On Wed, 2 Nov 2016, James Greenhalgh wrote: > Done in this revision, though this makes no difference to what gets tested > as these options are not added when compiling the probe functions > (check_effective_target_float16). These would need a patch like: > > diff --git a/gcc/testsuite/lib/target-supports.exp > b/gcc/testsuite/lib/target-supports.exp > index abc4ac0..36573cc 100644 > --- a/gcc/testsuite/lib/target-supports.exp > +++ b/gcc/testsuite/lib/target-supports.exp > @@ -2485,7 +2485,7 @@ proc check_effective_target_has_q_floating_suffix { } { > proc check_effective_target_float16 {} { > return [check_no_compiler_messages_nocache float16 object { > _Float16 x; > - }] > + } [add_options_for_float16 {}]] > } > > proc check_effective_target_float32 {} { > > To actually enable the tests.
Then I think such a change should be made (for all these functions, probably separately from the ARM patch) so we have proper test coverage for _Float16 on ARM. (I noted when adding the dg-add-options to these tests that I hadn't tested this properly enabled the _Float64x / _Float128 tests on powerpc64le and that that could do with powerpc maintainer testing.) -- Joseph S. Myers jos...@codesourcery.com