yade-dev team mailing list archive

[Chareyre <bruno.chareyre@xxxxxxxxxxxxxxx>]

numpy? I guess this is for comparing mathematical functions one by one, 
not for testing Yade.
There is sometimes this little detail that nobody suspect and which 
makes things go bad.

In order to really test yade, we would need to run some typical 
simulations and check that the results are the same. This is really hard 
to do, because we know the results _will not_ be the same actually. So, 
it would need to analyze statistics of results after repeated runs and 
see if the statistic is the same, or find something that is obviously 
incorrect from a physical point of view (the bug I just filled on 
instabilities is an example). Moreover, if we don't find the "good" 
simulation exhibiting the problem, it can remain hidden for some time.

For basic triaxial tests, I can tell results are at the same time 
realistic and in good agreement with previous results. Luc mentionned a 
possible problem in cohesive laws however, no idea if it is related to 
eigen yet.

I don't think we have manpower and time to do such hard testing before 
next release. So, we can drop wm3 completely now (lazy but risky), or 
keep both library support for some time (painfull but safe), so that if 
strange results are obtained we can always test wm3 build vs. eigen build.
Also if one guy needs last minute results before PhD defense, and find 
that his code behaves strangely when compiled with eigen, I'm sure he 
will be glad to switch to wm3...
Keeping wm3 until the next release should not be that hard, is it? We 
can perhaps quit typing wm3-friendly code after that?

-- 
r2247 broke compilation with wm3
https://bugs.launchpad.net/bugs/585141
You received this bug notification because you are a member of Yade
developers, which is subscribed to Yade.

Status in Yet Another Dynamic Engine: New

Bug description:
Please fix that if you want to continue using wm3 or say clearly that we can drop wm3 now (I am fine with that).

pkg/dem/DataClass/InteractionGeometry/Dem3DofGeom_SphereSphere.cpp: In member function 'virtual Vector3r Dem3DofGeom_SphereSphere::scaleDisplacementT(Real)':
pkg/dem/DataClass/InteractionGeometry/Dem3DofGeom_SphereSphere.cpp:89:11: error: no match for 'operator*' in '2 * diff'
pkg/dem/DataClass/InteractionGeometry/Dem3DofGeom_SphereSphere.cpp:90:1: warning: control reaches end of non-void function


pkg/dem/DataClass/InteractionGeometry/Dem3DofGeom_WallSphere.cpp: In member function 'virtual Vector3r Dem3DofGeom_WallSphere::scaleDisplacementT(Real)':
pkg/dem/DataClass/InteractionGeometry/Dem3DofGeom_WallSphere.cpp:46:11: error: no match for 'operator*' in '2 * diff'
pkg/dem/DataClass/InteractionGeometry/Dem3DofGeom_WallSphere.cpp:47:1: warning: control reaches end of non-void function