add modifications to handle gfortran on msvc#9429
Conversation
|
Not sure if you still want feedback given that you closed the PR, but: if it takes listing scipy submodules explicitly then that's a lot better than nothing, but it's then not a generic solution for "Fortran + Python on Windows". I haven't been following in enough detail, but how sure are you that this is necessary? |
|
As for the library order: the easiest solution could be to specify it
manually in setup.py. Does numpy.distutils preserve the library order
given by the setup.py scripts? If not, it could be made to do that, and
we could fix the library order on Scipy side.
|
It doesn't, and making it do so will be useful in other contexts as well. E.g. I just tried to move |
I inadvertently deleted the branch and github closed the PR. @rgommers The other issue that I am unsure how to address is how to gracefully handle DLL dependencies . Do you have any ideas on this? |
|
new pr: #9431 (gh wouldn't let me reopen this) |
This is currently wip (it fully works but needs to handle non msvc+gfortran). Feedback would be greatly appreciated to understand whether this is the correct approach.