Closed mcomella closed 4 years ago
We recently updated the parsing code assuming fenix-nightly is the only variant that's working (b/c of https://github.com/mozilla-mobile/perf-frontend-issues/issues/137) – I wonder if old results were reinterpreted with this change.
Triage: there's too much changing with how we measure things right now so we don't care too much: closing.
Screenshot of Samsung S10 results:
The discrepancy is worrying: is our methodology correct?
The old Pixel 2 results of 0.55 ms better match my local P2 than the current dashboard times of 0.43 ms.
Then again, we can just switch to mach perf-test and maybe this won't continue to be an issue.