas.