well I think it would be best to have a table that shows for each value of each criterion, the last succesful state (if known)
so, for example:
criterion ---- last known state Arch: - i686 OK on 2011.02.10 (tested by $user) - x86_64 OK on 2011.02.05 (tested by $user) - dual/i686 not tested yet - dual/x86_64 FAILED on 2011.01.25 (tested by $user)
Image type: - Core OK on 2011.02.10 (tested by $user) - Netinstall OK on 2011.01.18 (tested by $user)
and so on (see my first mail for all criteria)
So, since every test report is a combination of reports for specific things (like "i tested an install with the i686 core image, with manual partitioning, ntp clock setting, etc etc and it worked fine"), but it will cause updates for various properties in the table above.
Dieter
Won't that get troublesome though? I mean, someone tests it with i686, core, automatic partitioning and it works, but then someone tests it with i686, net, automatic partioning and it doesn't. At this point i686 will have worked, core will have worked and automatic partitioning will have worked, but only net won't have. But then someone might test it with x86_64 and that does work (ok so maybe using the core/net difference might be a bad idea? sorry) then net will have worked too, but not on i686 which might not be apparent from this data. Or am I just being picky? Tom