Ross A. Baker
Ross A. Baker
Thanks for the reproducible case. This is an excellent report. I sent a PR: https://github.com/slve/http4s-eof/pull/1. I think this clears up by sharing one client for the app instead of creating...
Okay, that was the only misuse I saw reviewing it last night. Going to give this the bug label and try to chase it more this weekend.
By way of comparison, there is a `readTimeout` and a `pooledConnectionIdleTimeout` in AHC. Jetty client has just an `idleTimeout`, and it's not clear which definition it means.
So if we added a `keepAliveTimeout` and made `idleTimeout` behave like blaze-client, it seems we'd be in a good place?
We had an issue at $WORK with a server that used round-robin DNS interacting badly with the JVM's DNS caching. We filled the pool against one cached instance. It was...
No, there's not yet anything like the `keepAliveTimeout` implemented, but I'm sure that would be exposed and configurable once implemented.
I'm calling pending tests bugs.
Perhaps the decoder could be relaxed, but that's an invalid query string. The `%` is a reserved character for percent encoding, and is supposed to be followed by two hex...
The valid characters are alphanumerics and: `/ ? : @ - . _ ~ ! $ & ' ( ) * + , ; =`. Everything else needs to be...
I think that might also be achievable by an appender, but I'm not opposed to a configurable logger if that's deemed a better way.