Windows support via msys2 #18
Conversation
|
Hi! This is the friendly automated conda-forge-linting service. I wanted to let you know that I linted all conda-recipes in your PR ( Here's what I've got... For recipe:
|
|
Hi! This is the friendly automated conda-forge-linting service. I just wanted to let you know that I linted all conda-recipes in your PR ( |
|
@conda-forge-admin, please rerender |
|
conda-forge/mumps-feedstock#25 is now merged. @conda-forge-admin, please re-render |
|
So we have a working ipopt DLL building on Windows now. I'm going to try also to build the exe to address #17 |
|
Looks like the exe requires AMPL, which we don't support on Windows yet: https://github.com/conda-forge/ampl-mp-feedstock |
|
@conda-forge-admin, please re-render |
| cd build | ||
|
|
||
| rem rm /mingw-w64/bin/make.exe | ||
| bash -c "CXXDEFS=-DCOIN_USE_MUMPS_MPI_H ADD_CXXFLAGS='-I/mingw-w64/include -L/mingw-w64/lib -L/mingw-w64/bin' ../configure --build=x86_64-w64-mingw32 --disable-linear-solver-loader --with-blas-lib='-lopenblas' --with-mumps-lib='-ldmumps -lmumps_common -lpord -lgfortran' --enable-dependency-linking --prefix=/mingw-w64" |
There was a problem hiding this comment.
disabling the linear solver loader is a bad idea - that's important functionality if you care at all about performance
There was a problem hiding this comment.
I'll take another look. I was getting unrelenting segmentation faults trying to get the linear solver loader to work, but I'm fully willing to admit the possibility of user error.
There was a problem hiding this comment.
It can be picky about finding the runtime libraries IIRC, and how the solver you're trying to load was built. I've forgotten what the status is of HSL providing Windows binaries to academic users. Dynamically loading the MKL version of Pardiso may also be flaky
| - ipopt.patch | ||
|
|
||
| - ipopt.patch # [not win] | ||
| - dll_makefile.patch # [win] |
There was a problem hiding this comment.
configure with --enable-shared, shouldn't need to manually patch makefiles
There was a problem hiding this comment.
I'm not sure that I have a Windows-compatible configure script from upstream to do this the right way.
There was a problem hiding this comment.
You do. I wrote the patches upstream that enabled it, 5ish years ago.
| - blas 1.1 {{ variant }} | ||
| - openblas | ||
| - metis 5.1.* # [not win] | ||
| - scotch # [not win] |
There was a problem hiding this comment.
there's no direct dependency on scotch on unix either
|
Will this be usable in downstream projects compiled with MSVC? I'm wondering because of the |
|
I've been working on compiling Ipopt with |
|
I will probably not have time to take another look at this until the new year. I may be able to lightly support someone else picking up the baton in the short term. |
|
Would you be supportive of also using clang-cl, etc. in addition to the msys2 toolchain? |
|
@chrisdembia I'll chime in, I think we'd accept any PR that works here natively with the tool chain conda-forge offers. |
|
Okay great. |
|
Closed in favor of #34 |
Compile Windows builds of ipopt using msys2.
Depends on conda-forge/mumps-feedstock#25