881615 - [socorro-crashstats] crash number different between Top Crashers and Signature Summary views

Status: VERIFIED FIXED

(Socorro :: Webapp, task)

Description

[Scoobidiver (away)]

The crash number is different between Top Crashers and Signature Summary views.

For instance, js::types::IdToTypeId(int) has 586 crashes in Top Crashers (see https://crash-stats-django.allizom.org/topcrasher/products/Firefox/versions/24.0a1/date_range_type/report/crash_type/browser/os_name/None?days=28) and 601 in Signature Summary when clicking the link in Top Crashers (see https://crash-stats-django.allizom.org/report/list?product=Firefox&range_value=28&range_unit=days&date=2013-06-10&signature=js%3A%3Atypes%3A%3AIdToTypeId%28int%29&version=Firefox%3A24.0a1).

I think it has something to do with the date and time (not 0:00 UTC) when Top Crashers are freeze for the day.

Comment 1

[Robert Helmer [:rhelmer]]

Note that this is also wrong on the PHP app but by a different amount:

https://crash-stats.allizom.org/topcrasher/byversion/Firefox/24.0a1/28/browser
https://crash-stats.allizom.org/report/list?range_value=28&range_unit=days&date=2013-06-10&signature=js%3A%3Atypes%3A%3AIdToTypeId%28int%29&version=Firefox%3A24.0a1

I don't see the issue on production, I wonder if it is isolated to staging...

Comment 2

[Robert Helmer [:rhelmer]]

Hm the django code here is very simple, I see a few odd things:

1) an unused date param is passed to the service; internally it looks at today's UTC date (should it be yesterday's?) for end_date, and subtracts "duration" days for start_date then passes these to the mware.

2) The signature_summary report seems to be listing products other than the one the selected products is for... this might account for the discrepancy

Comment 3

[Scoobidiver (away)]

(In reply to Robert Helmer [:rhelmer] from comment #2)
> 2) The signature_summary report seems to be listing products other than the
> one the selected products is for... this might account for the discrepancy
It's true with Nightly and Nightly UX but it doesn't seem the case here.

Comment 4

[Robert Helmer [:rhelmer]]

(In reply to Scoobidiver from comment #3)
> (In reply to Robert Helmer [:rhelmer] from comment #2)
> > 2) The signature_summary report seems to be listing products other than the
> > one the selected products is for... this might account for the discrepancy
> It's true with Nightly and Nightly UX but it doesn't seem the case here.

I mean for example in prod I see Thunderbird and MetroFirefox showing up in signature summary, when selected product is Firefox:

https://crash-stats.mozilla.com/report/list?range_value=28&range_unit=days&date=2013-06-10&signature=js%3A%3Atypes%3A%3AIdToTypeId%28int%29&version=Firefox%3A24.0a1

Schalk, is this intentional? If so, is the topcrash page expected to have this behavior as well?

Comment 5

[Scoobidiver (away)]

(In reply to Robert Helmer [:rhelmer] from comment #4)
> I mean for example in prod I see Thunderbird and MetroFirefox showing up in
> signature summary, when selected product is Firefox:
It's normal. Only Operating System, Uptime Range, Architecture, Process Type, and Flash Version should be filtered by the product and version of the crash query which is not the case here but a different issue.

Comment 6

[Robert Kaiser]

Rob, it's intentional right now, though I think there might be a bug open on changing that (and make this list of all products/versions only available with a click).

Comment 7

[Robert Helmer [:rhelmer]]

OK I see the problem here and it's data-related, although I think the real underlying issue makes this more confusing - the topcrashers page seems to detect when the underlying TCBS table was actually last updated, and sets the end_date to that.

You'll notice the URL has 2013-06-09 hardcoded in it if you click from https://crash-stats-dev.allizom.org/topcrasher/byversion/Firefox/24.0a1 -

https://crash-stats-dev.allizom.org/report/list?range_value=7&range_unit=days&date=2013-06-09&signature=js%3A%3Atypes%3A%3AIdToTypeId%28int%29&version=Firefox%3A24.0a1

While you can see the date is passed in the django version, the signature_summary view ignores it and always sets end_date to UTC today.

I am going to go ahead and make it honor the date passed by topcrasher page, but I find this behavior a bit confusing/misleading.. however it's not different than the PHP app so let's handle any potential changes in a separate bug.

Comment 8

[Robert Helmer [:rhelmer]]

r? https://github.com/mozilla/socorro-crashstats/pull/374

Comment 9

[[github robot]]

Commits pushed to master at https://github.com/mozilla/socorro-crashstats

https://github.com/mozilla/socorro-crashstats/commit/f301890bf6c4031716387f6b67b5d83eda07fe66
bug 881615 - pass product/version to signature summary, and also honor date

https://github.com/mozilla/socorro-crashstats/commit/bf5ebaa676101cdca69e4f020337a3c695b41fa9
Merge pull request #374 from rhelmer/bug881615-topcrashers-signature-summary-difference

bug 881615 - pass product/version to signature summary, and also honor d...

Comment 10

[[github robot]]

Commits pushed to master at https://github.com/mozilla/socorro-crashstats

https://github.com/mozilla/socorro-crashstats/commit/5b8f44b151887d9ce3635eec85bdfc515dc47422
bug 881615 - update api test to reflect signature summary filtering fix

https://github.com/mozilla/socorro-crashstats/commit/c3e06e7b4e964da20eeb9f171bf99f7e41db9f13
Merge pull request #381 from rhelmer/bug881615-topcrashers-signature-summary-difference

bug 881615 - update api test to reflect signature summary filtering fix

Comment 11

[Scoobidiver (away)]

It's not fixed for the given example in comment 0: 949 crashes in Top Crashers and 2542 crashes in the crash query.

Comment 12

[Scoobidiver (away)]

Verified on stage and prod.