ghc failed to build in 10.8.5

[apuignav]

Hi all,

I was trying to compile ghc and it fails with message

==> make -j1 install
gcc-4.2 -E  -m64 -undef -traditional -Wno-invalid-pp-token -Wno-unicode -Wno-trigraphs -P -DINSTALLING -DLIB_DIR='"/private/tmp/ghc-6Gskib/ghc-7.8.3/subfo/lib/ghc-7.8.3"' -DINCLUDE_DIR='"/private/tmp/ghc-6Gskib/ghc-7.8.3/subfo/lib/ghc-7.8.3/include"' -DPAPI_INCLUDE_DIR="" -DPAPI_LIB_DIR="" -DFFI_INCLUDE_DIR= -DFFI_LIB_DIR= '-DFFI_LIB="Cffi"' -x c -Iincludes -Iincludes/dist -Iincludes/dist-derivedconstants/header -Iincludes/dist-ghcconstants/header rts/package.conf.in -o rts/dist/package.conf.install.raw
cc1: error: unrecognized command line option "-Wno-invalid-pp-token"
cc1: error: unrecognized command line option "-Wno-unicode"
make[1]: *** [rts/dist/package.conf.install] Error 1
make: *** [install] Error 2

I have followed the Troubleshooting guide and everything looks ok from my side. I have also looked at the open issues and I cannot find anything similar to this.

The gist logs can be found in https://gist.github.com/3317213a78b987a0e59b

Please let me know if I can provide further tests to shed light on this matter.
Albert

[jacknagel]

Is there a specific reason you're building from source rather than using the bottle (which is the default)

Does it work if you brew install ghc --cc=gcc-4.9?

[apuignav]

Hi,

I just do brew install ghc, so I guess it doesn't get picked up...

Albert

[apuignav]

And just to add to this, brew install ghc --cc=gcc-4.9 works. However, I would expect the formula to work out of the box, can I provide some further information to help pinpoint why it's not doing so?

Albert

[jacknagel]

I suspect it hasn't come up because most people don't have apple gcc-4.2 installed and available as a fallback (since it is an Apple compiler, it has precedence over gcc-4.9).

I'll have to determine if it can be made to work or if we should add a fails_with for it.

[tdsmith]

Thanks again for reporting. The appearance of this failure is a consequence of the compiler priority issue now described more fully in #41055; since as of #39134 it is asserted that llvm no longer fails with this identical message, I think it is likely that gcc-4.2 will also succeed. Please let me know if I'm wrong!