Closed spall closed 2 years ago
Thanks for the fix - very nice and localised!
I forgot about the lack of printing of commands; not sure if you have an idea about that.
It's cosmetic, so happy to address it later - no idea what is causing it, as the code here doesn't seem related. Are you sure this is the diff that breaks it?
I tested that the new code at line 228 actually raises the hazard by making it always throw a hazard and running it. I wasn't sure how to add this to the existing test code since it is a race condition.
CmdRattleFinished now tells us when the cmd started, stopped, and the files it read, because we only care about the read files during our hazard re-checking. Then if the cmd is requested again addhazardset is called again near like 228 with those 3 pieces of info. And, it doesn't matter if the hazard set is changed, because we discard it after this check.