lab infra: sku labeling should be robust against preproduction HWIDs |
|||||
Issue descriptionre: crbug.com/853224 It seems that the sku labeling software used in the lab will crash if dummy values are used for preproduction HWID (which is allowed in EVT) This software should produce a "sku" label for any valid HWID input, including EVT HWIDs. Possible approaches are to fallback to probing for unspecified components or simply handle "unknown" fields. Not sure what component/assignee this should be, set to jrbarnette@ to reassign.
,
Jun 19 2018
> It seems that the sku labeling software used in the lab > will crash if dummy values are used for preproduction HWID > (which is allowed in EVT) Nothing crashes. But if the input HWID database information is garbage, the output label will be garbage, too. And Autotest doesn't care if the label is garbage, but the performance team *does* care. Hence the discussion to date. Dropping ownership: CrOS Infra doesn't care about this at all, _except_ that we get caught in the middle when this stuff goes wrong, so we want to be sure that whatever process changes get applied to address this, they will be good enough to keep the problem out of our hair for at least a couple of device cycles.
,
Jun 19 2018
who on the performance team cares? Can you assign this to that person, or cc them?
,
Jun 19 2018
> who on the performance team cares? Can you assign this to that person, or cc them? I assume the best contact would be cywang@, since he was the OP for bug 849908.
,
Jun 19 2018
,
Oct 12
A long term plan is needed, no work will be planned until a decision is made. |
|||||
►
Sign in to add a comment |
|||||
Comment 1 by haddowk@chromium.org
, Jun 19 2018