Stopping (indefinitely) stopped programs

[jpolitz]

I just had an interesting experience of writing a world program without and on-tick, and not being able to convince the REPL that it had stopped.

This is because in between event handlers, the world is dormant, and the way we pause/break programs needs the program to wake up and notice its been paused in order to notify the world of the break.

This comes from our re-use of pause as break. Break should be a distinct operation that immediately calls the failure continuation of the current evaluation (so it knows that its done, without relying on timeouts), and simply blocks any new computation of the same run that starts later.

I think pause still needs to have the behavior it has now, because it needs access to the actual continuation of the computation in order to restart it. It might be possible to make this asynchronous as well, but there aren't any known issues with the behavior of pause so this isn't as urgent.

[shriram]

I actually had a problem with stopping/pausing too, and I think it was a similar config.

[jpolitz]

Related commits, for future reference; not quite all the way there yet.

4298a7c

c6db335

0165650

[jpolitz]

Marking this as closed, as I think 3f29c2b solves this problem.