stonylohr
commented
2 years ago Thanks, @michaelkirk. My recollection is that I had addressed the items that turned out to be clear, and that the remaining ones were murkier and fine to leave for another day.
Long version: Many remaining items seemed to boil down to the fact that Rust trig functions give results that sometimes differ by 1 in the least significant digit from their C++ counterparts for the same input, which means that tests designed to test exact 0 special cases often end up missing, resulting in larger than ideal difference from the expected result. It's not clear to me that there's much to be done about it without taking impractical measures like writing an alternate implementation of all the trig functions. It's not even clear to me that the Rust versions are less right than the C++ versions. My hunch is that it's a question how to handle digits that can't be fully resolved with the precision available, and that each would end up being closer about half the time, though I haven't chased that theory.
michaelkirk
[x] I added an entry to
CHANGES.mdif knowledge of this change could be valuable to users.There hasn't been any recent changes, but I never released the fixes @stonylohr made last year. I should do that.
Let me know if there's anything else you'd like me to wait for @stonylohr, otherwise I'll plan on releasing in a couple days and getting it into whenever the next georust/geo release is.