Issue metadata
Sign in to add a comment
|
Regression range on Seriff-o-Matic sometimes do not match with the actual range |
||||||||||||||||||||||
Issue descriptionExample: https://sheriff-o-matic.appspot.com/chromium.perf/examine/chromium.perf.Android%20Nexus5X%20Perf.blink_perf.layout. In Sheriff-o-Matic, the blink_perf.layout regression range is shown as: chromium: 490956 - 495228 Then I look at the 1st bot "Android Nexts5X Perf" and the 1st run failing blink_perf.layout is #444: https://uberchromegw.corp.google.com/i/chromium.perf/builders/Android%20Nexus5X%20Perf/builds/444 And #444's range is: 496768 - 497032 Notice that those 2 do not overlap. I think it is very important to have correct range shown on sheriff-o-matic. As a sheriff myself, I often trust the range shown there and do not actually double check inside the 1st failing run. I noticed this by accident because I was checking a different failing test in similar range. A mismatch would result in a bisect failure, which results in more questions about which part went wrong and much more work. So I am giving this pri-1. Re-adjust if needed.
,
Aug 24 2017
Hmm, while double checking the links, I am not sure why clicking on the example link does not go to the examine page directly. Filed a separate bug for that ( issue 758764 ).
,
Aug 29 2017
This unfortunately has been happening for a while. We definitely need to up the priority on this though. |
|||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||
Comment 1 by zh...@chromium.org
, Aug 24 2017