679182 - Querying for 6.0 signatures shows data but clicking on the links results in data not found (Search needs to use the release channel)

Status: VERIFIED FIXED

(Socorro :: General, task)

Description

[Marcia Knous [:marcia]]

STR:

1. Load a query searching for RDFContentSinkImpl::OpenObject(unsigned short const*, unsigned short const**) - http://tinyurl.com/3zpkjz6. It finds 157 crashes. 

2. Click on the crash signature link. I get the following on the page:

Results within 2 weeks of 08/15/2011 15:46:05, where the crash signature contains 'RDFContentSinkImpl::OpenObject(unsigned short const*, unsigned short const**)', and the product is one of Firefox, and the version is one of Firefox:6.0 and the crashing process was of any type.

There were no reports in the time period specified. Please choose a different time period or use advanced search to select a custom time period.

3. If I search for it under all versions - http://tinyurl.com/4xatqpf - it finds the signatures and I can get to the reports.

Do we need to remove 6.0 from the list now that there are entries for all the betas (if I query for all the betas it does not find any data for that signature)

Comment 1

[Laura Thomson :laura]

This is a bug (regression) in search as it's passing the wrong parameters to report/list.  

The search is passing in version=Firefox:6.0, and it should pass version_string=6.0 (I think, please confirm) e.g.
https://crash-stats.mozilla.com/report/list?product=Firefox&version_string=6.0&query_search=signature&query_type=exact&query=RDFContentSinkImpl%3A%3AOpenObject%28unsigned%20short%20const*%2C%20unsigned%20short%20const**%29&reason_type=contains&date=08%2F15%2F2011%2015%3A46%3A05&range_value=1&range_unit=weeks&hang_type=any&process_type=any&do_query=1&admin=1&signature=RDFContentSinkImpl%3A%3AOpenObject%28unsigned%20short%20const*%2C%20unsigned%20short%20const**%29

Comment 2

[Robert Helmer [:rhelmer]]

(In reply to Laura Thomson :laura from comment #1)
> This is a bug (regression) in search as it's passing the wrong parameters to
> report/list.  

The product:version thing is fine; the above just removes the version restriction (version_string is silently ignored).

I think the problem here is that search should be restricting to release channel, but it is not. We're looking into it now.

In the new world, that search is supposed to look for Firefox 6.0 on the release channel. Has Firefox 6 been put on the release channel yet? I'm not really sure why we're even showing it in the menu if not...

Comment 3

[Robert Helmer [:rhelmer]]

Adrian and I went over the logic for searching "new" style reports, he is on it.

Comment 4

[Robert Helmer [:rhelmer]]

Also just to clarify - there are no reports in the reports table for this signature for Firefox 6.0 release for the specified time period. The bug is that search is not restricting correctly, so it's showing betas which report themselves as "6.0" now (which is the whole point of the "newtcbs" and 2.2 release).

Comment 5

[Laura Thomson :laura]

(In reply to Robert Helmer [:rhelmer] from comment #4)
> Also just to clarify - there are no reports in the reports table for this
> signature for Firefox 6.0 release for the specified time period. The bug is
> that search is not restricting correctly, so it's showing betas which report
> themselves as "6.0" now (which is the whole point of the "newtcbs" and 2.2
> release).

That's good and makes me feel more comfortable.  Sorry for the flaky initial report, too.

Comment 6

[[DEACTIVATED] Adrian Gaudebert]

Attachment: Search now uses the release_channel

Use the release_channel field in search, depending on the product:version values. And removed some duplicate code.

Comment 7

[Lonnen :lonnen]

Comment on attachment 553357 [details] [diff] [review]
Search now uses the release_channel

ran it, no new errors and a seemingly appropriate number of results.

Comment 8

[[DEACTIVATED] Adrian Gaudebert]

Last patch landed on trunk (r3453: http://code.google.com/p/socorro/source/detail?r=3453 ) and branch 2.2 (r3452: http://code.google.com/p/socorro/source/detail?r=3452 ).

