[conda] compile calculix for osx and windows

[looo]

as spooles should be available for win/linux/osx it's time to test calculix build. But it is more diffficult than I expected.

If anyone sees something obviously wrong, please report:
https://github.com/conda-forge/staged-recipes/pull/6827

[looo]

Finally a first build for windows. Maybe someone wants to test this.

Code: Select all

conda install calculix -c freecad

https://anaconda.org/freecad/calculix/files

[looo]

tried it myself and it seems like there is missing a ".exe" But after renaming the file it seems to work.
In FreeCAD there are still come problems:

Code: Select all

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:\Users\hans_\conda\envs\freecad\Library\Mod\Fem\femguiobjects\_ViewProviderFemSolverCalculix.py", line 257, in calculixFinished
    majorVersion, minorVersion = fea.get_ccx_version()
  File "C:\Users\hans_\conda\envs\freecad\Library\Mod\Fem\femtools\ccxtools.py", line 696, in get_ccx_version
    m = re.search(r"(\d+).(\d+)", ccx_stdout)
  File "C:\Users\hans_\conda\envs\freecad\lib\re.py", line 182, in search
    return _compile(pattern, flags).search(string)
TypeError: cannot use a string pattern on a bytes-like object

[looo]

https://anaconda.org/conda-forge/calculix/files
Calculix is now uploaded to the conda-forge channel.

[Kunda1]

https://anaconda.org/conda-forge/calculix/files
Calculix is now uploaded to the conda-forge channel.

Sweet!
Until repology incorporates conda-forge, is there a list (of links) of dependencies that are on conda-forge that FreeCAD currently uses ?

[looo]

Sweet!
Until repology incorporates conda-forge, is there a list (of links) of dependencies that are on conda-forge that FreeCAD currently uses ?

Yes, I am trying to have this list updated: https://github.com/FreeCAD/FreeCAD_Cond ... /README.md

[UR_]

Maybe someone wants to test this

Just ran your build against guido's official test suite from here:

http://www.dhondt.de/ccx_2.14.test.tar.bz2

And this is the result log

error.ccx2.14looo.txt

(29.02 KiB) Downloaded 34 times

In comparison to ddfem's build

error.ccx2.14ddfem.txt

(8.27 KiB) Downloaded 33 times

As said, ccx is a pretty sensible beast
(especially in regard of code optimizations of fortran parts, perhaps worth a look)

[looo]

can you share the script for running this test? I guess it's not always necessary to have matching output. For example Eigenvectors can have different values for the same result using different solvers. But looking at file eg b31.inp there shouldn't be a difference in the result of a static computation...

[looo]

ok looking at b31.frd and comparing the results shows one obvious problem:

Code: Select all

b31.frd
...
 -1        12-1.25000E-0011.25000E-0010.00000E+000
 -1        13-1.25000E-001-1.25000E-0010.00000E+000
 -1        141.25000E-001-1.25000E-0010.00000E+000
 ...

Code: Select all

b31.frd.ref
...
 -1        12-1.25000E-01 1.25000E-01 0.00000E+00
 -1        13-1.25000E-01-1.25000E-01 0.00000E+00
 -1        14 1.25000E-01-1.25000E-01 0.00000E+00
 ...

So it seems we have a formatting problem here and directly comparing the results is not really possible.

edit: This problem is related to the number of exponential digits printed. Somehow the default value is 3, but calculix needs 2. If anyone knows how to fix this, please add a note here: https://github.com/conda-forge/calculix ... k/issues/3

[bernd]

we had similar problem already but I can not find it ...