gcc_impl_<target> contains stuff it probably shouldn't contain?

[h-vetinari]

Looking at the content of gcc_impl_linux-64 (and removing the SOVERSION distinctions), there's a bunch of things in that output that probably shouldn't be there?

x86_64-conda-linux-gnu/lib/libasan.so
x86_64-conda-linux-gnu/lib/libasan_preinit.o
x86_64-conda-linux-gnu/lib/libatomic.so
x86_64-conda-linux-gnu/lib/libgfortran.so       # !!!
x86_64-conda-linux-gnu/lib/libgomp.so           # !!!
x86_64-conda-linux-gnu/lib/libgomp.spec
x86_64-conda-linux-gnu/lib/libhwasan_preinit.o
x86_64-conda-linux-gnu/lib/libitm.so
x86_64-conda-linux-gnu/lib/libitm.spec
x86_64-conda-linux-gnu/lib/liblsan.so
x86_64-conda-linux-gnu/lib/liblsan_preinit.o
x86_64-conda-linux-gnu/lib/libquadmath.so       # ???
x86_64-conda-linux-gnu/lib/libsanitizer.spec
x86_64-conda-linux-gnu/lib/libtsan.so
x86_64-conda-linux-gnu/lib/libtsan_preinit.o
x86_64-conda-linux-gnu/lib/libubsan.so

In particular, libgfortran and libgomp have separate outputs themselves, and libquadmath is also used by gfortran, without ever bringing gcc into the game.

[h-vetinari]

OK, I guess this works because gfortran_impl depends on gcc_impl

ctng-compilers-feedstock/recipe/meta.yaml

Line 382 in 8853ee7

- gcc_impl_{{ cross_target_platform }} >={{ gcc_version }}

And since it's in the target-specific folder, it shouldn't be picked up by accident. Still scratching my head a bit about the layering, but happy to close if there's nothing to do...