)

Co-authored-by: Michael Graeb <[email protected]>

…l regression. (#486)

Co-authored-by: Bret Ambrose <[email protected]>

Co-authored-by: Michael Graeb <[email protected]>

- Remove the word "test" from any test name
- Add the word "decoder" to decoder tests
- Lots of fixup in h1_encoder tests:
    - Rename some tests to make more sense
    - Always assert that aws_h1_encoder_message_init_from_request() succeeds or fails. Previously a lot of tests weren't ACTUALLY testing anything!!! Fortunately no actual bugs were uncovered here.
    - Use the "bad request" helper in more places
    - Add "Host" header to all requests

Some cleanup before embarking on major plumbing...

While developing HTTP/1, we stumbled into the pattern of grouping variables under `thread_data` or `synced_data`, so it was more explicit how they should be accessed. But we never moved ALL the variables into the proper groups. We had a TODO about this. Now it is TODONE.

Also: Fix some flaky integration tests by accepting any "2xx" status-code, not just literal "200. I observed these tests failing when they sometimes got 202, instead of 200: h2_sm_acquire_stream & h2_sm_acquire_stream_multiple_connections. But I applied the same fix to any other test I found that strictly checked for 200.

**Issue:**

aws-sdk-cpp needs to send streaming requests of unknown content length. This isn't currently supported in our HTTP/1 client. (our HTTP/2 client *does* already support this)

**Background:**

The HTTP/1 client already supported these use cases:
- Send streaming request body, but declare Content-Length up front
- Send chunked request body, but user must declare length of each chunk via `write_chunk()` API.

But it wasn't possible to send a stream of unknown length without doing extra copies (read to intermediate buffer, then send that buffer via write_chunk() API).

**Description of changes:**

Support input streams of unknown length via chunked encoding.

When reading from the stream, leave a bit of space at the start and end of the aws_io_message buffer for the chunk prefix and suffix. Then go back and write in the prefix and suffix afterwards. No unnecessary copies are involved.

Co-authored-by: Michael Graeb <[email protected]>

Co-authored-by: Michael Graeb <[email protected]>