Hacker News Reader

Comment by tavert | Hacker News Reader

Try compiling SciPy from source using Microsoft dev tools. Or libgmp, the standard library for arbitrary precision integer arithmetic.

View on HN

I'm going to have to pretty much instantly admit defeat with SciPy: https://github.com/scipy/scipy/issues/5461#issuecomment-2871...

That issue is much more complicated than what most FOSS projects deal with.

I agree with you, building a lot of non-Windows stuff on Windows sucks. It's getting better, though.

With libgmp, it looks like you should use http://www.mpir.org/ but I don't know how this works with programs that expect gmp.

> That issue is much more complicated than what most FOSS projects deal with.

R or Julia would have the exact same problems if they cared at all about MSVC.

> I agree with you, building a lot of non-Windows stuff on Windows sucks. It's getting better, though.

I actually didn't say that at all. It's the Microsoft dev tools that suck. GCC works fine.

That's easy. For an extra challenge, try python-mapnik (libgmp is just one of the dependencies).

Anything that depends on both boost and gdal is bound to be pain. I once spent 22 hours getting a build of GDAL on FreeBSD with the drivers I needed.

Easy how? Otherwise the scipy maintainers would be doing it.

Relatively easy. I gave you an example, that has several dozens of dependencies, with several different build systems, of which not all of them support building with VS.

And that's targeting py3. For py2, you would need to build with VS2008, which some packages flat out refuse (for example, icu wants at least VS2010).

How many of them don't support MSVC due to requiring features that MSVC can't handle (like inline assembly or Fortran), vs not wanting to deal with build system contortions?

Honestly, I have no idea about the current status. Last time I tried, it was some two years ago and it was a major PITA. Since then, when I need newer versions, I just use either Linux or Mac.

Reply on news.ycombinator.com