[Dune-devel] parallel tests in dune-common

Dominic Kempf dominic.r.kempf at gmail.com
Mon Dec 1 12:44:00 CET 2014


Hey all,

d) would be so easy to implement: just return 0 instead of 77 in the
preprocessor else statement and be done. I personally don't deem it
important to know whether a test passed or whether it passed trivially,
because it was skipped.

The return-77-issue seems like a very good example of an autotoolism to me
and we shouldn't copy the old bahviour IMO.

Dominic

On Mon, Dec 1, 2014 at 11:59 AM, Carsten Gräser <graeser at mi.fu-berlin.de>
wrote:

> Hi Christoph,
>
> Am 01.12.2014 um 11:43 schrieb Christoph GrĂ¼ninger:
> > Hi Carsten,
> > I know. The reason it, they are not disabled by the build system but
> > return 77. This indicates a skipped test for the Autotools, in CMake
> > this was not supported until CMake 3.0.
> > Maybe we should decide in general, whether the build system should
> > disable the test silently or that it should indicate the skipped test. I
> > prefer to get "x test skipped" and would like to make the change. The
> > problem is, that it will only work properly with CMake 3.0.
> > Or we use a hack with a script that evaluates the return value and
> > changes a 77 into a 0. Mhh.
> given the following options
>
> a) Let the tests fail until we require cmake 3.0
> b) Silently disable the tests if cmake <3.0
> c) Provide a hack that disables pretends that the tests have passed.
> d) Make tests pass if their preconditions are not matched
>    (i.e. sequential dune)
>
> my ranking would be b) > c) > d) > a).
>
> Best,
> Carsten
>
>
>
> _______________________________________________
> Dune-devel mailing list
> Dune-devel at dune-project.org
> http://lists.dune-project.org/mailman/listinfo/dune-devel
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.dune-project.org/pipermail/dune-devel/attachments/20141201/c40ec198/attachment.htm>


More information about the Dune-devel mailing list