bryancall commented on PR #13873: URL: https://github.com/apache/trafficserver/pull/13873#issuecomment-6306707704
#13545 moved this test to Uranium, so I merged master and moved the fix there. `tests/uranium_tests/h2/h2active_timeout.py` gets the same read-until-close client as before: it times from the preface and prints CONNECTION_TIMEOUT only when the close lands between the configured timeout and 15 seconds. `tests/uranium_tests/h2/test_http2.py` now passes the timeout it configures (`ACTIVE_TIMEOUT = 3`) instead of a separate `"4"`, and the step timeout goes from 10 to 30 seconds so it outlasts the client's wait. I could not make the old failure happen under Uranium: master passed 320 of 320 runs with `inactivity_check_frequency` at 2 and every core busy. Over 168 runs, the close landed between 3.55 and 3.90 seconds, so master's fixed 4 second sleep has as little as 0.1 seconds to spare. The master AuTest failure in the description had the client exiting after 4.02 seconds. This change removes that margin rather than fixing a failure I can reproduce today. The new check still fails when it should: 4 of 4 with the timeout disabled, and 4 of 4 with the timeout at 1 second while the client expects 3. Could you take another look? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
