[Libreoffice] minutes of tech. steering call ...
michael.meeks at suse.com
Thu Nov 24 08:26:48 PST 2011
+ Thorsten, Stephan, Mitch, Eike, Andras, David,
Bjoern, Kendy, Caolan, Michael S, Petr, Rainer,
Lionel, Tor, Markus
* Completed Action Items:
+ add merged msi details to 3.5 release page (Andras)
+ add unexpected mingw regressions to MingW 'most-annoying' (Rainer)
+ send ssh keys for Linux build machine access to Fridrich (Caolan)
+ concrete debug proposal for the public list (Stephan)
+ poke at and 'adopt' some easy hacks:
+ tinderbox owners should change names & Thorsten to prune server (All)
+ looks lots prettier ~done
* Pending Action Items
+ [in progress] default to TM safe (non-TDF) branding (Thorsten)
+ [in progress] enable on-line updates for QA in cross-compiled dailies (Kendy)
+ [in progress] extract 64bit build hardware from firewall (Kendy / Admins)
+ come up with a list of QA heros for next meeting (Rainer)
+ ask Christian wrt. Mac / PPC (Fridrich)
+ python/wizards merge for 3.5 & strip old versions (Bjoern)
* Action Items
* Release Engineering update (Petr)
- coming soon, kicking the tires of the build process /
+ commit deadline is Monday Nov 28 (next Monday)
+ feature freeze is 1 week later
+ Please check most annoying 3.5 bugs:
+ Windows error / during installation
+ extension registration / terminal launch issues
* Java 7 (Stephan)
+ requests / concern about it
+ already enabled for Linux
+ we should enable for Windows
+ duplicate the entry from XML file from Linux -> Windows
& sanity check.
+ rumoured instability with Java 7
+ is it generally fixed / still broken ?
AA: + enable Java 7 for Windows, post qa list to give it a try (Stephan)
* Unit tests issue (Markus [Moggi])
+ calc unit tests take around 30 seconds now
+ move some of them to tinderboxes / defer building
+ we have tons of filter tests - ~20 files to import & calc etc.
+ make subsequentcheck - the solution (Stephan)
+ the JUnit UNO API tests are a pain (Michael)
+ can we use 'make build' instead ?
+ concerned to run some basic tests every time
+ have lots of tests, that are nice, but don't need to
be executed every time.
+ no need for more different test rules (Bjoern)
+ add them to subsequentcheck
+ if we run just cppunittests in subsequenttests (Bjoern)
+ fast & easier to debug
+ ie. disable Java, and avoid undebuggable uno tests
+ tests should be executed with tinderboxes, useful (Markus)
+ have found problems, l10n issues etc.
+ number of ways to improve it (Caolan)
+ Bjoern has a nice write-up of how to debug it here:
+ can't we just run 'make build' and then 'make'
+ want to extend the tests yet further ...
AA: + move some slow tests over to 'make subsequentcheck' (Markus)
+ should be a ~trivial makefile rename / tweak
+ make subsequentcheck afterwards
* Officially kill bit-rotted pch work (Norbert in absentia)
+ eager to kill them, they're unmaintainable, bit-rots
with each commit (Kendy)
+ tried to use them once 6 momths ago, failed miserably (Tor)
AA: + remove them, or file easy hack for the same (Norbert)
* NlpSolver (Michael)
+ un-maintained, Sun extension, no live up-stream
+ flatten it into git as a module instead
* Cross-compiling MSIs building (Mitch)
+ slow but steady, continuing build.pl etc. re-factor
* QA Update (Rainer)
+ had to leave early.
* bug priority handling (Lionel)
+ vs. wrong icon issue - severity
AA: + help out diving deeper into it next week (Caolan)
+ is it a missing copy-constructor ?
* Cor's bug reporting context ?
+ table of who found what 3.5 bugs, how good they are ?
+ T-shirt prizes ?
+ resources low for triage, can we easily
assess quality of bug-reports for a fair metric ?
+ discuss with Rainer next time.
* 3.5 most annoying bugs ...
michael.meeks at suse.com <><, Pseudo Engineer, itinerant idiot
More information about the LibreOffice