Comment 9

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

1) https://crash-stats.allizom.org/query/query?product=Firefox&version=Firefox%3A6.0&range_value=1&range_unit=weeks&date=08%2F15%2F2011+15%3A46%3A05&query_search=signature&query_type=exact&query=RDFContentSinkImpl%3A%3AOpenObject%28unsigned+short+const*%2C+unsigned+short+const**%29&reason=&build_id=&process_type=any&hang_type=any&do_query=1 no longer finds the 6.0beta builds 
2) Used PowerFuzzer, and /query is fine -- no new issues there

What else should I verify?

Comment 10

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

3) We think that https://crash-stats.allizom.org/query/query?product=Firefox&version=Firefox%3A6.0%28beta%29&version=Firefox%3A6.0b5&version=Firefox%3A6.0b4&version=Firefox%3A6.0b3&version=Firefox%3A6.0b2&version=Firefox%3A6.0b1&version=Firefox%3A6.0a2&range_value=1&range_unit=weeks&date=08%2F15%2F2011+15%3A46%3A05&query_search=signature&query_type=exact&query=RDFContentSinkImpl%3A%3AOpenObject%28unsigned+short+const*%2C+unsigned+short+const**%29&reason=&build_id=&process_type=any&hang_type=any&do_query=1 (which selects *all* non-shipping Firefox 6.0 builds on all platforms, *should work*, but it's not, currently.

Comment 11

[[DEACTIVATED] Adrian Gaudebert]

Attachment: Search now uses the release_channel

Fixed the SQL, fixed some problems in the algorithm. Tested on khan and it return results now.

Comment 12

[[DEACTIVATED] Adrian Gaudebert]

Attachment: Search now uses the release_channel v3

Previous patch was wrong, here is the good one. Sorry about that.

Comment 13

[Robert Helmer [:rhelmer]]

Comment on attachment 553401 [details] [diff] [review]
Search now uses the release_channel v3

Looks reasonable to me, let's get it pushed to stage and do more intensive QA on the data.

Comment 14

[Lonnen :lonnen]

Comment on attachment 553401 [details] [diff] [review]
Search now uses the release_channel v3

I'm passing on this. My khan set up is pooched and continually times out when I try to test this.

Comment 15

[Robert Helmer [:rhelmer]]

Trunk:
Committed revision 3459.

2.2 branch:
Committed revision 3460.

Comment 16

[Robert Helmer [:rhelmer]]

Strategy I used to test this in dev:

1) load topcrasher report for a mix of:
* old, f.e. 8.0a1
* new release, f.e. 5.0
* new beta f.e. 6.0(beta), 6.0b6
2) copy signature from report, note count 
3) do Advanced Search for "is exactly"
4) verify count is close 
5) click signature through to /report/list, count should be similar (total will not be displayed if not paginated unfortunately)

Comment 17

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

While testing this, I found:

* bug 679304
* bug 679306
* bug 679309
* bug 679310

I want to re-test once a few of those have landed, at least.

Comment 18

[Matt Brandt [:mbrandt]]

QA verified.

Using the steps to reproduce in comment 1:
1. Searching for "RDFContentSinkImpl::OpenObject(unsigned short const*, unsigned short const**)" (http://goo.gl/9aD2k) returns 88 matches on stage
2. Clicking on the signature in the returned results correctly displays a report with 88 signatures

Also, comment 10:
1. Appears to be working correctly now. Previously it wasn't returning any results  -> https://crash-stats.allizom.org/report/list?product=Firefox&version=Firefox%3A6.0%28beta%29&version=Firefox%3A6.0b5&version=Firefox%3A6.0b4&version=Firefox%3A6.0b3&version=Firefox%3A6.0b2&version=Firefox%3A6.0b1&version=Firefox%3A6.0a2&query_search=signature&query_type=exact&query=RDFContentSinkImpl%3A%3AOpenObject%28unsigned%20short%20const*%2C%20unsigned%20short%20const**%29&reason_type=contains&date=08%2F15%2F2011%2015%3A46%3A05&range_value=1&range_unit=weeks&hang_type=any&process_type=any&do_query=1&signature=RDFContentSinkImpl%3A%3AOpenObject%28unsigned%20short%20const*%2C%20unsigned%20short%20const**%29

Also:
Verified the above 2 using different signatures and different releases.

Lastly (comment 17):
The following bugs have been pushed off to 2.3
* bug 679304
* bug 679306
* bug 679309