> That regresses on gnat.dg/specs/atomic1.ads for aarch64/-mabi=ilp32, > missing the error on line 13. The missing error on line 9 is > preexisting.
That's sort of expected, since the point of the patch is to make the 2 situations equivalent wrt atomicity. Clearly the test is not portable and will fail on targets for which pointer size and word size are not equal. -- Eric Botcazou
