Currently (see bug 1346068 comment 14) agent tests are disabled in the browser because the test cases are not suitable for the browser and the $262.agent framework is insufficient for dealing with a browser. Work items: - fix $262.agent framework upstream (tc39/test262#928) - fix the test cases upstream (ditto) - fix our implementation of $262.agent in js/src/tests/test262-host.js and js/src/tests/test262/shell.js - remove the opt-outs in js/src/tests/jstests.list
Created attachment 8851569 [details] [diff] [review] bug1349863-agent-for-browser.patch This has a new harness with a new API and rewritten test cases. It is *roughly* right and it passes locally when run in a browser with a simplified context, but fails on try, see the links in previous comments. The error is invariably that 'no output' is produced for the test cases. I think that what's most likely happening is that when $262.agent.start() returns to the event loop, as it must for the workers to be created properly, and also when it tries to wait for results from the workers, then the test runner may conclude that the test is done (because the script returned). But the only way to determine that is either to see test output, or to time out.
Assignee: nobody → lhansen
You need to log in before you can comment on or make changes to this bug.