Release Summary:

Resolved issues:

Description of changes:

The daily scheduled CI run detected flakiness in udp::fuzz_test. I changed the test in my forked repo with the failed input. The test would run successfully locally in my EC2 instance, but failed in GithubAction. The changed code can be found in this snippet. I believe that the reason for such failure in GithubAction is because the EC2 instance runs faster than the one that GithubAction is ran on. I then increase the test duration from 120 seconds to 300 seconds, and the test succeeded:

test stream::tests::request_response::udp::fuzz_test has been running for over 60 seconds
...
test stream::tests::request_response::udp::fuzz_test ... ok
test result: ok. 327 passed; 0 failed; 2 ignored; 0 measured; 0 filtered out; finished in 253.08s

I then run the same test but update the test duration to 180 seconds, and the test succeeded as well. Hence, I conclude that the udp::fuzz_test flakiness is due to the test duration. Bolero sometimes generate a client and server that needs a long time to run the test, which is causing the flakiness. Hence, increase the test duration will mitigate that.

Call-outs:

• I intentionally choose 180 seconds over 300 seconds because we should increase the duration by the smallest amount which will make the test works. If that is proven to be not enough, then we can increase it again.
• I also fixed a typo in the comment that I detected.

Testing:

Already mentioned the test method in the section of Description of changes.

Analysis:

The failed input looks like:

(
    Client {
        delays: Delays {
            read: 1.058095ms,
            write: 1.560203ms,
            shutdown_write: 949.615µs,
            shutdown_read: 1.532273ms,
            drop: 1.61057ms,
        },
        count: 3,
        concurrency: 2,
        max_read_len: 29444,
        max_mtu: Some(
            3153,
        ),
    },
    Server {
        delays: Delays {
            read: 569.478µs,
            write: 1.385789ms,
            shutdown_write: 1.391925ms,
            shutdown_read: 746.348µs,
            drop: 800.292µs,
        },
        count: 4,
        max_read_len: 1,
        max_mtu: Some(
            2252,
        ),
    },
    [
        Request {
            count: 5,
            request_size: 98312,
            response_size: 95544,
        },
    ],
)

The server's max_read_len is set to 1 due to fuzz test randomness, which makes the server to read extremely slow. I believe that's why this test is taking a long time.

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.