[Libreoffice-qa] LibreOffice QA Call 2012-07-27

Hi all,

here are the minutes of the LibreOffice QA call on 2012-07-27 14:00 UTC.
Additions and corrections welcome.

attendance: Petr, Rainer, Norbert, Bjoern

completed action items:
+ - Invite active bugwranglers to next call/QA list, CC Rainer (Bjoern)
   - authorative close of bugzilla versioning thread (Rainer/Petr)
   - adapt scripts for new versioning (Markus)

pending action items:
   - write update scenario testcase in Litmus/MozTrap (Kendy)
   - Set Cor up with the Community/Forum maintainers at the distros
     to better propagate Hackfests, Bug Hunting Sessions etc.
     RedHat, Debian, Gentoo still missing
+ - Invite active bugwranglers to next call/QA list, CC Rainer (Bjoern)
   - merge 3.5 and 3.6 in one big bibisect repo (Bjoern)
   - recheck and tweak bibisect details (Bjoern)
   - research how it is done on SUSE/if there is a ready-to-use-extension
     for bugzilla email search improvements
   - Ping cloph if we can make that switchable to say "Bug" instead of
     "EasyHack" when explcitly requested (Bjoern)
   - new Bug Hunting Sessions (Cor)
   - betas for Fedora (Caolan)
                        
structured manual testing (Yifan/Petr?)

bug wrangling (Rainer):
   - bugzilla contract improvements (Rainer):
     - while there is some inital latency in a query, results are much quier
       and reliably to browse now
     - next up: OpenID for Bugzilla

community building/communication:
AI - send list of active wranglers (Rainer) (already done)

bibisect for tinderboxes (Norbert):
   - basically works on Linux, OSX. Windows is WiP
   - early tests encouraging:
     - 100 builds at ~400MB
     - Linux a bit smaller, OSX a bit bigger
     - at current growth would mean:
       - 60MB/day 300MB/week or 15GB/year
       - not blocking daily incremental uploads at that size
       - even smaller than uploading a full daily build
     - 'sampling out' with git rewrite should solve that
     - release branch builds still need solving
     - possible tweaks:
       - sdk generates quite big diffs by having timestamps
       - python seems to generate a diff contribution everytime too
       - fixing these would make the delta even smaller
       - l10n builds would be huge
         - OTOH does l10n change that often? (Bjoern)
       - builds with symbols are only plus 30MB, might be well worth it
     - we still need to have an upload location
       - clarify with Thorsten: can we do that on gimli (Norbert)
     - bytemark machine is waiting for use
       - should continue Bjoerns bibisect series (Norbert)

The next call is on 2012-08-10 1400 UTC.