-fmax-errors not support by clang

[pbrady]

https://github.com/lanl/singularity-eos/blob/main/CMakeLists.txt#L449 unconditionally adds -fmax-errors=3 which leads to the following warning:

clang: warning: argument unused during compilation: '-fmax-errors=3' [-Wunused-command-line-argument]

[Yurlungur]

Good catch---we should add a clang build to our CI... In the meantime, @mauneyc-LANL, @rbberger and/or @dholladay00 maybe we add that in with a generator expression for non-clang builds only?

[mauneyc-LANL]

This is my fault. Usually when developing I put this in so that I have a reasonable shot of locating the compiler error before my terminal burns through main memory printing out 1000s of incomplete template noise.

We should really just remove it altogether; any objections? @Yurlungur @dholladay00 @rbberger

[Yurlungur]

None---let's just remove it. Thanks, @mauneyc-LANL

[mauneyc-LANL]

Good catch---we should add a clang build to our CI... In the meantime, @mauneyc-LANL, @rbberger and/or @dholladay00 maybe we add that in with a generator expression for non-clang builds only?

We should use the intel compilers in our CI if we arn't already, too

[Yurlungur]

Yeah @dholladay00 brought that up recently. I think he's working on making that happen.

[dholladay00]

Yeah @dholladay00 brought that up recently. I think he's working on making that happen.

I am fixing some issues with intel builds, but I haven't put in an intel build in our CI. I think that would require additional spacks buildouts and interacting with the project space in ways I am not very familiar with. Perhaps @ktsai7 or @rbberger could look into adding such a thing to our CI or replacing a current gnu build with an intel.

[ktsai7]

@dholladay00 sure, we can add an intel CI job, but do we want a new CI job or replace an existing one? Which intel compiler?

[dholladay00]

@ktsai7 I'm looking and there are several gnu Skylake tests. I'd say replace test_gnu_skylake with an intel test that turns on all of the bells and whistles (fortran, closure, kokkos and kernels, eospac, spiner), but I could be convinced to just add this in addition.

[ktsai7]

I'm fine with either way, but I do need to know which intel compiler do you want to use?

[mauneyc-LANL]

@ktsai7 @dholladay00 we should be using the oneAPI compilers (classic intel C/C++ compilers are losing support), but there's still broad downstream use of classic compilers.

The oneAPI compilers use a clang frontend, so having a compile with just a clang++ would be good to be out in front of the eventual changeover.