Ian Henriksen
Ian Henriksen
Nevermind, that was just a testing failure causing a false success. Marking this as a draft again for more debugging.
Okay, this still needs some restructuring for clarity, but it is actually working now.
Good enough. There's more to do to get thread sanitizer working correctly with our threadqueues, but this is a solid step forward.
This would matter any time you want to use both numpy dtypes and numpy types as keys in a dictionary. I've been wanting to do something like this to map...
Fixed in https://github.com/sandialabs/qthreads/pull/287/commits/233bde8f699bbca0101ade21234a8b40e6a8b397
This appears to have been resolved by one of the other thread sanitizer patches. Probably #206. This can also be closed when that one's merged.
This also shows up inside the hash map when running the `syncvar_prodcons` test with thread sanitizer on. Specifically: https://github.com/sandialabs/qthreads/blob/d6ce514a70c65b74c5e04906615ec51c7f288e0f/src/hashmap.c#L412 and https://github.com/sandialabs/qthreads/blob/d6ce514a70c65b74c5e04906615ec51c7f288e0f/src/hashmap.c#L219
Same thing was going on with `syncvar_prodcons`. Just added a commit that resolves this in #235.
Nevermind. It was just intermittent enough to look solved. The `syncvar_prodcons` intermittent failure is fixed though. Still investigating.
This hasn't been happening for a long while. Tentatively closing.