Digging in deep with #165 I noticed the only reason we need this "nursery is now exited by parent event was due to the .run_in_actor() style task-per-actor API we're debating dropping as in #290. Without this event, errors from remote actors are effectively raised right away since there is no collection of task per actor* style teardown - this is the cause of a test suite massacre if I just remove it and run the CI 😂
If we were to remove this:
spawn tasks move on immediately to a soft_wai() which is just as fine a spot to block right?
is there any reason to ensure error checking from remote tasks is delayed until an ActorNursery.__aexit__()? No right, since all errors should be raised immediately (presuming no IPC msging failures) at a higher level then the process spawning/lifetime management?
Digging in deep with #165 I noticed the only reason we need this "nursery is now exited by parent event was due to the
.run_in_actor()style task-per-actor API we're debating dropping as in #290. Without this event, errors from remote actors are effectively raised right away since there is no collection of task per actor* style teardown - this is the cause of a test suite massacre if I just remove it and run the CI 😂If we were to remove this:
soft_wai()which is just as fine a spot to block right?ActorNursery.__aexit__()? No right, since all errors should be raised immediately (presuming no IPC msging failures) at a higher level then the process spawning/lifetime management?More thinking to do on this fo sho 🏄🏼