Compile without fatal warnings on 2.13.0

[rossabaker]

• Reimplement the case classiness of Request by hand to work around False positive unreachable code warning with sealed abstract class scala/bug#11457.
• Work around Seq aliasing in a spec
• Eliminate early initializer in TomcatSpec, which no longer appears to be necessary

[rossabaker]

If we gave Request this treatment, we might consider doing the same to Response for symmetry, though I don't think there were any binary changes here, so then again, maybe not. Discuss.

[rossabaker]

I'm not a huge fan of these interfaces, but:

• Product makes hash code easier and was there before.
• Spark folks may be depending on Serializable. I don't judge.

[rossabaker]

That F is erased and I'm surprised this didn't generate a warning but _ didn't work. Maybe _[_]? Ugh.

[aeons]

The unapply on Request seems to fail < 2.13

[hamnis]

Maybe this would help: https://github.com/alexarchambault/data-class

[rossabaker]

I just mentioned data-class on another ticket. 😄 It's can also be a compile-time only dependency, which makes it more attractive for our use case. I'm running out of time for the night, but I'll give that a poke tomorrow.

[hamnis]

I tried doing this now, and it failed with conflict on withXXX

[hamnis]

Seems like the only change needed for this to work is changing the pattern match for org.http4s.client.middleware.Logger and org.http4s.server.middleware.Logger

def prelude = message match {
      case r: Request[F] =>
        s"${r.httpVersion} ${r.method} ${r.uri}"

      case Response(status, httpVersion, _, _, _) =>
        s"$httpVersion $status"
}

[hamnis]

@rossabaker Should I create a new PR to get this moving again?

[rossabaker]

@hamnis That would be wonderful. Thank you!

[rossabaker]

Superseded by #2933.