Refactor BaseInterface context and stop.

[fleming79]

No description provided.

[x42005e1f]

There is aiologic.lowlevel.async_sleep_forever(). It is much more efficient and also does not suffer from the math.inf issue (since it uses a different approach to waiting; you can check _time.py if you want).

[fleming79]

Okay thanks for that. I didn't even think to check if you'd provided one.

Are you planning to adjust your _MAXIMUM_SECONDS_PER_SLEEP based on what Pyodide decided? Not perfect, but it is an improvement on the current behavior.

[x42005e1f]

Yes, I think that makes sense. Most of my analysis has focused on standard event loops, so for now, the code assumes that the overflow issue is handled by the event loop itself. I probably cannot do it today, since I am too busy, but I might be able to fix it one of the next few days.

[fleming79]

It can probably just be an issue to leave open. It is a pretty rare issue.

[x42005e1f]

But it is also an issue that can cause damage almost silently. The probability of an issue occurring is just one component of the risk it poses. For example, many thread-safety issues are very rare, but even so, they are taken quite seriously.