v1.6.0

Bump version number to 1.4.1-async-await.3 (grpc#1303)

Motivation:

We plan on tagging a release soon.

Modifications:

- Bump the version to 1.4.1-async-await.3

Result:

The version in the default user-agent string will match the released
version.

Bump version number to 1.5.0 (grpc#1294)

Motivation:

We plan on tagging a release soon.

Modifications:

- Bump the version to 1.5.0

Result:

The version in the default user-agent string will match the released
version.

Bump version number to 1.4.1-async-await.2 (grpc#1291)

Motivation:

We plan on tagging a release soon.

Modifications:

- Bump the version to 1.4.1-async-await.2

Result:

The version in the default user-agent string will match the released
version.

Bump version number to 1.4.2 (grpc#1280)

Motivation:

We plan on tagging a release soon.

Modifications:

- Bump the version to 1.4.2
- Add a commit template for version bumps, because the commit is always
  the same (unless I'm adding a commit template as well).

Result:

The version in the default user-agent string will match the released
version.

Bump version number to 1.4.1 (grpc#1267)

Motivation:

We plan on tagging a release soon.

Modifications:

- Bump the version to 1.4.1

Result:

The version in the default user-agent string will match the released
version.

[async-await] Code generation for "simple, but safe" wrapper client c…

…alls. (grpc#1261)

This PR adds codegen support for the "simple, but safe wrappers" that were part of the proposal for async/await support, added in grpc#1231.

Codegen has also been re-run for the example Echo service.

Bump version number to 1.4.0 (grpc#1256)

Motivation:

We plan on tagging a release soon.

Modifications:

- Bump the version to 1.4.0

Result:

The version in the default user-agent string will match the released
version.

Bump version number to 1.3.0 (grpc#1222)

Motivation:

We plan on tagging a release soon.

Modifications:

- Bump the version to 1.3.0

Result:

The version in the default user-agent string will match the released
version.

Merge pull request from GHSA-2jx2-qcm4-rf9h

Motivation:

If the server replies early, currently the GRPCWebToHTTPServerCodec
takes it upon itself to suppress the rest of the data coming from the
client.

This design does not match with HTTP/2, where an early return from the
server does not suppress this data. As a result, there is no particular
reason to have implemented this beyond optimisation.

Sadly, this optimisation also opens up a crash vector, because the
codec also suppressed the .end message. When combined with the fact that
grpc-web re-uses the server state machine, we exposed the server to a
crash, because it could see two .heads with no intervening .end.

Modifications:

Stop suppressing reads, continue to police them and pass them through.

Result:

Remove a crash vector, stop errors being fatal.

Co-authored-by: Cory Benfield <[email protected]>