New thorn CactusTest/TestCrayPointers

Comments (10)

1. 

   Erik Schnetter reporter

   • changed status to open
   • removed comment

   • 2012-10-25T14:26:23+00:00
2. 

   Ian Hinder

   • removed comment

   I think we have too many test thorns. Can you create a single thorn which contains a larger number of tests?

   • 2012-10-25T16:46:44+00:00
3. 

   Roland Haas

   • removed comment

   Since we plan to use this feature in the flesh eventually, we might want to not add such a test right now but instead rely on autoconf to detect this for us once we add that feature. If autoconf fails then we must fall-back to the current scheme of calling Fortran thorns. We then would need no specific test case since all Cactus thorns will be tests I think.

   • 2012-10-27T19:27:58+00:00
4. 

   Erik Schnetter reporter

   • removed comment

   Regarding "too many thorns": Please suggest another thorn into which to fold this test. All other thorns relate to certain features, and none subsume Cray pointers. We could conceivably merge test thorns, but I suggest to not make this a prerequisite for testing Cray pointers, as these are independent issues.

   Regarding autoconf: Autoconf support for Fortran is meagre. One can use Fortran 77 only, and this only because Cactus extends certain autoconf macros that otherwise support only C or C++. (We are using a very old version of autoconf.) I just checked, and running a Cray pointer test via autoconf is not trivial, as Fortran compiler flags seem to get lost. Testing the C/Fortran calling interface is also not possible from autoconf, since this is determined later (and basically outside autoconf).

   I believe this thorn has value as-is, if only to allow us get feedback about systems where Cray pointers don't work before we use them.

   • 2012-10-28T21:42:17+00:00
5. 

   Ian Hinder

   • removed comment

   Why not have a single thorn which contains all these tests? I think these thorns logically test features of the flesh, which is often thought of as a thorn called "Cactus". Why not just have a thorn called CactusTest in the CactusBase arrangement which contains all these tests? Then we wouldn't have to keep creating repositories for every new class of test for the flesh. Thoughts?

   • 2012-10-29T10:34:43+00:00
6. 

   Roland Haas

   • removed comment

   Fewer test thorns would make sense. A number of these tests in this arrangement test compile time errors as well as of run time errors (eg. the Cray pointer one would do so, and the testloop thorns also test compile time errors).

   • 2012-10-29T10:46:30+00:00
7. 

   Bruno Mundim

   • removed comment

   While I agree that grouping several tests in one thorn may be beneficial, I think the discussion in this line goes beyond the scope of this ticket. I would add this proposed thorn to the ET thornlist and open a new ticket where we could put our thoughts together on what needs to be done to group all the tests in one single thorn, if we really decide to do so.

   • 2012-10-31T22:41:50+00:00
8. 

   anonymous

   • removed comment

   I agree.

   • 2012-10-31T23:25:16+00:00
9. 

   Erik Schnetter reporter

   • changed status to resolved
   • removed comment

   Applied.

   • 2013-03-08T14:25:39+00:00
10. 

    Roland Haas

    • edited description
    • changed status to closed

    • 2019-02-21T20:09:36+00:00
11. Log in to comment