yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

> For me, clang is the primary compiler since a few months (it is SOO 
> much faster; I only use g++ when I want to compile with openmp), I 
> consider it a supported compiled. Therefore I would like you to 
> install clang (lucid package in yade-external) and fix your test; I 
> would do it, but honestly I don't know what it tests. We can just put 
> the tolerance up a bit, so that it passes with clang.
Increasing tolerance is not a good idea, since 10e-3 is already high. 
Thanks to your numbers, I could see the problem : it was comparing a 
residual velocity in a situation close to static equilibrium. In other 
words : "is numerical noise the same?". Now we know that numerical noise 
is different with different compilers. Lesson learnt.

I disabled this test in last commit, and included more explanations of 
what happens in the script:

/"Motion of a "sinusoidal" beam made of cylinders. The test checks the 
position and velocity of the free end of the bending beam subjected to 
gravitational load. It is similar to 
scripts/test/chained-cylinder-spring.py but with less elements. 
positions and velocity are compared during the transient oscillations, 
only positions are compared for the larger time since residual velocity 
is compiler-dependent (see 
https://lists.launchpad.net/yade-dev/msg06178.html)."
/
I won't spend time on clang++ now, sorry. I'm strugling with enough 
taucs/pardiso/fortran compilation issue with gcc at the moment. With 
clang I would be left with no support at all from library developers.
Just curious : did you compile with clang + CGAL feature yet?
> > p.s. What is pristine, sorry?
> Virginal, d'origine... (dict?)

Oh, it was just english! Thanks. :-)

Bruno