03:40 <SeanTAllen> patroclos: awesome!

10:23 <Candle> vaninwagen: tcpdump shows that the server is sending the first FIN packet. I agree that sending to an already closed connection is reasonable behaviour though. I have an example that's working, and curl is sending the FIN packet as it should, when it's finished receiving the http response.

11:39 <vaninwagen> Candle: i see. so, the server sending the first fin is strange here, right? because it does so before sending the actual response?

13:23 <Candle> vaninwagen: https://gist.github.com/8e9e4fbab749d5f5a504251d481e18bd is all working, as expected.

13:28 <Candle> I've been trying to make a minimal testcase - but, so far, unable to reproduce the exact issue. The failure mode *seems* to be around calling the 'fun ref _do_the_thing' - it seems to return from that function and then apply, and then terminate the connection from the server end.

14:51 <SeanTAllen> Candle I dont know the code but from what you are describing.

14:51 <SeanTAllen> I would imagine that there is an async callback that happens

14:51 <SeanTAllen> BUT... whatever shuts the connection down is doing it synchronously,

14:51 <SeanTAllen> and thereby "jumping the line"

14:51 <SeanTAllen> That's a guess

14:52 <SeanTAllen> If an actor needs to wait for all async behaviors that might already be waiting to run, then it should do its shutdown async as well.

14:52 <SeanTAllen> but, i dont know the code at all, its just a guess from what you are describing

14:52 <Candle> There is an async callback; the gist has an example of that happening, and it working.

14:56 <Candle> ^ Yes, that's what I think was happening, but have yet to get a unsable testcase!