I suspect we are running the wrong browser on linux-chromeos-rel

The test uses the test_chrome binary [1], which is the same binary as the chrome binary [2]. So as long as the builder is using the right configs for chromeos, the test should be picking up the chromeos build of chrome during the test.

[1] Note "../../tools/perf/run_tests -v --browser=exact --browser-executable=./test_chrome ..." in the log in https://chromium-swarm.appspot.com/task?id=3c8bb68810501d10&refresh=10&show_raw=1
[2] https://cs.chromium.org/chromium/src/chrome/BUILD.gn?type=cs&q=test_chrome&sq=package:chromium&l=348

I don't think the bug is about the browser binary, it's about Telemetry picking up a wrong browser back-end for the binary, here: https://cs.chromium.org/chromium/src/tools/perf/core/perf_benchmark.py?l=54

This is a known bug: crbug.com/788442 and crbug.com/799950.

Maybe we should bump up the priority of crbug.com/799950?

+Dirk Pranke
+Ben Pastene

I am not really sure about the status, anyway, from what I could see:


some efforts on Chrome Operations team to "Add ChromeOS VM tests to the CQ" - Issue 840967
and telemetry_perf_unittests for ChromeOS, e.g. https://chromium-swarm.appspot.com/task?id=3e2ed68eb37f6410&refresh=10&show_raw=1

Some bots are using the pool 'Chrome-CrOS-VM' - 
https://cs.chromium.org/search/?q=Chrome-CrOS-VM&sq=package:chromium&type=cs

To be clear, the problem reported originally has to do with platform-detection on the linux-chromeos-rel bot, correct? That bot is a bit unique in that it builds desktop-linux binaries that executes chromeos code-paths. I'm not surprised that there's OS mixup in the test.

However, telemetry_perf_unittests was removed from the bot a few months ago in  bug 836447 . So this shouldn't be a problem anymore. I'm currently working on reenabling the suite on a different chromeos bot, one that runs the test in a true chromeos environment (which won't be the linux-chromeos build that prompted this bug). I'm going to dupe this bug into my efforts there.