On 16 Jun 2015, at 10:20, Bitbucket <[email protected]> wrote:

> 1 new commit in Cactus:
> 
> https://bitbucket.org/cactuscode/cactus/commits/3ae5f993e10c/
> Changeset:   3ae5f993e10c
> Branch:      master
> User:        Roland Haas
> Date:        2015-06-16 08:15:33+00:00
> Summary:     Cactus: forbid self-referential orderings, do not add cyclic deps
> 
> * use C99 features in CCTKi_ScheduleAddRow
> 
> Affected #:  1 file
> 
> Repository URL: https://bitbucket.org/cactuscode/cactus/

Hi Roland, 

This looks like it causes a number of test failures, e.g.

https://build.barrywardell.net/job/EinsteinToolkit/527/testReport/junit/(root)/TestGlobalReduce/TestAvg_1procs/

The problem seems to be in CactusTest/TestGlobalReduce/schedule.ccl:

> schedule TestGlobalReduce_Initial at CCTK_INITIAL
> {
>   LANG:C
> } "Initialize the input arrays to values"
> 
> schedule TestGlobalReduceC_Avg at CCTK_POSTINITIAL after TestGlobalReduceC_Avg
> {
>   LANG: C
> } "Test out GF reduction"
> 
> schedule TestGlobalReduceC_Maximum at CCTK_POSTINITIAL after 
> TestGlobalReduceC_Avg
> {
>   LANG: C
> } "Test out maximum reduction"

and there is indeed a circular dependency.  This is probably a copy-paste 
error.  My initial reaction would be to remove the "after 
TestGlobalReduceC_Avg" in the second one, since POSTINITIAL is after INITIAL 
anyway.  

Is this the correct fix?

Can this sort of thing be checked at compile time in addition to runtime?

-- 
Ian Hinder
http://members.aei.mpg.de/ianhin

_______________________________________________
Users mailing list
[email protected]
http://cactuscode.org/mailman/listinfo/users

Reply via email to