Richard Hodges

You are either: - initiating two or more async_shutdowns at once, or - you are not preserving the lifetime of the websocket stream, or - some other race condition

Difficult to tell from here. Any chance you can roll a minimal 1-page program that exhibits the problem?

What is the text in the assertion message? What is actually failing?

I have it compiled on my home machine. I'll leave it running in a debugger while I head to the office. I'm currently seeing this output: ``` post result:0 reason_or_reply=...

``` (beast::bind_front_handler([self] (beast::error_code ec) {}) ``` this looks a bit suspect. The `bind_front_handler` isn't necessary here as you're capturing `self` in the function object itself.

Are we able to build a compilable single file stress test to try to reproduce?

We do have internal discussions on design improvements that could be made. We’re not entirely happy with the relationship between parser and body_type, for example. So the door is certainly...

None of this is nonsense. I think our view is that users would love an API that looks a lot like python’s requests library. This is actually on my todo...

I have some good news for you. - I've written all this stuff already (including connection recycling) - I do have it on my roadmap to produce such a library...

OK, I'm actually on holiday at the moment, so I'll start while my wife isn't looking. She'll get mad with me if she catches me working ;-)