There is a problem impacting emulator WebAPI tests randomly that results in sporadic failures. This happens frequently enough that it negatively impacts the ability of the sheriffs to deal with these tests, and we'd like help in resolving it.

The bug is https://bugzilla.mozilla.org/show_bug.cgi?id=898074 - Frequent "raise TimeoutException(message='socket.timeout', status=ErrorCodes.TIMEOUT, stacktrace=None)" during WebAPI tests

The symptoms of the problem are that randomly, data that is written to Marionette's socket by the client fails to cause the remote debugger's onDataAvailable callback to be fired (http://mxr.mozilla.org/mozilla-central/source/toolkit/devtools/server/transport.js#138). Once this happpens, the client can write to the socket multiple times without error, and none of the writes will trigger onDataAvailable, until the client closes the socket. At that point, onDataAvailable is called a final time, and all the data that's been written to the client is read.

This problem has never been observed with desktop Firefox, with the b2g desktop build, or with device builds, under similar circumstances. dhylands and I have spent time eliminating Marionette and the remote debugger as potential causes of the problem. Because we've not seen it except with the emulator, we think it likely that the problem lies somewhere with it, but it's also possible that the problem lies somewhere in gecko, or in the interface between the two.

If you're interested in helping make FirefoxOS testing more stable, and have some time or experience to delve into the emulator's and/or gecko's networking layer, we'd love your help!

Feel free to ping jgriffin on #ateam for more details.

Thanks,

Jonathan

_______________________________________________
dev-b2g mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-b2g

Reply via email to