Taylor R Campbell <riastr...@netbsd.org> writes: > What you've done is fine, but what I do -- and what I suggest everyone > do, though we don't have any mandatory rule -- is:
Yeah, I realize that filing a PR and making the test xfail would be nicer, but my question -- which you answered was about whether the corner-cutting of not filing a PR, and not marking the test xfail, was problematic enough that I shouldn't, vs making the job harder and maybe not getting to it. I very much appreciate the process point -- and I did do full atf runs after the commit to make sure it didn't have unintended consequences -- but it's a tough call how much is helpful vs not when looking at work achieved per hour. This was the first time I wrote an ATF test, or I forgot the last time. I did this because it was making proj regression tests fail. The bug was not really an edge case in terms of a slightly wrong answer; it was spectacularly wrong. But it's fixed in current and in the 10 pullup queue, and locally tested on 9, to be submitted when I'm sure the 10 pullup was ok. It remains to fix remquof and remquol. For those I might just file a bug. It would be great if Someone were to diff our libm with FreeBSD and see if we are missing more fixes. Or to write sort of fuzzing tests with random inputs where we can run on FreeBSD and then NetBSD and compare answers.
