minutes of ESC call ...
Michael Meeks
michael.meeks at collabora.com
Thu May 8 08:29:41 PDT 2014
* Present
+ Christian, Stephan, David, Thorsten, Adam, Miklos, Kendy, Michael M,
Eike, Andras, Markus, Eilidh, Caolan, Kohei, Bjoern, Miklos,
Michael S, Robinson, Norbert, Jacobo
* Pending Action Items:
+ crashtest box - more research (Moggi/Cloph)
[ calc is working, writer problems can't easily be nailed down,
have to use multiple processes ]
[ calc numbers increased a bit - but not as bad as expected ]
+ will try filing a lot of small string changes as easy hacks (Astron)
+ encourage icon authors to put them into git (Astron)
+ fix the Java bridge for MSVC2012 for 4.3 (Stephan/Michael S.)
* Release Engineering update (Christian)
+ 4.3 status
+ looking so/so - reports that Windows / Alpha wouldn't run at all.
+ need to poke at the windows build environment, or a general
problem with the Alpha.
+ are we past the failing unit tests oddness ?
+ the failing tests are not reproducible in a debug environment
related to a parallel build (most likely).
+ created a new tinderbox with this baseline, which reports only
to tinderbox; hitting failing checks.
+ not always the same tests fail.
+ sd/ filters & sd/ import tests, also seen in sc/
+ more data? (Michael)
+ memory ? which exception fails ? which test ?
+ SEGV from 'make' (Cloph)
http://tinderbox.libreoffice.org/cgi-bin/gunzip.cgi?tree=MASTER&full-log=1399551601.15078
+ seems to fail on the first line etc.
+ re-running make passes in the same tree.
+ sounds like missing UNO components ? or dependencies (Bjoern)
+ maybe good to test 20-100 times after the build completed.
+ and if they're fine - its missing dependencies.
+ if you keep running 'make check' on a full-build
its unstable (Stephan)
+ slowcheck is a bit more stable than a subsequentcheck (Miklos)
+ if suspected with a failing UNO component ? (Kendy)
+ just add a printf
+ Windows tinderboxes do incremental builds and these
fail too (Sberg)
http://tinderbox.libreoffice.org/cgi-bin/gunzip.cgi?tree=MASTER&brief-log=1399545019.7798#28531
AI: + turn the SAL_WARN into a fprintf & send to Cloph ? (Kendy)
AI: + run the 'make slowcheck' 20x from the top-level & report (Cloph)
+ 4.2.x builds with older compiler & Windows version
+ 4.3 VS 2012 + Windows Server 2012
+ 4.2 VS 2010 + Windows Server 2008 rc2
+ 4.3.0 feature freeze:
+ week after next.
+ 4.2.4 RC2 status
+ announced today as final, enabled the update too.
+ update was broken
+ none of the 4.2.x was updated so far, until today.
+ discovered the ver. included the git hash there.
+ thanks to report from Michael Stahl.
+ we lean on QA to check this cf. MozTrap ...
AI: + poke Sophie to query that (Cloph)
+ BlueZ 5 bits get merged.
+ status of KDE4 support ? (Bjoern)
+ 4.2.3 deadlocked under KDE right away
+ best to poke Lubos - requires some KDE patch.
+ 4.1.6 announce / retrospective
+ Update enabled from 4.1.x -> 4.1.6 today.
+ good to see MAB's migrated to 4.2.x - 4.1.x is dead.
+ Android / iOS remote
+ patch for Android Remote for pointer feature,
+ also (re-)adding an About Dialog
+ with version; discussion ongoing.
+ 4.2.4 removed StarOffice XML export (Michael S)
+ bit unexpected
+ bug was around saving calc as old format (Kohei)
+ opted to remove rather than fix / work around.
+ was not added to release notes: we should add it.
