sente/reconnect doesn't reset backoff-ms attempt number when chsk is closed

[danielcompton]

From my testing, it seems like when calling sente/reconnect from a closed chsk, the backoff-ms attempt number isn't reset. The use case for this is displaying to the user the number of seconds before reconnecting, but giving them the option to reconnect faster if they want to. Currently, they can click a reconnect button, but if the channel socket is closed then nothing will happen, and Sente will try to reconnect on it's normal schedule.

Is this understanding correct, and is this a bug?

[ptaoussanis]

Hey Daniel,

Your understanding's correct - though the behaviour's not technically a bug. chsk-reconnect!'s currently documented as: "Drops connection, allows auto-reconnect. Useful for reauthenticating after login/logout."

i.e. chsk-reconnect! was originally added as a quick hack as a way of dropping an existing connection to allow it to come back up. Didn't have in mind to allow it to speed up reconnect attempts of already-dropped connections, though that's a totally reasonable expectation from the name.

Would be happy for a PR to adopt the behaviour you're describing (+ update the docstring). Might not be an obvious/easy way to do this btw, so feel free to leave this to me if it's looking like it might be tricky.

Can try take a look at this myself at the weekend. Thanks for the report, the current behaviour's definitely surprising.

[danielcompton]

I'll have a look tomorrow with clearer eyes, but I think this might be part of the solution (also a note to self):

(reset! nattempt_ 0)
(.clearInterval js/window @kalive-timer_)

[ptaoussanis]

but I think this might be part of the solution

The nattempt_ looks good, but note that the keep-alive timer and backoff timers are different. The backoff timer isn't currently accessible for cancelling; it's just an ephemeral thing created in set-exp-backoff-timeout!.

So we may need to add a new backoff-timer_ atom (like kalive-timer_) that we do track to allow cancelling.

[lfn3]

So I've taken a swing at this, and what I went with is adding the connect! fn to the record and using that directly while clearing the interval timer. This means the timer gets reset but the number of attempts doesn't (which seems desirable?).

Does that sound good to you guys?

[danielcompton]

Open a PR and we can take a look :)

[ptaoussanis]

Fixed in the SNAPSHOT I'm busy uploading now...