reqwest::blocking::ClientBuilder does not have a function tcp_keepalive()

[thegatsbylofiexperience]

HI,

On the current version 0.10.9 I have a windows baset reqwest client connecting is return a weird error, which it was not in v0.10.8. I have setup the client to be reused on multiple requests to the server.

thread '' panicked at 'called Result::unwrap() on an Err value: reqwest::Error { kind: Request, url: Url { scheme: "http", host: Some(Ipv4(127.0.0.1)), port: Some(1234), path: "/some_path", query: None, fragment: None }, source: hyper::Error(Connect, ConnectError("tcp set_keepalive error", Os { code: 0, kind: Other, message: "Success." })) }'

I am trying to use the blocking client builder to add the tcp keepalive timeout. However it does not have one available.

From the change log it seems like it was an added feature in 0.10.9.

I am happy to help with a PR if you want me to have a go at adding this to reqwest. Please let me know.

Thanks.

[seanmonstar]

It looks like the option wasn't added to the blocking::ClientBuilder. If you want to add it in a PR, that'd be welcome :)

[rneswold]

I'm having a similar problem on NetBSD 9. I'm using the huebridge crate. When I specify reqwest 0.10.8, I can control the Philips lights. When I upgraded to 0.10.9, I get this:

HueError(Reqwest(reqwest::Error { kind: Request,
                                  url: Url { scheme: "http",
                                             host: Some(Ipv4(192.168.1.4)),
                                             port: None,
                                             path: "/api/.../lights/8/state",
                                             query: None,
                                             fragment: None },
                                    source: hyper::Error(Connect,
                                                         ConnectError("tcp set_keepalive error",
                                                                      Os { code: 42,
                                                                           kind: Other,
                                                                           message: "Protocol option not available" })) }),
                  State { next_error: None,
                           backtrace: InternalBacktrace { backtrace: None } })

[Martichou]

@rneswold can we see an example of your code ?
Have you tried using HEAD instead of 0.10.9?

[rneswold]

reqwest is getting used indirectly by huebridge. So I have stuff like:

if let Err(e) = bridge.set_light_state(light, &state) {
    warn!("light {} : error {:?} sending state {:?}", light, e, state)
}

huebridge hasn't changed -- I've been using 0.3.0 for months. Here's where it's defined:

https://github.com/NikoWoot/huebridge/blob/a3be3a2b851df97003e9fb3157a69c2ff08b789f/src/bridge.rs#L118

I haven't tried a non-released version. Is there an easy way to do that with cargo?

[seanmonstar]

Sorry about the breakage. The problem comes from reqwest defaulting to setting TCP_KEEPALIVE starting in release 0.10.9, and hyper's HttpConnector fails the connection if an option is unavailable. So, here's my plan:

• Release a new version of reqwest disabling TCP_KEEPALIVE by default.
• Get hyper to stop failing if setting an error fails, and instead just log a warning.
• Once it's fixed in hyper, we can re-enable TCP_KEEPALIVE by default.

[seanmonstar]

Reverting the default is in a PR here: #1113

The actual title of this issue is related to a missing method, which was already fixed on master, so I'm going to close here.

[rneswold]

Sorry about the breakage.

Not a problem. Thanks for being so responsive.