[Libreoffice-qa] minutes of ESC call ...

Michael Meeks michael.meeks at suse.com
Thu Apr 4 08:58:35 PDT 2013


* Present
	+ Andras, Joel, Bjoern, Stephan, David, Astron, Thorsten,
	  Michael M, Caolan, Eike, Cedric, Kendy, Petr

* Completed Action Items
	+ poke website guys wrt. feedback menu-item / page (Michael)
	+ switch ESC time to CET to avoid changes (Michael)
        + Bugzilla attachments not set to autodetect (Tollef)
        + instead have the global headers in one central global directory
                + check behaviour of git-cherry-pick (Miklos)
        + SLES11 / temporary PTF for the server ? using SP3 toolchain ? (Kendy)
		+ Markus worked around it on his own machine
	+ find a slot in the schedule for 3.6.7 (Petr)
	+ poke UX / users list suggesting removal of inherited number formats (Markus)

* Pending Action Items
	+ Icons for Lionel + template manager (Astron/UX)
		+ Issa produced something - pending Lionel's input
        + disable Rhino / Beanshell unless in experimental mode (Michael M)
        + look at gradient / clipart issues (Michael)
        + helping out with code-pointers for UI bugs (Kendy)
        + need design for copying styles between templates (Astron/UX)
                + either in that dialog or a new dialog
                + also issue with only editing templates that are in the mgr
        + decided: change the default install directory to
          'LibreOffice 4' on master (Pmladek)
        + instead have the global headers in one central global directory
                + write up for the mailing list (Bjoern)
			+ [ half way through ]
	+ Personas - update / de-couple built-in URL (Kendy)
		+ legacy link will continue to work for a while.
        + buy Windows build hardware / hosting (Norbert)
		+ still investigating options
	+ update mac SDK configure check on master (Norbert)
	+ change -4-0 calc default to not use cached values (Eike, Kohei, Markus)
		[ Markus plans to submit on Monday ]

* Release Engineering update (Petr)
        + 4.0.2 retrospective
		+ it's out, few complex fixes left waiting for rc2
		+ otherwise seems to have gone well
        + 3.6.6 rc2 update
		+ tagged yesterday, builds for linux available,
		  Windows still building.
		+ looks promising.
	+ 4.0.3 rc1 - ~two weeks out April 15th
	+ 3.6.7 date: proposed for July 17
	+ 4.1 deadlines
		+ feature-freeze May 20th, Alpha1 two weeks before

* WikiHelp for 4.0 (Andras)
	+ what's the status of it, lots of complaints it's not on-line
	+ help appreciated (Kendy)
AI:		+ get wikihelp credentials to Andras so he can do it (Kendy)

* UX input (Astron)
	+ still pending a style copying design
	+ looking at a design-bugs initiative: easy-hacks focused on design
	+ would demand for rotation in writer be satisfied by
	  rotation of 90 degree multiples ?
		+ a very very common case; already have mirroring

* Removing C++ examples from the SDK (Bjoern)
	+ tried to get the C++ examples to work, tried on Windows
		+ a real pain to use that on Windows
		+ someone approaching the project from far away
	+ problem: people may try extensions, find it hard to use and leave
		+ whereas it is far easier to hack on the core itself
	+ where was the trouble, compiling SDK, or extensions ? (Stephan)
		+ example, compiling the C++ - problem
		+ setup of the SDK needed regardless of language (Stephan)
			+ needs attention & love
			+ from there a minor step to having C++ working
		+ is it a gbuild related regression (Michael?)
			+ SDK uses it's own build-system: dmake
			  tweaked to work with gnumake, still build.pl
			  based
			+ worked for Stephan on Windows recently fine.
	+ tried a native visual-studio environment, but failed (Bjoern)
		+ immediate crash on run; as well.
		+ build system we have is not ideal, building your
		  own doesn't work so well either.
	+ python does not require an SDK at all, no IDLs etc. (Bjoern)
		+ getting started is easier
	+ in practise it is always easier to go to the core; Java or C++ (Thorsten)
	+ making things easier for building extensions:
		+ provide a way to build extensions in our core build ?
		+ don't want to bundle extensions though (Stephan)
	+ easier to build the whole of LibreOffice instead of with the SDK
		+ could we not make it easy to build extensions inside
		  the main build ?
		+ with gnumake deps - don't need to build all of it
	+ we need someone to care for and love the SDK (Stephan)
		+ native platform builds
		+ the build-system is not set in stone -
		  it needs to change.
		+ eg. pre-compiling the headers
		+ external makes it harder to use internal impl. details.
	+ liblibreoffice might help (Michael)

