915209 - [prod][stage] Firefox 23.0.1 queries not returning results for /topcrasher/

Status: RESOLVED INVALID

(Socorro :: General, task)

Description

[Matt Brandt [:mbrandt]]

I'm not sure of the width of the issue but I'm seeing a few instances on prod where I would expect results to be return am seeing none.

Firefox 23.0.1
Returns no results:
https://crash-stats.mozilla.com/topcrasher/products/Firefox/versions/23.0.1/date_range_type/build/crash_type/all/os_name/None?days=28

Returns results:
https://crash-stats.mozilla.com/query/?product=Firefox&version=Firefox%3A23.0.1&range_value=1&range_unit=weeks&date=09%2F11%2F2013+15%3A00%3A00&query_search=signature&query_type=contains&query=&reason=&release_channels=&build_id=&process_type=any&hang_type=any

Firefox 24.0b
https://crash-stats.mozilla.com/topcrasher/products/Firefox/versions/24.0b/date_range_type/build?days=7
EMPTY: no crashing thread identified; corrupt dump shows 20586 results

Returns no results
https://crash-stats.mozilla.com/report/list?product=Firefox&version=Firefox:24.0b&query_search=signature&query_type=contains&reason_type=contains&date=2013-09-11&range_value=28&range_unit=days&hang_type=any&process_type=any&signature=EMPTY%3A+no+crashing+thread+identified%3B+corrupt+dump

Comment 2

[Matt Brandt [:mbrandt]]

Thanks Kairo - my mistake, I duped my own bug for the 24.0b problem

Steps to reproduce for the Firefox 23.0.1 bug:
1. Returns no results - https://crash-stats.mozilla.com/topcrasher/products/Firefox/versions/23.0.1/date_range_type/build/crash_type/all/os_name/None?days=28

2. Returns results - https://crash-stats.mozilla.com/query/?product=Firefox&version=Firefox%3A23.0.1&range_value=1&range_unit=weeks&date=09%2F11%2F2013+15%3A00%3A00&query_search=signature&query_type=contains&query=&reason=&release_channels=&build_id=&process_type=any&hang_type=any

Expected:
/topcrasher/products/Firefox/versions/23.0.1 should return results

Comment 3

[Lonnen :lonnen]

If this is already happening on prod, will the staged code make it worse? If not, I do not think we should block for this. Even if its important -- when we get a fix for it we can do another release.

Comment 4

[Matt Brandt [:mbrandt]]

(In reply to Chris Lonnen :lonnen from comment #3)
> If this is already happening on prod, will the staged code make it worse? If
> not, I do not think we should block for this. Even if its important -- when
> we get a fix for it we can do another release.

+1 .. my main concern was/is trying to understand what changed in the env and to get it sorted out.

Comment 5

[Erik Rose [:erik][:erikrose]]

Matt,

(1) returns no results because it's a by-build-date query (rather than a by-crash-date one) against a released version of Firefox (rather than a nightly or what-have-you wherein build dates have a discriminating meaning). You can tell by the "/build/" segment in the URL. Kairo says this is expected behavior, though it wasn't obvious to me either.

(2) is an advanced search for crashes by crash date. Since it's by crash date, it works, regardless of whether you're querying against a released version of Firefox or not.

So I don't think this is a real "bug" as such, though I could certainly see morphing it into an enhancement to make things work more obviously. I bet the priority will be low, however, since it apparently doesn't confuse our actual users. Let's chat tomorrow morning, and we can decide whether to "invalid" it. Cheers!

Comment 6

[Robert Kaiser]

I think we should tell in the UI in a better way that we don't have a by-build-date report for that version if we are calling up such a combination, and that is probably worth a bug.

Otherwise, comment #5 is spot-on.

Comment 7

[Stephen Donner [:stephend] Not actively reading bugmail]

(In reply to Erik Rose [:erik][:erikrose] from comment #5)
> Matt,
> 
> (1) returns no results because it's a by-build-date query (rather than a
> by-crash-date one) against a released version of Firefox (rather than a
> nightly or what-have-you wherein build dates have a discriminating meaning).
> You can tell by the "/build/" segment in the URL. Kairo says this is
> expected behavior, though it wasn't obvious to me either.

STR:

1. Load https://crash-stats.mozilla.com/home/products/Firefox
2. Choose "23.0.1" from the "Current Versions" pulldown/select
3. Click on "Top Crashers," near the bottom of the page

Actual Results:

If you have the Web Console open when loading that (# of days in the parameter doesn't matter - 7 or 28, the failure, below, is the same), you'll see the Web app issue GETs like this:

[17:51:18.337] GET https://crash-stats.mozilla.com/correlation/signatures?correlation_report_type=core-counts&product=Firefox&version=23.0.1&platforms= [HTTP/1.1 400 BAD REQUEST 104ms]
[17:51:18.339] GET https://crash-stats.mozilla.com/correlation/signatures?correlation_report_type=interesting-addons&product=Firefox&version=23.0.1&platforms= [HTTP/1.1 400 BAD REQUEST 91ms]
[17:51:18.341] GET https://crash-stats.mozilla.com/correlation/signatures?correlation_report_type=interesting-modules&product=Firefox&version=23.0.1&platforms= [HTTP/1.1 400 BAD REQUEST 99ms]

Comment 8

[Matt Brandt [:mbrandt]]

Thank you :ErikRose, let's do chat tomorrow about the nuances - I do see what you what you're saying regarding the by-build-date vs. by-crash-date queries.

The 400s that stephend is seeing (comment 7) may be worth a new bug?

Comment 9

[Robert Kaiser]

(In reply to Stephen Donner [:stephend] from comment #7)
> STR:
> 
> 1. Load https://crash-stats.mozilla.com/home/products/Firefox
> 2. Choose "23.0.1" from the "Current Versions" pulldown/select
> 3. Click on "Top Crashers," near the bottom of the page

Those STR only work if you have selected "by build date" before. If you have selected "by crash date", then there is no bug and everything works fine.

The only issue is that there is no "by build date" report for 23.0.1 because it's a release. As said before, that's correct behavior, we should probably have a more helpful message on the page though when you get there.

Comment 10

[Erik Rose [:erik][:erikrose]]

We'll add a message to the empty page, saying something like 'Crashes by build date are unavailable for this version. Please switch to "By Crash Date" to see data for this version.'

Comment 11

[Peter Bengtsson [:peterbe]]

Sorry for sounding lazy, but can you check that this is still reproducible?

Comment 12

[Matt Brandt [:mbrandt]]

(In reply to Peter Bengtsson [:peterbe] from comment #11)
> Sorry for sounding lazy, but can you check that this is still reproducible?

Per comment 9 we can close this.