I've cherry-picked the unit test commits from #31791 to check whether the same behaviour (rounding negative numbers around -n.5 to -n.5 instead of either -n.0 or -(n-1).0) can be observed on NetBSD with current libc implementation (not libzigc).
Test current libc rintl implementation #32070
mihael:add-rintl-unit-tests into master
Instead of testing all the `rint` functions in the same test declaration, have a test declaration per `rint` function and `special`/`normal` split.
libzigc/math: Skip rintl unit tests on PowerPC
Also, switch to `expectEqual` for comparing zeros in `testRintNormal` and `testModf` instead of relying on the combination of `expect` and `isNegative/PositiveZero`. In case of test failure, we want to get the actual value in the log output.
libzigc is never built on NetBSD. The libc API tests run against native libc even on such platforms. See: #31923
@alexrp wrote in #32070 (comment):
libzigc is never built on NetBSD. The libc API tests run against native libc even on such platforms. See: #31923
Oh, then it's a problem with the native libc implementation / environment on NetBSD and not the implementation I added on #31791. I guess we ran into a bug?
Please, feel free to cancel these runs if you can. I'll finalize the other PR and add a note on this, but I'm glad it doesn't regress rintl.
Pull request closed
No due date set.
No dependencies set.
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?