i was able to induce another (deadlock that was pr...
# development
a
i was able to induce another (deadlock that was predicted by the TODOs added in that commit), everything still seems to apply when based off master so i'm in a new branch now. looking into addressing the TODO here: https://github.com/pantsbuild/pants/blob/5d850fcadeb3da5e3a3fa075595f9e3bf2e32a47/src/rust/engine/process_execution/src/local.rs#L317. i started by removing the blanket
unsafe impl<T> Send for Resettable<T> {}
and
unsafe impl<T> Sync for Resettable<T> {}
and replacing with those trait bounds, and that only broke the ResettablePool that I'd regrettably turned into a Resettable<CpuPool> already. i also made Resettable<T> actually lazy again (that's what induced the other deadlock, which happens even without pantsd enabled)