On Fri, Mar 31, 2023 at 01:02:18PM -0400, Andrew MacLeod wrote: > I guess it figures the recip is safe to put in, there will not be a divide > by zero.
I think the problem was that 1/d was hoisted before the loop; as long as it is guarded with the d > 0.01 or e > 0.005 condition, it is fine. The test probably should have been a runtime test, doing the main stuff in some other noipa function and doing fetestexcept after it or something similar. > I guess the test is no longer testing what it should be? > > And yes, we could set he param back to 1 for the test... > add --param=ranger-recompute-depth=1 makes the "issue" go away :-) for > now. That looks reasonable unless we rewrite the test into runtime one (but we'd then need to double check that it was really miscompiled and would fail back then in 4.0). Jakub