py-pytorch: builds now failing for 10.12 and 10.13

[mascguy (Christopher Nielsen)]

This port was rev-bumped, due to a breaking ABI change related to opencv4 4.5.5. However, builds are now failing on 10.12 and 10.13 - even though they succeeded previously - and despite no updates to py-pytorch.

The compilation errors appear to be (potentially) utterly trivial:

/opt/local/var/macports/build/_opt_bblocal_var_buildworker_ports_build_ports_python_py-pytorch/py39-pytorch/work/pytorch-1.8.1/third_party/benchmark/src/complexity.cc:82:10: error: variable 'sigma_gn' set but not used [-Werror,-Wunused-but-set-variable]
  double sigma_gn = 0.0;
         ^
1 error generated.

​https://build.macports.org/builders/ports-10.13_x86_64-builder/builds/147862/steps/install-port/logs/stdio

It's a bit unclear as to why a set-but-unused variable is now a compilation error, when that was not the case previously. And this doesn't appear related to the Clang version either: Whether this port is compiled with older Xcode Clangs (8.0/9.0) - or MacPorts Clang 13 - this case is now flagged as an error [on 10.12/10.13].

In other words, yes: I've tried blacklisting Xcode Clangs < 1001, to ensure Clang 13 is used for compilation on 10.13. Yet, we're still seeing this.

However, the compilation defaults on 10.14 and higher are such, that this case is NOT flagged as an error. Does anyone happen to know why the defaults are different for 10.14 and above, vs. 10.13 and below... regardless of which Clang version is used? (I could probably figure this out, with enough time spent digging through the changes to both our portgroups - and base 2.7.2 - but hoping someone else might know off the top of their head.)

As for the fix, it may be a simple matter of ensure set-but-unused variables are not flagged as errors. So potentially no big deal there.

But I'm more curious as to where this behavior has changed, between our portgroups and base 2.7.2. Thoughts?