minutes of ESC call ...
michael.meeks at suse.com
Thu Apr 5 07:55:05 PDT 2012
+ Tor, Rainer, Andras, Fridrich, Caolan, David, Michael,
Bjoern, Kendy, Michael S, Petr, Elidh, Lubos, Thorsten,
* Completed Action Items
+ Hamburg Hackfest
+ blog some artwork for 'going to the LibreOffice hack-fest' (All)
+ TDF blog entry publicising the Hamburg hack-fest (Fridrich)
+ extract 64bit build hardware from firewall (Kendy / Admins)
+ abandoned / getting new hardware from NGO
+ string dump / dissection scriptage (Michael)
+ re-file fdo#47857 as a bug in the KDE file picker (Michael)
+ drop exception specs from new string constructors (Lubos)
* Pending Action Items
+ [nearly done] review and re-close 3.4.x MAB fixed in 3.5.x (Rainer)
+ add a bold heading to GSOC page: LibreOffice is easy & fun to hack (Cedric)
+ rename VCL API to make it GetBeamerFoo & fix (Michael)
+ poke RedHat security guy wrt. keys (Caolan)
* Action Items review
* Release Engineering update (Petr)
+ 3.5.2 retrospective
+ released today, thanks for great work
+ smooth release, almost no delays
+ plenty more annoying / critical 3.5 bugs to fix
+ 3.5.3 rc1 deadline - April 16th (next Monday)
* GSOC update (Fridrich)
+ applications have to be submitted before tomorrow
evening european time (12:00 California time)
+ bug students if they want to apply
+ have 12 applications as of now
+ 3x for smartphone control of slideshow
+ good to encourage students to spread out.
AA: + mail the mailing list wrt. filing urgency (Fridrich)
* MSI cross-compile building (Tibbylickle)
+ wine is building nicely
+ cross-compilation is busted as of today
+ perl stuff seems straightforward
AA: + help Ellie get mingw build going (Tor)
* exception / compilation issue (Lubos)
+ current situation doesn't make sense, exception specifications
are there, but errors triggered only for clang compiler
+ either enforce them for gcc -and- clang, or not for either ?
+ potential for breaking things if we just remove them (sberg)
+ Solaris compiler did enforce them until we abdoned it a year ago
+ so code state is still reasonable
+ problems debugging what goes wrong with wrong exceptions (Lubos)
+ remove gcc option that disables checking for debug build ?
+ reason checking exceptions turned off
- makes binaries even larger (Caolan)
AA: + turn exception checking on for gcc in debug builds (Lubos)
+ makes clang building happier
* re-discuss default of encryption option (Thorsten)
+ do we need to re-visit this, cf. 2011/08/18
consensus to back-port etc. ?
+ question of defaults, code to handle read/write old & new
style encryption - should we stick with known poor old-style
ODF encryption to improve interop.
+ other office suites can use the old encryption
+ fewer bugs reports expected for conservative users
+ legacy crypto issue shows as
'file is corrupted' (Norbert)
+ there is a configuration setting for default encryption (Michael S)
+ but no UI => add a new UI option for encryption.
+ not for 3.5.x code-line.
+ choosing a known-weak algorithm is not ideal (Norbert)
+ chaff work + new encryption are much stronger (Michael)
AA: + switch default back for 3.5, keep as in 3.6 + add UI option (Thorsten)
+ MS filters don't support encryption at all (Thorsten)
* Apple developer enrollment
AA: + working on it as we speak (Thorsten)
* update our license header for new code to MPLv2
AA: + update the template (Michael)
* unowned easy-hacks ... (Bjoern)
+ all but 12 bugs now have owners
- Bjoern manually filed them from the wiki
- can't own all of them
*** Anything not owned by people will be closed in one week ***
+ please check & CC yourself on any you want to save:
* new bytemark tinderboxen (Norbert)
+ currently being setup
+ 2x boxes kindly provided by Bytemark
+ 1x windows
+ installed cygwin,
+ need a VC++ ...
+ 1x Debian
+ do generic linux builds
+ run 4x VM's for BSDs ?
+ Other ideas (Caolan)
+ or run startup times under callgrind [!]
+ and build regressions / memory usage metrics ?
+ or magically do bibisect packing ?
+ few 100'Gb's of storage ...
+ Michael S
+ had convwatch tests revived ...
bitmap output comparisons to avoid layout regressions ?
+ we have another Linux box coming on-line soon
+ TDF / SUSE hardware
AA: + there / ready / firewall issues getting solved (Thorsten)
AA: + give bytemark access to Caolan (Norbert)
* QA update (Rainer / Bjoern / Markus)
+ business as usual - nothing to report
+ pleased with enough Windows daily builds
+ how can we get new code contributors to contribute to bug fixing ?
+ discuss in person at Hackfest
+ turn more simple bugs into easy hacks ?
* updating internal packages for windows
+ libxml etc. are easy but ICU real killer (Caolan)
+ limiting factor is win build magic
+ they distribute project files for latest
+ project files, different to 2008
+ no tool to downgrade MS compiler versions
+ perl script to downgrade it doesn't work
anymore for new ICUs
+ tried with Tor, to build with VS2010
+ all managed code crashes: climaker bustage
+ no time to track down, prolly wrong run-time
+ other things to update
+ glib2, cairo, openssl, etc.
+ for linux, basically follow MacOS/X model (Caolan)
+ for anything that OS/X considers a system library
+ for universal builds: drop libxml cf. OS/X
AA: + pull up versions for 3.6 (Fridrich)
+ fix name clashes with the system
* 3.5 most annoying bugs ...
+ jump from consolidating the 3.4 issues:
+ 73 open (of 193), older 66/183 58/168 57/157 58/153 52/145 49/140 44/132 41/124
38% 36% 35% 36% 37% 36% 35% 33% 33%
* 3.5 bugs tagged with 'regression'
+ 134(-1) bugs open of 443(+11) total
* Component count net *
+ Writer - 51 (+2)
+ Presentation - 18 (+3)
+ Writer / RTF - 14 (-4)
+ LibreOffice - 13 (+0)
+ Drawing - 10 (+0)
+ Database - 8 (+1)
+ Spreadsheet - 2 (-2)
+ Basic - 2 (+0)
michael.meeks at suse.com <><, Pseudo Engineer, itinerant idiot
More information about the LibreOffice