[conda] compile calculix for osx and windows

[looo]

Has anybody successfully compiled calculix for osx? I am wondering if it makes sense to try this [1]. Also I am trying to compile calculix for windows with the mingw-toolchain. I think @easyw was successful on windows. Maybe we can transfer the instruction into a build-instruction for conda.

I guess in theory it should be possible to support all 3 major platforms, but some help would be nice! [2]

[1] https://travis-ci.org/conda-forge/stage ... tification
[2] https://github.com/conda-forge/staged-recipes/pull/6827

[looo]

For windows it looks like it is a bit more difficult. spooles and arpack are not yet available:

https://github.com/conda-forge/spooles- ... a.yaml#L17
https://github.com/conda-forge/arpack-feedstock/pull/22

So I guess this will need some time until we can compile calculix with flang and clang...

[bernd]

have you found these:

https://forum.freecadweb.org/viewtopic. ... 7&start=10
https://forum.freecadweb.org/viewtopic. ... 9&start=80

[peterl94]

I've always just used the homebrew package built using this formula: https://github.com/brewsci/homebrew-sci ... lix-ccx.rb

[looo]

Thanks for the instructions. I hope osx won't need too much iterations to get the builds succeed. For Windows arpack is now available build with the mingw-toolchain. So I will try to use the same toolchain for spooles and calculuix too.

[UR_]

ccx is a really sensible piece of software, showing it's weak spots only after doing tons of tests.

So let's speak this guy ("kwip" from mecway forum):

I did some research finding other projects having issues with ARPack which is kind of EOL. ocatve, scilab were forking their own implementation. There is some effort of consolidation, namely arpack-ng. I'll try to build CCX against this arpack version. Maybe we are lucky. I'll report back when I've put the required dependencies together.

http://mecway.com/forum/discussion/comm ... mment_2511


Results can be found here:
https://github.com/ddfem/ccx_mingw64/commits/master

Especially if it comes to non static calculations, this build is really appreciated.

[looo]

There is some effort of consolidation, namely arpack-ng.

actually the arpack-package on conda-forge uses sources from arpack-ng [1]. There is also another arpack (arpackpp [2]). Do you know for which part ccx uses arpack and for which it uses spooles?

[1] https://github.com/conda-forge/arpack-feedstock
[2] https://github.com/conda-forge/arpackpp-feedstock

[UR_]

for which part ccx uses arpack and for which it uses spooles?

For an overview please refer to docs

http://www.dhondt.de/ccx_2.14.pdf

see page 570 chapter 9.6:

9.6 Filling and solving the set of equations, storing the results
In this section a distinction is made between the types of analysis and the solver
used:
• for linear static calculations with SPOOLES or the iterative solver the appropriate routine is prespooles.c
• for nonlinear static or dynamic calculations (which implies the use of SPOOLES or the iterative solver) routine nonlingeo.c is called. This includes all thermal calculations.
• for frequency analysis without cyclic symmetry routine arpack.c is called.9.6 Filling and solving the set of equations, storing the results 571
• for a frequency analysis with cyclic symmetry conditions the appropriate routine is arpackcs.c
• for a buckling analysis arpackbu.c is called
• for linear dynamic calculations (i.e. modal dynamic analysis) the routine is dyna.c
• finally, for steady state dynamics calculations the routine is steadystate.c
...

[Kunda1]

https://repology.org/metapackages/?sear ... es_newest=

[looo]

I was able to build spooles for windows on my machine. There is still a problem with recursive copying the header files (appveyor):
https://ci.appveyor.com/project/conda-f ... 8069#L1889

If anyone knows a better way to copy header files, please let me know.