Re: MXE/MSVC

[Michael Goffioul]

On Mon, Jun 17, 2013 at 12:05 AM, John W. Eaton <address@hidden> wrote:

On 06/16/2013 07:53 PM, Michael Goffioul wrote:

Here's a patch that enables BLAS compilation with MSVC.

John, what do you want me to do with those patches. Should I keep them
separated from MXE or do you want me to integrate it into it? I'm fine
with keeping things separated, I just want to know how I'm gonna handle
those when they start piling up.

I think we should include all the changes for MSVC in mxe-octave.  Jordi should be able to give you push access to the archive on hg.octave.org.  I would do it myself, but I'm not sure exactly what needs to be done to set it up.

If the change is a patch that is specific to MSVC and that requires MSVC to work (no reasonable way to #ifdef, for example) then we will need to fix mxe-octave to apply system-dependent patches.  Currently, all patches in the src directory are applied for all systems.  That design decision made sense for the original MXE because it was only targeting one system (mingw cross).  I suspect we may need more flexibility than that.

That's seomthing I've added in my first patch to MXE. A patch can also have the name "$(MXE_SYSTEM)-$(1)-*.patch" and it will only be applied in the corresponding system. I used that to patch libiconv.

 

 Ideally, we would get all the patches in the mxe-octave src directory applied upstream so we could remove them from mx-octave.  I'd like to work toward that, but I'll be very (pleasantly) surprised if we will ever be able to eliminate all patches from mxe-octave.

This may work for MinGW, but I doubt it'll work for MSVC, as it's regarded as the evil compiler (experience tells me that common answers are "get yourself a real compiler", "MSVC is not free", "it works in MinGW, why should I change my code"...). 

Michael.