Re-enable HTTP and Redis tests

[itowlson]

@fibonacci1729 Right now, this fails because the test response body appears empty (when it is expected to be a "Hello world" message). I can't see what I need to change though - my guess is that something needs to flush or wait longer or something, but I thought collect() would do that for me. Would you mind giving it a once-over to see where my understanding has gone awry please? Thanks (and sorry for loading this on your plate)!

Re-enables HTTP and Redis tests. Thank you Brian and Joel for your help!

[fibonacci1729]

I think the root cause is that into_http() and body collection are happening outside run_concurrent(). It seems like the component model's async stream-based body delivery only works while the store is actively running concurrent code. Once run_concurrent returns, the internal stream channels stop being driven, so the body appears empty.

This seems to work:

    let (status, resp_body) = store
        .run_concurrent(async |store| {
            let (res, ()) = tokio::try_join!(
                async {
                    let res = http_service.handle(store, request).await??;
                    let res = store.with(|store| res.into_http(store, async { Ok(()) }))?;
                    let status = res.status();
                    let body = res.into_body().collect().await?;
                    anyhow::Ok((status, body.to_bytes()))
                },
                async {
                    io.await.map_err(|e| anyhow::anyhow!("{e:?}"))
                }
            )?;
            anyhow::Ok(res)
        })
        .await??;

    assert!(status.is_success());
    assert_eq!(resp_body.as_ref(), b"Hello, world!");

[dicej]

Right, as discussed in spinframework/spin#3454, the AsyncFn passed to Store::run_concurrent must not return until it has everything it needs from the guest, including any stream content to be written by the guest after it returns a value. Once that AsyncFn returns, Store::run_concurrent will return, and no guest code can run until/unless Store::run_concurrent is run again.

[itowlson]

Oh that might affect the Redis trigger then - when I did that I just called run_concurrent - it sounds like that would have a bug if a Redis guest spawned a task.

[itowlson]

Thanks all - I appreciate both the explanation for understanding, and the helpful fix!