homebrew-macos-standard-gcc_spkg: library not found for -lmpc while building gcc

[mkoeppe]

On macOS with a bunch of homebrew packages installed and with ./configure --with-system-gcc=no:

[gcc-9.2.0]   	  cc1-checksum.o libbackend.a main.o libcommon-target.a libcommon.a ../libcpp/libcpp.a ../libdecnumber/libdecnumber.a libcommon.a ../libcpp/libcpp.a  -liconv ../libbacktrace/.libs/libbacktrace.a ../libiberty/libiberty.a ../libdecnumber/libdecnumber.a   -lmpc -lmpfr -lgmp   -lz
  [gcc-9.2.0]   ld: library not found for -lmpc
  [gcc-9.2.0]   collect2: error: ld returned 1 exit status
  [gcc-9.2.0]   ld: library not found for -lmpc
  [gcc-9.2.0]   collect2: error: ld returned 1 exit status
  [gcc-9.2.0]   make[6]: *** [cc1plus] Error 1
  [gcc-9.2.0]   make[6]: *** [cc1] Error 1
  [gcc-9.2.0]   make[6]: Target `all' not remade because of errors.
  [gcc-9.2.0]   rm gfortran.pod gcc.pod
  [gcc-9.2.0]   make[5]: *** [all-stage2-gcc] Error 2
  [gcc-9.2.0]   make[5]: Target `all-stage2' not remade because of errors.
  [gcc-9.2.0]   make[4]: *** [stage2-bubble] Error 2
  [gcc-9.2.0]   make[4]: Target `stage3-bubble' not remade because of errors.
  [gcc-9.2.0]   make[3]: *** [all] Error 2
  [gcc-9.2.0]   ********************************************************************************
  [gcc-9.2.0]   Error building gcc-9.2.0
  [gcc-9.2.0]   ********************************************************************************
  [gcc-9.2.0]   
  [gcc-9.2.0]   real	29m43.664s
  [gcc-9.2.0]   user	215m9.194s
  [gcc-9.2.0]   sys	11m33.054s
  [gcc-9.2.0]   ************************************************************************
  [gcc-9.2.0]   Error building package gcc-9.2.0
  [gcc-9.2.0]   ************************************************************************
  [gcc-9.2.0] Full log file: /Users/mkoeppe/s/sage/sage-rebasing/worktree-algebraic-2018-spring/logs/pkgs/gcc-9.2.0.log

To test: tox -e local-homebrew-macos-standard-gcc_spkg

(from #26426)

[dimpase]

What's the point of building gcc on MacOS? One cannot build Sage without Apple's or Conda's clang, anyway, as Apple polluted its system headers with blocks, which are not understood by unpatched gcc.

[jhpalmieri]

That's a good question. The Sage gfortran package builds, unlike gcc, although it seems to lead to some doctest failures.

[mkoeppe]

Replying to dimpase:

One cannot build Sage without Apple's or Conda's clang, anyway, as Apple polluted its system headers with blocks, which are not understood by unpatched gcc.

Do we know if this is still true with current gcc versions? It's not like macOS is some obscure platform.

Anyway, it's not important to fix for 9.1.

[dimpase]

Replying to mkoeppe:

Replying to dimpase:

One cannot build Sage without Apple's or Conda's clang, anyway, as Apple polluted its system headers with blocks, which are not understood by unpatched gcc.

Do we know if this is still true with current gcc versions? It's not like macOS is some obscure platform.

You can try with Homebrew's gcc, but note that ​https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78352 is still open.

[mkoeppe]

I think this may just have been a problem caused by a bad .homebrew-build-env. With #29607, gcc installs correctly and in the build so far I only see a failing givaro.

[mkoeppe]

The givaro build failure is now #29613.

[jhpalmieri]

I agree, gcc builds with #29607.