panic in `Regex::regex()` about nlprule HOT 5 CLOSED

Sure! Release CI is running for 0.6.3 now.

from nlprule.

Hi, thanks for the issue. Refreshingly, something which I couldn't have foreseen :)

I believe the easiest fix is just to replace lazycell with once_cell, which is a more well used dependency anyway and has the get_or_init API which fits our use case.

I went ahead and implemented this: #69. Can you check if this solves the problem? Binaries from the latest release should still work.

from nlprule.

A bit of context:

Rules::suggest(...) is called in parallel over a shard immutable reference to a single &Rules instance. Looking into lazycell the ordering semantics look ok from the first glance, the API usage just seems incorrect (even in their example).

If thread A calls borrow() while thread B is just about to write the value to the UnsafeCell, it will be in LOCK state, which will return a None.
The only way I can think of doing this correctly is calling borrow() repeatedly in the else case until it succeeds.

from nlprule.

Just had a peek into the once_cell crate, and there is a significant amount of non trivial synchronization logic happening, but I ran it a few times and it works consistently 🎉 - so #69 is good to go :) Thanks for the fast turnaround 👍

from nlprule.

Could you craft a release, that's the last thing pending for a cargo-spellcheck 0.8.0 which is again needed for paritytech/parity-bridges-common#832

from nlprule.