... even though we *should* be able to.
If we used effective triples at the marked sites, we would fail:
- test/Driver/{darwin-dsymutil.c, darwin-debug-version.c}
-ccc-print-bindings is expected to print 'x86_64-apple-darwin10', but prints 'x86_64-apple-macosx10.6.0' instead. - test/Driver/netbsd.c
We see -mcpu=arm1022e instead of arm926ej-s in a single run. - test/Driver/linux-as.c
An extra '-mfpu=neon' argument appears in a single run. Probably harmless -- the test could be relaxed.
This is part of on ongoing effort to use effective target triples
whenever we can in the clang driver (see: [cfe-dev] Lessening the
driver's reliance on default target triples).
Is the test correct?