IIRC, wptserve runs on a different port, so we might need to set up another port forward from the device if we never did that before ... jbudorick, does that sound plausible?
Chatted with Stefan (thanks for filing the bug!)
Indeed WPT is not properly set up for Android. All the WPT testharness baselines on Android are layout dumps instead of harness output.
Looks like the key is in comment 5 and 6: add the two ports used by wptserve to android's port forwarding list. I'll give a try and hope it's as simple as that. Thanks so much for the help, John and Dirk!
Hi Dirk, looks like after fixing this issue, the current 1h55m timeout for Android builders isn't quite enough as WPT are now actually running (rather than dumping layout right after document is loaded).
My CL: https://crrev.com/c/777932
and an android_blink_rel build: https://build.chromium.org/p/tryserver.chromium.android/builders/android_blink_rel/builds/4716
My hunch is that 2h30m (9000s) might be enough, but I'm not sure.
Is it OK to increase the builder timeout? Or do we need to cut down SmokeTests (layout tests run on Android)?
Ahh, after a closer look, the two failed builds look exactly the same as described in issue 784303. The webkit_patch step actually only took 20min (enabling WPT does increase the execution time, just a bit), but the bot just hung there until it finally timed out. Should've looked more carefully. Thanks a lot for the pointer, Dirk!
Android try bots are currently flakily timing out (purple) due to issue 784303, making it very difficult to correctly test the fix and rebaseline the web platform tests that are enabled on Android. It's not impossible, but even if we land the fix, it wouldn't make much difference in practice if issue 784303 is unresolved.
Marking as blocked.
Comment 1 by foolip@chromium.org
, Nov 17 2017