+ failing tests - not running; and tests passed
and back-ported the fix to the -4-2 branch (Kohei)
+ one of the JUnit tests failed recently (Michael S)
AI: + Add the removal to 4.2 release notes (Cloph)
* 4.3 Features Page (Michael)
+ https://wiki.documentfoundation.org/ReleaseNotes/4.3
+ please update it !
* Should we simplify version numbering? (Kendy)
+ So far major.minor.micro, but due to our release plan, all our versions
are continuous improvements - does it make sense to still keep 'major'?
[ ie. should we do 4.3.0 -> 5.0, 4.4.3 -> 6.3, etc.? ]
+ Technical POV reasoning for stale 1st version number (Bjoern)
+ used to mean stable ABI
+ agreed - not such an issue these days.
+ technical reason is ABI stability, vague anyway.
+ Agreed, technically no need for it (Stephan)
+ What is the cost of change ? (Cloph)
+ book publishers unhappy with this
+ Scripts - implied dependencies: extensions, versions etc. (Bjoern)
+ potentially extensions break ?
+ update mechanism independent of versions uses git hash (Cloph)
+ download website - not so much of a problem (Cloph)
+ extensions that have a max OO version dependency is
essentially broken anyway (Stephan)
+ not expecting any problems there.
+ potentially extension website needs work ?
+ user profile: contains only major version '4' (Miklos)
+ some 'migration' / profile migration issues
+ only happened with the 4.0 release.
+ every major version would then require a profile
migration.
+ either get rid of it and keep '4' as magic forever (Sberg)
+ .so version numbers in libraries - may break (Bjoern)
+ new major .so version # => it costs packagers some time.
+ split packaging, and theoretical inter-package linkage etc.
+ can de-couple .so version from product version (Michael S & Norbert)
=> a reasonably small technical cost overall.
+ we discover that stuff later though (Thorsten)
+ but unfortunate to do this for 4.3, defer to 4.4
* Hackfests (Bjoern)
+ Gran Canaria Hackfest update / report (Bjoern)
+ http://skyfromme.wordpress.com/2014/04/29/libreoffice-las-palmas-hackfest-2014/
+ interest in python use with LibreOffice again
+ interested in GSOC students meeting up and/or coming
from the same area.
+ Achievements in the wiki page:
https://wiki.documentfoundation.org/Hackfest/GranCanaria2014#Achievements
+ Brazil LibreLogo workshop:
+ http://librelogo.org/2014/05/05/workshop-in-brazil/
+ Paris Hackfest:
+ June 27-28th
+ please add yourself to the wiki:
+ https://wiki.documentfoundation.org/Hackfest/Paris2014
+ HackFest 'pack' / box / Virtual Machines / etc. ?
+ idea of the Virtual-Boxes is rather good (Kendy)
+ don't want to discourage that.
+ Peli-cases for hack-fests would be cool
+ 1x powerful machine for icecream master & slave
+ fast switch & some ethernet cables, power blocks etc.
+ At Gran Canaria - wifi was somewhat flaky etc.
+ unable to use icecream - no wired connection.
+ Still have hardware to try out (Bjoern)
+ can bring to Paris - some quadro / quad CPU thingit
(TDF-owned Sun Ultra 24 with Q9650)
+ if we buy new hardware: something like tinderboxes
we have or ...
+ alternatives (from ebay):
+ refurbished HP DL165 with 24 cores at 1.5K EUR at 1HU/15kg
+ refurbished Lenovo W520s at ~1K EUR each (and plane-suitable)
+ total cost sub Eur 5k
AI: + add it to the next Board budget extensions (Michael)
+ bit expensive if H/W sits idle rest of the year (Bjoern)
+ happy to have it at his place as a tinderbox the rest of
the time (Kendy)
+ something that would be permitted to travel with a person
on a plane (Thorsten)
+ problems getting people to accept that stuff at the
target location etc. would be a shame to loose it.
AI: + budget for Amazon VM's for hack-fests (Thorsten)
* LibreOffice Conference Call for Papers (Michael)
+ please think of what to submit:
http://nabble.documentfoundation.org/LibreOffice-Conference-2014-Call-for-Paper-td4100170.html
+ May 15th deadline ...
