👍LGTM, finally we can fix #51 . Maybe we should also decrease read_timeout for better responsiveness?

Hmm - I thought about that. I removed it on my machine and the CPU rose a little ~9%, but I didn't look very deeply at it.

Tbh, I'm a little afraid to lug this in as well. Looks like this next version will have lots of core changes. As I understand it, the timeout doesn't run the risk of us losing packets or anything like that, right? I think the only change right now would be that we might quit faster? Or am I missing something?

I think the only change right now would be that we might quit faster? Or am I missing something?

That's my understanding, the faster we exit select the more times we will read the atomic bool, which takes CPU. It's not an easy decision :)

We can play with the values in a followup if 1s of quitting is not problematic for now.

A little bit late... but I just found out that we can patch dependencies by using this feature of cargo or that feature of cargo

@zhangxp1998 - that was my first thought, but according to our friends on ##rust on freenode, this does not work for binary crates installed from cargo. :(

I saw the other issue you opened and will fix the various forks, ugh! :)