On 8/15/25 10:34 AM, Vineet Gupta wrote:
On 8/15/25 06:13, Jeff Law wrote:
On 8/14/25 6:32 PM, Vineet Gupta wrote:
Changes since v1:
    - Fix scan pattern for math-nearbyint-1.c
    - Per Andrew's suggestion gate flag save/restore on -ftrapping-math
Still failing CI:

https://github.com/ewlu/gcc-precommit-ci/issues/3776#issuecomment-3190438630

I generally do the A/B testing locally but this time I was relying on CI -
forgot to mention this in the patch.
The new failures are due to conditional emit of flags save/restore as suggested
by Andrew - so round_64 built with unsafe math optimizations doesn't see
additional flags as we wanted.
OK to push after readjusting the scan pattern ?
Yes.
Jeff

Reply via email to