Support for pervasive timeouts

[reem]

We need to provide timeouts for all of our blocking APIs and ways to set timeouts for all internal blocking actions.

[seanmonstar]

Indeed, and we had a PR offering timeouts, but the IO reform currently means there is no API to do so for TcpStreams.

[reem]

There is now going to be a stable way to timeout on condition variables, so we could implement something.

[seanmonstar]

Link?

On Wed, Apr 1, 2015, 2:13 PM Jonathan Reem notifications@github.com wrote:

There is now going to be a stable way to timeout on condition variables,
so we could implement something.

—
Reply to this email directly or view it on GitHub
#315 (comment).

[reem]

rust-lang/rust#23949

[seanmonstar]

I see. So, to support a timeout using those, I imagine having to kick off 2 threads, one that is parked for X ms, the other that tries to read(), and we use a select! {} on the 2 receivers.

Doesn't sound cheap...

[Wilfred]

If you're not happy using select! in hyper itself, how about adding an example to the README? It took me a little fiddling to get it working, so having an example would have helped me.

[wezm]

Once set_read_timeout is stabilised it can be used to add timeout support, yes?

• http://doc.rust-lang.org/nightly/std/net/struct.TcpStream.html#method.set_read_timeout
• http://doc.rust-lang.org/nightly/std/net/struct.UdpSocket.html#method.set_read_timeout

[seanmonstar]

Yep! I'd like to try to start nowish on an API, and put it behind a Cargo feature timeouts. Then anyone could use it if they're using a nightly rustc, and we can maybe provide feedback after trying real use. I've just been busy on other parts, but PRs always welcome!