524507 - need a variety of reports to get at the long tail of crash reports

Status: RESOLVED INCOMPLETE

(Socorro :: General, task)

Description

[chris hofmann]

maybe start with this being a brainstorming bug, then spin off individual good ideas into reports that can be built.

the basic problem is that we currently focus mostly on analysis of "top 100" crashes and reports and queries that get at those stack signature.  these are crashes that happen around more than 150 times a day.

we have a very long tail of crashes that happened that we don't have much visibility into.

I'll add an update that describes the problem in more detail soom.

Comment 1

[chris hofmann]

Attachment: profile of the crash data long tail.

as mentioned above we only studying the top 20-100 list of crashers for major releases which means crashes that happen around 150 times a day or more.

the long tail for stack signatures stretches well beyond that.  there are about

19763 unique signatures per day on 3.5.3
and 
11114 unique signatures per day on 3.0.14

the tools would help us to understand the nature of this long tail and how it differs among releases.

we also have single stack signatures that translate into many actual bugs, and many signatures that could be fixed by a single bug fix.  we need tools that help disect and make the correlations for these kinds of signatures.

Comment 2

[chris hofmann]

one tool might make a pool of crashes that have 'time since last crash less than two minutes, then produce a top 100 signatures of start up related bugs.

I'll try to spin this off into a separate bug with some example analysis and an example report soon.

Comment 3

[chris hofmann]

https://bugzilla.mozilla.org/show_bug.cgi?id=470827 is an example of a report that would help understanding of the long tail

Comment 4

[chris hofmann]

[Bug 518287] twice as many "no signature" and other crashes in firefox 3.5.3 as 3.0.14  is another blackhole area that is part of understanding this problem

Comment 5

[Laura Thomson :laura]

Closing in favor of more specific bugs.