679328 - Ability to search for raw version value

Status: RESOLVED WORKSFORME

(Socorro :: General, task)

Description

[Robert Kaiser]

After bug 679182, the search filters for our "display versions", which in the case of betas and release are actually either raw version + buildid + release channel or raw version + release channel(s) filters.

In some cases, like linking from external reports, or digging into the more raw data, it would be good if we could directly search for the saw version and e.g. get everything with a "6.0" version, regardless of build ID and release channel (or can manually set a build ID filter through the build ID field of the search).

Comment 1

[Robert Kaiser]

For reference, what I'm testing with right now is http://test.kairo.at/socorro/2011-08-14.buildcrashes.html and the queries it has linked in the upper block. Some examples:

https://crash-stats.mozilla.com/query/query?product=Firefox&version=Firefox%3A6.0&range_value=1&range_unit=days&&date=2011-08-14+23%3A59%3A59&query_type=contains&query=&reason=&build_id=20110804030150&process_type=any&hang_type=any&do_query=1
That should be all crashes from the 6.0b5 build on that day, covering 10836 reports, says "2,861 Results" (i.e. signatures) in the old code.

https://crash-stats.mozilla.com/query/query?product=Firefox&version=Firefox%3A6.0&range_value=1&range_unit=days&&date=2011-08-14+23%3A59%3A59&query_type=contains&query=&reason=&build_id=20110809231223&process_type=browser&hang_type=any&do_query=1
Browser-only crashes from a build that was not released as a beta, covering 14 crashes, showing 9 signatures listed in the old code.

Both return zero in the new code. I probably can work around it for now by setting version to "All" in the query, but in theory we could build an Aurora or Nightly build in the same second as a beta and therefore get the same build ID, or someone can run his own builds and force a build ID that matches a different one. We also tend to still get some 10-digit (to the hour) build IDs in crashes where our normal ones are 14-digit (to the second), I guess some Linux distro or so might still have an old patch in there that restricts that ID.

Comment 2

[Robert Kaiser]

With the new code, both URLs give no results with that additional filtering. The build IDs I used seem to be fairly unique, though (collisions are possible but mostly theoretical):

https://crash-stats.mozilla.com/query/query?product=Firefox&version=All&range_value=1&range_unit=days&&date=2011-08-14+23%3A59%3A59&query_type=contains&query=&reason=&build_id=20110804030150&process_type=any&hang_type=any&do_query=1
Says "2,861 Results", which matches the old query above.

https://crash-stats.mozilla.com/query/query?product=Firefox&version=All&range_value=1&range_unit=days&&date=2011-08-14+23%3A59%3A59&query_type=contains&query=&reason=&build_id=20110809231223&process_type=browser&hang_type=any&do_query=1
Looks like the one from above did earlier, and also has 9 signatures.

Given that, I'll switch my reports to use version=All as a workaround for the moment, until a better fix is available here.

Comment 3

[Robert Kaiser]

Super Search has actually fixed this (but doesn't support the reverse any more, which is bug 1131068).