Finn Bear

> I just looked into your patch some more. The `makeArgs`/`storeArgs` solution is really clever. Thanks! (it's just a workaround for a lack of a way to force the Go...

> Is there any scenario in which a string pointer could point to the stack? Yes, although I admit it may not be common. `//go:noescape` on `valueGet`, `valueSet`, and `valueDelete`...

> can `handleEvent` also reuse slices to avoid slice allocations? This can be dangerous if `handleEvent` is re-entrant (it theoretically can be, if the Go code causes a JavaScript immediate...

> My understanding of your latest proposal is that if the capacity is a small constant value (16 in this case), the slice allocation will be on the stack. Is...

> Another question: how can we make sure `makeArgs` is inlined? In my understanding, there is no way to force Go to inline something other than in-lining it manually. If...

FWIW the [documentation on `http1_header_read_timeout`](https://docs.rs/hyper/latest/hyper/server/conn/struct.Http.html#method.http1_header_read_timeout) implies to me that the timeout would start immediately and stop when the header was fully transferred. I would have expected to see "If a...

_This is a restatement of https://github.com/hyperium/hyper/pull/3185#issuecomment-1506298523 with additional details._ > think, from talking to several people, many find it clearer for there to be a separate timeout. Clarity aside, isn't...

Note that, in debug mode, you start with more points. I am aware that you are having an issue (#214) *compiling* in debug mode though :wink:

Thanks! We noticed this issue on our end and have a fix in the next release. I recommend commenting out the audio-related code if you just need the texture-related code...

Hi, We usually recommend an ``-based solution. Even if the client could be served statically, it would break after every update.