python: add C++ compiler support for distutils

[veprbl]

Motivation for this change

Fix #18729

Things done

• Tested using sandboxing
  (nix.useSandbox on NixOS,
  or option build-use-sandbox in nix.conf
  on non-NixOS)
• Built on platform(s)
  • NixOS
  • OS X
  • Linux
• Tested compilation of all pkgs that depend on this change using nix-shell -p nox --run "nox-review wip"
• Tested execution of all binary files (usually in ./result/bin/)
• Fits CONTRIBUTING.md.

---

[mention-bot]

@veprbl, thanks for your PR! By analyzing the history of the files in this pull request, we identified @cstrahan, @vbgl and @FRidh to be potential reviewers.

[FRidh]

We should of course try to fix the situation on Darwin, while at the same time stick as much as possible to upstream.

It's not yet clear to me whether this situation is being dealt with upstream. Furthermore, if we fix an issue, we should fix it for all versions of the interpreter that are affected.

[FRidh]

Is this patch accepted upstream? How does this affect non-Darwin? Does this fix anything on non-Darwin?

[FRidh]

and how about other Python versions?

[veprbl]

You can see in the original issue http://bugs.python.org/issue1222585 that distutils were considered frozen at the time, so this has became a won't fix. This Arfrever person was maintaining it outside of the source tree. The patch is supposed to fix things that get compiled with non-gcc on any platform, but the main target is of course Darwin.

There are also patches presented for python 3.3 and 3.4, it doesn't look like it made to python 3.5. I personally don't consciously use any python3 software and I don't know if they use distutils or if they use C++ code at all.

[FRidh]

You can see in the original issue http://bugs.python.org/issue1222585 that distutils were considered frozen at the time, so this has became a won't fix. This Arfrever person was maintaining it outside of the source tree. The patch is supposed to fix things that get compiled with non-gcc on any platform, but the main target is of course Darwin.

If its not accepted upstream, then I think we should only use it when it is necessary. If I understand correctly, it is necessary on Darwin, so I suggest using an optional stdenv.isDarwin for appending this patch and the others as well.

[veprbl]

How about optional !(stdenv.cc.isGCC or false)?

[FRidh]

That's fine as well.

[knedlsepp]

I don't understand the "or false" part. Wouldn't !(stdenv.cc.isGCC) yield the same result?

[veprbl]

or false is a special syntax for the case when isGNU is not defined to prevent evaluation from failing

[FRidh]

has this patch/issue been reported upstream (numpy)?

[veprbl]

I made this patch myself. I don't see the point of reporting the issue before the main distutils patch is merged to python2, which won't happen.

[FRidh]

This patch, which has landed on master, breaks building scipy on Linux.

[FRidh]

it should probably be lib.optionals (isPy27 && !(stdenv.cc.isGNU or false))

[FRidh]

has this patch/issue been reported upstream (cython)?

[veprbl]

cython/cython#1491

[FRidh]

thanks

[veprbl]

I don't understand this part very well. The hunk of the original patch that I removed, namely:

--- Makefile.pre.in
+++ Makefile.pre.in
@@ -455,7 +455,7 @@
        *\ -s*|s*) quiet="-q";; \
        *) quiet="";; \
    esac; \
-   $(RUNSHARED) CC='$(CC)' LDSHARED='$(BLDSHARED)' OPT='$(OPT)' \
+   $(RUNSHARED) CC='$(CC)' LDSHARED='$(BLDSHARED)' CFLAGS='$(CFLAGS)' \
        $(PYTHON_FOR_BUILD) $(srcdir)/setup.py $$quiet build

 # Build the platform-specific modules

seems to set CFLAGS in these config vars, however this chunk seems to make it so that the CFLAGS value from the makefile is ignored. So this looks like an inconsistency in original patch.

Perhaps a gentoo bug https://bugs.gentoo.org/show_bug.cgi?id=548776 needs to be tried out.

[veprbl]

Yes, because this patch depends on the other one

26 окт. 2016 г., в 10:45, Frederik Rietdijk notifications@github.com написал(а):

@FRidh commented on this pull request.

In pkgs/development/python-modules/numpy.nix:

@@ -12,6 +12,12 @@ in buildPythonPackage (args // rec {
buildInputs = args.buildInputs or [ gfortran nose ];
propagatedBuildInputs = args.propagatedBuildInputs or [ passthru.blas ];

• patches = lib.optionals isPy27 [

• See cpython 2.7 patches.

• numpy.distutils is used by cython during it's check phase

• ./numpy-distutils-C++.patch
  it should probably be lib.optionals (isPy27 && !(stdenv.cc.isGNU or false))

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or mute the thread.