cfallin commented on Issue #2581:
So this was it then? I'm really happy we were able to finally pin this down!
Could you add a comment describing the reason for the initial zeroing -- that NaN != NaN so
x == x
doesn't always generate all-1s?
jlb6740 commented on Issue #2581:
So this was it then? I'm really happy we were able to finally pin this down!
Could you add a comment describing the reason for the initial zeroing -- that NaN != NaN so
x == x
doesn't always generate all-1s?@cfallin .. Yes, looking like that was it. I haven't done a lot of testing so we need to monitor. I will put a comment.
cfallin commented on Issue #2581:
This bug has made me realize that we should be running tests under valgrind on CI -- it would've deterministically found this issue and made our lives a little easier! I'll create an issue for that.
jlb6740 commented on Issue #2581:
This bug has made me realize that we should be running tests under valgrind on CI -- it would've deterministically found this issue and made our lives a little easier! I'll create an issue for that.
Yes, I was thinking the same thing. Either that and/or this memory sanitizer. Clean up the current issues and make sure it runs to completion.
jlb6740 edited a comment on Issue #2581:
This bug has made me realize that we should be running tests under valgrind on CI -- it would've deterministically found this issue and made our lives a little easier! I'll create an issue for that.
Yes, I was thinking the same thing. Either that and/or memory sanitizer. Clean up the current issues and make sure it runs to completion.
Last updated: Dec 23 2024 at 12:05 UTC