[Dune] [Dune-Commit] dune-grid r6593 - releases/2.0/dune/grid/albertagrid/test

Jö Fahlke jorrit at jorrit.de
Fri Apr 9 11:30:23 CEST 2010


Am Fri,  9. Apr 2010, 11:04:47 +0200 schrieb Christian Engwer:
> But what does then the following output mean?
> 
>       std::cout << "The following stuff is expected to fail currently, because the recursive" << std::endl
>                 << "bisection algorithm of alberta will run into an endless recursion for" << std::endl
>                 << "certain grids.  See Flyspry#569." << std::endl;
> 
> Actually it is not expected to fail, but to produce a segfault?! In
> that case the test is truely failing, because the bug is fixed.

Well, a segfault is kind of a failure (argh, too many "not-nots").  And Martin
did fiddle with Alberta to make the segfault go away, at least for the grid
used in the test.  As I understand it, there are (and always will be) certain
grids for which the test would segfault.  This is inherent in the recursive
bisection algorithm.

> So either we remove the test, or fix it.
> 
> I'm awaiting patches ;-)

I'm going to fix the test such that passing becomes the expected behaviour,
and of course I'll improve the message above as well.  This way we will notice
when someone fiddles with Albertas heuristic again in such a way that the grid
tested here does no longer work.

Just give me some time to set up a dune-2.0 environment, so I can actually
check that I don't introduce some stupid error.  You will get the patch today,
promise.

Jö.

-- 
Entstauben ist Mord!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 828 bytes
Desc: Digital signature
URL: <https://lists.dune-project.org/pipermail/dune/attachments/20100409/9b8fce2d/attachment.pgp>


More information about the Dune mailing list