feat(server): dropping a Response will write out to the underlying stream

[seanmonstar]

No description provided.

[coveralls]

Coverage increased (+1.15%) to 85.45% when pulling 5176922 on drop-res into 7f5e038 on master.

[Ryman]

The drop tests don't seem to actually test the body itself is flushed?

The branching on PhantomData looks like a neat technique!

[reem]

This is inspired.

[reem]

That said we can get better perf by just doing TypeId::of::<W>() == TypeId::of::<StatusFresh>(), which will get the branch optimized away because monomorphization.

[seanmonstar]

@Ryman what do you mean? That flush was called on the MockStream?

[reem]

should lose the _ now that it is used

EDIT: nvm because the suggestion in the other comment makes it unused again

[Ryman]

@seanmonstar I mean that the tests seem to only check with a zero byte body in the response, I wonder if it's a benefit to include a test that actually ensures a body is dealt with appropriately?

[seanmonstar]

@Ryman ah sure. Though, since it's a chunked the response, that last chunk written before flushing must be the 0\r\n\r\n chunk, which the tests are checking for.