If you look in perf_try/api.py:468 the recipe seemingly depends on tracing/bin/parse_metric_cmdline which doesn't exist. Since that always fails, the results step always gets warnings appended although they only show up in the stdout or the posted bisect results:
@@@STEP_CURSOR@Results@@@
@@@STEP_STARTED@@@
@@@STEP_TEXT@No values from test with patch, or none from test without patch.
{
"post_data": {
"bisect_bot": "win_8_perf_bisect",
"buildbot_log_url": "http://build.chromium.org/p/tryserver.chromium.perf/builders/win_8_perf_bisect/builds/2287",
"cloud_link": "",
"command": "src/tools/perf/run_benchmark --browser=release blink_style.top_25 --verbose",
"metric": null,
"status": "completed",
"try_job_id": null,
"warnings": [
"No values from test with patch, or none from test without patch.\n Output with patch: ........."
]
},
"status_code": 200,
"text": "No try job found.\nrequest_id:583fb61800ff0df4f34d2ed5a30001737e6368726f6d65706572660001636c65616e2d6a657373696d622d333065643561373900010106\n"
}
We should look at whether this is still necessary to parse out the values, and if not at least remove the dead code.
Comment 1 by simonhatch@chromium.org
, Dec 15 2016