965481 - Generate topcrasher reports for B2G

Status: RESOLVED FIXED

(Socorro :: Database, task)

Description

[Robert Helmer [:rhelmer]]

ADI for B2G is not on the horizon, let's make it possible to generate topcrasher reports for Firefox OS anyway.

Comment 1

[Robert Kaiser]

Yes, we should do this.

That said, to get meaningful FxOS/B2G reports, topcrasher or otherwise, we need bug 928051 as well.

Comment 2

[Laura Thomson :laura]

Bug 928051 is also on the list of things needed for the quarterly goal of "fix b2g reporting", along with this one.

Comment 3

[Robert Helmer [:rhelmer]]

Hm so why are we getting some top crash reports at all?

https://crash-stats.mozilla.com/topcrasher/products/B2G/versions/18.0b?days=7
https://crash-stats.mozilla.com/topcrasher/products/B2G/versions/29.0a1

Very few, but still.

Where is topcrash dependent upon ADI right now? The most obvious problem I see right now is that incoming reports are not consistent with the data we get from ftpscraper, so things aren't getting into reports_clean and therefore not into topcrasher (there are some small number that do match up, which is why I believe we are getting the few topcrash results)

If nobody is sure I can answer my own question, just hoping somebody knows :) I think as a first step I will try loosening up the population of product_versions for b2g by loosening up the dependency on releases_raw which is populated by ftpscraper, and see if it starts working.

Comment 4

[Robert Helmer [:rhelmer]]

(In reply to Robert Helmer [:rhelmer] from comment #3)
> Hm so why are we getting some top crash reports at all?
> 
> https://crash-stats.mozilla.com/topcrasher/products/B2G/versions/18.0b?days=7
> https://crash-stats.mozilla.com/topcrasher/products/B2G/versions/29.0a1
> 
> Very few, but still.
> 
> Where is topcrash dependent upon ADI right now? The most obvious problem I
> see right now is that incoming reports are not consistent with the data we
> get from ftpscraper, so things aren't getting into reports_clean and
> therefore not into topcrasher (there are some small number that do match up,
> which is why I believe we are getting the few topcrash results)
> 
> If nobody is sure I can answer my own question, just hoping somebody knows
> :) I think as a first step I will try loosening up the population of
> product_versions for b2g by loosening up the dependency on releases_raw
> which is populated by ftpscraper, and see if it starts working.

I think bug 928051 goes much further in loosening up the restrictions - going to do something simpler and more targeted to test this particular bug.

Comment 5

[Robert Kaiser]

(In reply to Robert Helmer [:rhelmer] from comment #3)
> Hm so why are we getting some top crash reports at all?

Hah, I wasn't even aware that any would make it at this point. Perhaps we made them work already with some other changes - IIRC we were just not running the matview calculation when ADI were not present. We of course get no data for /daily or the other graphs, but it looks like we'd calculate TCBS which is good.

Of course we need bug 928051 as mentioned before we get really decent data there.

Comment 6

[Robert Helmer [:rhelmer]]

So with the work in bug 928051 so far, I am able to get topcrasher results to appear using local "fakedata" set - I think we're correct here in that ADI isn't being used for the current topcrasher page, as long as I'm not missing something in my assumptions about the data.

Comment 7

[Robert Helmer [:rhelmer]]

topcrasher reports work now for B2G 18.0 (which is where the ZTE "roamer2" crashes are going, for the set of releases we've mapped). We didn't need to do anything with ADI \o/