My recent patch to check_effective_target_arm_v8_neon_ok_nocache didn't take into account that when compiling the test itself, the option "-march=armv8-a" is added to the flags used for the effective-target check, so it caused more multilibs to be skipped than needed to be.
This patch uses "march=armv8-a" to compile the code in that check, allowing multilibs that don't use conflicting options to run, and pass, tests that are gated by it as they did before my patch went in. Tested with gcc.target/neon tests for mainline arm-none-eabi with many sets of multilib options; OK for mainline and 4.9? Janis
2014-08-18 Janis Johnson <jani...@codesourcery.com> gcc/testsuite/ * lib/target-supports.exp (check_effective_target_arm_v8_neon_ok_nocache): Add "-march-armv8-a" to compile flags. Index: gcc/testsuite/lib/target-supports.exp =================================================================== --- gcc/testsuite/lib/target-supports.exp (revision 214084) +++ gcc/testsuite/lib/target-supports.exp (working copy) @@ -2587,7 +2587,7 @@ { __asm__ volatile ("vrintn.f32 q0, q0"); } - } "$flags"] } { + } "$flags -march=armv8-a"] } { set et_arm_v8_neon_flags $flags return 1 }