* GSOC update (Cedric)
	+ keep improving the Ideas page, update on Monday

* An understanding on adding new committers (Thorsten)
	+ typical practise was to give a small domain of code
	  to commit to, asking for non-trivial commits on the
	  list etc.
	+ would like a feed for new people with commit rights (Stephan)
	+ can we make those lists public (Bjoern)
	+ mail the ESC when people are added, for people where we
	  have any doubts mail first.
AI:	+ mail list of committers to ESC (Bjoern)

* gerrit auto-building - how to trigger that etc. (Thorsten)
	+ Mac is being setup, still hunting for a nice/cheap
	  win service.
	+ be interested in how to trigger that (Michael)

* Python unit tests (Bjoern)
	+ although hard to debug, never have to debug most of them
		+ rare occurences they fail.
	+ story sprang out of Dresden sprint (Thorsten)
		+ 2x volunteers wanted to help write unit tests,
		  but only really comfortable in Python.
		+ David hacked something nice up.
		+ concerned wrt. level of push-back
	+ the purpose of unit tests is to break & then to show you
	  something useful (Michael)
	+ we can only gain from knowing something broke (Bjoern)
		+ but tests that fail for unclear reasons are
		  worse than useless, distracting, time consuming (Stephan)
		+ the Java tests are exactly like this, code written
		  by a non-core developer, problem with the test => very
		  hard to debug.
		+ the effect is that you disable the old/ugly tests
		  they randomly fail, so just disable them.
		+ concern that tests from people only capable of writing
		  tests in python might create a similar problem.
	+ Java tests written at Sun: two categories (Bjoern)
		+ UNO API tests - awful - can't debug, synthetic
		  synarios all over the place.
		+ specific tests - much more useful
	+ concern wrt. un-helpful push-back to new changes (Thorsten)
	+ should get it into master
		+ however - introducing a test framework means:
			+ helping people debug the problems involved
			+ supporting and maintaining your unit tests
		+ for now we should treat python tests as subsequenttests
		  alongside the Java ones with similar debugging issues.
		+ and evolve that as we see how they pan out.
		+ thank-you to David O for stimulating the discussion
		  & doing the work

* results/status of automated load testing (Markus)
	+ commitments from Kendy, Eike to fix some of the few
	  bugs found
AI:	=> publish the results as/when (Markus)

* QA update (Joel)
        + 108 unconfirmed vs. 4.0, help appreciated with triage.
                http://tinyurl.com/a3csc2o (cf. topic in #libreoffice-qa)
	+ over 1000 bugs reported in February
		+ over 6000 comments in a month
		+ fighting to keep up with bug reports.
	+ trying to work out what to do about extensions:
		+ they're not our bugs, we don't control them
		+ look under 'extensions and filters' component.
		+ come up with a solution for users.

* Open 4.1 MAB / regressions
        + 4 (of 10) older 5/10 5/10 4/8 4/8 2/5 1/2 3/3
        + https://bugs.freedesktop.org/showdependencytree.cgi?id=60270&hide_resolved=1

* Open 4.0 MAB / regressions
        + 18 (of 112) older 17/111 19/109 19/106 17/105 14/99 15/96 16/94 16/91 13/79
             16%             15%    18%    18%    16%    14%   16%   17%   18%   22%
        + https://bugs.freedesktop.org/showdependencytree.cgi?id=54157&hide_resolved=1

* 3.6 most annoying bugs ...
        + 73 (of 218) older 74/216 78/217 77/215 77/214 75/211 76/210 57/189 44/175
             33%             34%    36%    36%    36%    36%    36%    30%    25%
        + https://bugs.freedesktop.org/showdependencytree.cgi?id=44446&hide_resolved=1

* Bibisected bugs open: whiteboard 'bibsected'
        + 24 (of 83) older 27/83 25/78 29/78 32/79 27/72
        + re-built the query to avoid pre bibisected mess
                http://bit.ly/VQfF3Q

* all bugs tagged with 'regression'
        + 291(+1) bugs open of 1414(+13)total

        * ~Component   count net *
        + Writer       - 96 (-3)
        + Presentation - 29 (+0)
        + Spreadsheet  - 28 (+1)
        + LibreOffice  - 28 (+1)
        + Database     - 24 (+1)
        + Drawing      - 20 (+0)
        + Crashes      - 16 (-2)
        + Borders      - 15 (+0)
        + Migration    -  3 (-2)
        + Basic        -  2 (+0)

        + https://bugs.freedesktop.org/buglist.cgi?keywords=regression%2C%20&keywords_type=allwords&resolution=---&query_format=advanced&product=LibreOffice&list_id=36764
        + Migration: https://bugs.freedesktop.org/showdependencytree.cgi?id=43489&hide_resolved=1

-- 
michael.meeks at suse.com  <><, Pseudo Engineer, itinerant idiot




More information about the Libreoffice-qa mailing list