Steven R. Brandt
sbrandt at cct.lsu.edu
Wed Nov 7 13:01:41 CST 2018
Yes, that was it. I was missing an update. Thanks.
On 11/7/18 12:50 PM, Roland Haas wrote:
> Hello Steve,
> cctk_alignment is new in the slew of Carpet related changes that were
> just recently pushed. Those changes concerned Cactus, Carpet and PUGH.
> Can you make sure you are running the very newest version of PUGH?
> To answer your question: cctk_alignment should always be 1 for PUGH:
> commit 1df75fdbf6f80805f58a0237dd0bc277bae75207 (HEAD -> master, origin/master, origin/HEAD)
> Author: Erik Schnetter <schnetter at gmail.com>
> Date: Thu Feb 2 16:33:42 2017 -0500
> PUGH: Initialize new cGH fields “cctk_alignment” and “cctk_alignment_offset”
> diff --git a/PUGH/src/GHExtension.c b/PUGH/src/GHExtension.c
> index 9d3a641..e5b76ad 100644
> --- a/PUGH/src/GHExtension.c
> +++ b/PUGH/src/GHExtension.c
> @@ -210,6 +210,8 @@ static void PUGH_InitGHBasics (cGH *GH)
> GH->cctk_ubnd[dim] = GFExtras->ub[mypGH->myproc][dim];
> GH->cctk_ash[dim] = GH->cctk_lsh[dim];
> + GH->cctk_alignment = 1;
> + GH->cctk_alignment_offset = 0;
>> I'm debugging a modification to Cactus, and something I've done has
>> tripped an assertion error when running the RK2-central.par test. The
>> message is:
>> TestMoL_InitVars: Assertion `cctki0_ialn>0' failed.
>> The variable cctki0_ialn seems to get filled in from
>> cctkGH->cctk_alignment, which is (as far as I can see) only set in
>> mode.cc by carpet.
>> The test in question, however, uses PUGH.
>> Valgrind confirms that just before the assertion error, I am tripping a
>> conditional depends on uninitialized value error. However, I don't see
>> this same problem when running with the unmodified Cactus source--so
>> somehow I've broken something relating to this infrastructure.
>> How and where is cctk_alignment supposed to get set when running PUGH?
>> What should its value be? Thanks.
>> Users mailing list
>> Users at einsteintoolkit.org
More information about the Users