I'm seeing a bug when using the latest salsa with dada-- I believe the steps to repro are:
create an input
create a memoized fn that returns a tracked struct using that input
modify something unrelated (new revision)
re-execute memoized fn, getting the tracked struct
read from that tracked struct
this results in a panic that the tracked struct is outdated, but I think the problem is that the revision has not been updated when the memoized value was reused. This pattern isn't one I expect to support in the future but it should work on salsa as is. Will investigate deeper and create a repro test tomorrow! (hopefully)
I'm seeing a bug when using the latest salsa with dada-- I believe the steps to repro are:
this results in a panic that the tracked struct is outdated, but I think the problem is that the revision has not been updated when the memoized value was reused. This pattern isn't one I expect to support in the future but it should work on salsa as is. Will investigate deeper and create a repro test tomorrow! (hopefully)