A push is still needed, I think, as in the other PR? But maybe I'm mixed up. @sbc100 ?

This looks good to me. The idea is that we want to keep the runtime alive as long as there is an outstanding timeout. Here we call reutimeKeeplivePush when we first set the timeout, and we only pop it if/when we fail set the timeout once again in the callback.

I won't if we can make a test case for this that fails with the old/broken code?

A push is still needed, I think, as in the other PR? But maybe I'm mixed up. @sbc100 ?

This looks good to me. The idea is that we want to keep the runtime alive as long as there is an outstanding timeout. Here we call reutimeKeeplivePush when we first set the timeout, and we only pop it if/when we fail set the timeout once again in the callback.

I won't if we can make a test case for this that fails with the old/broken code?

Actually looking more closely I don't see a bug in the old code. The new code maybe avoids the extra Pop/Push for each iterations, but the old code seems to be perfectly correct.

I do see one difference in that, under the old code, the value of the keepalive counter will be zero while the callback itself is run . This means that if the user called exit() during callback, the application would exit, rather than staying alive to service the next iteration of the loop.

With the new code the loop is still considered alive for the duration of the callback, so calls to exit won't actually exit the application during this time.

I think perhaps locally the new code make more sense, since it requires and explicit cancellation of the loop to exit the runtime.

Here we call reutimeKeeplivePush when we first set the timeout

But there is no push in the new code (the push is deleted, and the pop is moved around). Am I reading it wrong?

Here we call reutimeKeeplivePush when we first set the timeout

But there is no push in the new code (the push is deleted, and the pop is moved around). Am I reading it wrong?

The push is below on line 106. This push now remains active until one of the callbacks returns 0, which is this is the desired effect.