4:15:12gendl__just made a👆 naive implementation which apparently works for my use case, may or may not be useful as a reference. It's naive because it doesn't do anything about "fairness" and it'd be possible for the lock to get swamped by constant overlapping readers (i.e.in an application where the reader count never gets to zero, a writer might be locked out indefinitely). So some kind of "fairness" heuristic should probably be
4:31:28remexreshould I expect writing to the uiop/launch-process:process-info-input of an exited-but-not-waited process to error in some way? unless i'm messing something up, I'm not getting any kind of condition