+ Adam - paper on interop. status with DOCX and yearly update
+ Caolan - what I did last year
+ Miklos - fixed writer inconsistencies
+ Jacobo - Accessibility bits
+ Cloph - Release engineering
+ Cloph - build / tinderbox setup
+ Stephan - Compiler and Me
+ Eike - mix of things done
...
* QA (Norbert)
+ NeedAdvice:
+ Font rendering issues in KDE: https://bugs.freedesktop.org/show_bug.cgi?id=59818
+ help appreciated
+ buzilla migration
+ when its good enough will go back to Tollef for a real date.
* QA stats:
+ Thanks to Tommy for migrating MAB's from 4.1 -> 4.2
+ https://bugs.freedesktop.org/page.cgi?id=weekly-bug-summary.html
+171 -181 (-10 overall)
many thanks to the top bug squashers:
tommy27 22
Foss 16
Jean-Baptiste Faure 15
Maxim Monastirsky 13
Joel Madero 13
Julien Nabet 13
Kohei Yoshida 11
Jorendc 7
Cor Nouws 7
m.a.riosv 6
Caolán McNamara 5
* Open 4.3 MAB
+ 3/7 2/5 1/2 1/2 0/1 0/1 1/1 1/1 1/1
+ https://bugs.freedesktop.org/showdependencytree.cgi?id=3D75025&hide_resolved=3D1
* Open 4.2 MAB
+ 49/165 18/128 19/124 18/123 24/120 20/112 19/111 20/111 18/104 17/102
29% 14% 15% 14% 20% 17% 17% 18% 17% 16%
+ https://bugs.freedesktop.org/showdependencytree.cgi?id=3D65675&hide_resolved=3D1
* Open 4.1 MAB
+ 43/179 76/208 78/208 78/208 79/208 79/206 84/206 85/206 86/204 81/198
24% 36% 37% 37% 37% 38% 40% 41% 42% 40%
+ https://bugs.freedesktop.org/showdependencytree.cgi?id=3D60270&hide_resolved=3D1
* Bibisected bugs open: whiteboard 'bibsected'
+ 46/207 47/201 47/198 46/197 39/189 38/185 40/185 39/182 39/181 40/180
+ http://bit.ly/VQfF3Q
* all bugs tagged with 'regression' (vs. 2 weeks ago)
* Michael Stahl added the missing components to the list here,
and since this causes some change of datum / uplift anyway;
I also dropped NEEDINFO / UNCONFIRMED regressions from the stats.
+ 318(-5) bugs open of 2543(+32) total
+ these numbers continue to be accurate for the full
set of all regressions.
+ These numbers are thus updated and somewhat incomparable.
~Component count net *
Writer - 85 (+3)
Spreadsheet - 47 (+3)
Crashes - 21 (+1)
Database - 21 (+0)
Presentation - 21 (+0)
Libreoffice - 20 (-3)
UI - 20 (-7)
Borders - 15 (+1)
Graphics - 14 (-1)
Drawing - 12 (-1)
Chart - 10 (+0)
Print / PDF - 8 (-2)
Filters - 9 (-3)
Extensions - 5 (+0)
Formula - 3 (+0)
Linguistic - 3 (+0)
Installation - 2 (+0)
Migration - 2 (+0)
BASIC - 2 (+0)
framework - 1 (+0)
sdk - 0 (+0)
Impress Remote - 0 (+0)
+ http://bit.ly/15mM2Yn - for devs ( no NEEDINFO / UNCONFIRMED )
+ https://bugs.freedesktop.org/buglist.cgi?keywords=3Dregression%2C%20&=keywords_type=3Dallwords&resolution=3D---&query_format=3Dadvanced&product==3DLibreOffice&list_id=3D36764
--
michael.meeks at collabora.com <><, Pseudo Engineer, itinerant idiot
More information about the LibreOffice
mailing list