[Libreoffice-qa] minutes of ESC call ...
michael.meeks at collabora.com
Thu Jun 4 08:46:46 PDT 2015
+ Sophie, Norbert, Stephan, Eike, Miklos, Robinson, Kendy, Noel,
Bjoern, Jacobo, Muthu, Jay, Andras, Michael M, Caolan
* Completed Action Items:
+ kick-off a set of "who wants what test infrastructure" thread (Michael)
+ poke 2x certification candidates (Stephan)
[ working through mailing list setup with Italo & waiting responses ]
+ chasing un-translated 4.4 string (Sophie)
[ a commit with hard-coded string, made translatable later but not
back-ported due to the freeze => should break the rule & back-port
it's an important string in the UI.
Please back-port commit:
+ add ooenv MALLOC_CHECK etc. to script wrapper for tinderboxes (Norbert)
[ Would (by now) only improve the build tooling -> axe it ]
* Pending Action Items:
+ find some developer/sysadmin'y type people to train in CI maintenance (Cloph/Norbert)
[ had a chat after the call with Cloph - started this; ongoing ]
+ move problem writer tests from check -> subsequentcheck for sw.all (Michael S)
+ encourage QA guys to test mork / thunderbird bits on Win64 (Robinson)
[ encouraged people to build there; have only 1x tinderbox that
can make builds but isn't.
In the process of buying bare-metal Win64 infra boxes (Norbert)
+ pending delivery etc. couple of weeks ?
+ spec. decided on - and in TDF procurement
Who runs the Win64 tinderbox ? (Michael)
AI: + Thorsten - can we get builds up-loaded (?) ]
* Release Engineering update (Robinson)
+ 5.0.0 - Beta 2
+ going to tag shortly, and get builds out soon.
+ Status ? (Norbert)
+ need Cloph's help to trigger builds (Robinson)
+ good to ping Bjoern/Rene/Norbert when tagging (Norbert)
+ build/sign machines have v. limited access (Norbert)
+ Late Feature Status (Michael)
+ LibreOfficeKit / Online tweaks (Kendy)
+ mostly fixes this week etc.
+ OpenGL / double-buffer RenderContext (Kendy)
+ similar - lots of fixes, some other bits for review in gerrit
+ gtk3+ theming, embedded video, cut/paste (Caolan)
+ no concerns.
+ toolbar search / wrap-around (Kendy)
[ got its reviews & went in ]
+ text wrapping for chart labels (Marco)
[ punted to 5.1 -> remove]
+ orcus with gnumeric import (Markus)
[ build issues -> will not push to 5.0 ... -> remove. ]
+ pending more QA work.
+ 4.4.4 RC2 status
+ freeze next week / last fixes ?
+ Android & iOS Remote
* Menu changes / help & documentation handling (Jay / Sophie)
+ Interface / communication concerns between Help & Development (Sophie)
+ How do we tell docs team wrt. change ? (Norbert)
+ docs team don't work on help files (Sophie)
+ dev. team thinks docs team is working on it & they are unable
to edit the help & are under-resourced
+ talked wrt. what is technically necessary to edit help files (Kendy)
+ Pushing in changes to re-org the menu bar (Jay)
+ changes would affect docs as well as help.
+ how to approach that ?
+ editing the help XML files - easier than changing
menu XML descriptions (Miklos)
+ documentation more problematic than help (Jay)
+ not reasonable to ask dev to fix documentation screenshots (Norbert)
+ important to fix the limbo problem between dev & docs
+ Changes made will affect people too (Jay)
+ good to do that in 5.0 (Kendy)
+ hoping to get them in there, completing toolbars / context menus (Jay)
+ completed the menu-bars for writer, calc, impress, draw
+ reviewing the changes currently (Sophie)
+ don't want to loose functionality.
+ Wiki Help makes it easy & familiar to change the help (Kendy)
+ grow the community of those able to lead improvements in the help.
+ technically barrier to entry stops those who have good writing skills
but perhaps less technical / manual-XML editing familiarity.
+ not there yet unfortunately.
+ no easy path for editable 5.0
+ what is the mechanism for notifying docs generally ? (Norbert)
+ an issue and a wiki-page about this (Sophie)
+ Currently a wiki page with 4.4 and 4.5 UI / string changes (Jay)
AI: + link it from the developer pages (Sophie)
+ ideally we'd automatically generate screenshots for docs (Michael)
+ would test the UI as well potentially (Norbert)
AI: + create the GSCO idea - Kendy love to mentor (Norbert)
+ fixing help for 5.0 ?
AI: + help to re-arrange the help XML for the changes (Jay)
+ minimal English & content required (Norbert)
+ Olivier Hallot has done this in the past - be a good contact ?
+ cherry-pick this to 5.0 ? (Kendy)
+ no objections.
+ best to do it in 5.0 than a minor release (Norbert)
+ late feature approved
+ get it in for Beta 2
+ modulo fixing the help before release.
+ modulo Sophie's review - revert any feature loss.
* UX Update (Kendy)
+ switched OS/X default to Breeze icons from Sifr
+ much more complete than Sifr; as suggested by Mac guys.
+ Jay's GSOC student - working on sidebar improvements
+ will also spend time on change-tracking sidebar.
+ would love a chart sidebar (Michael, Norbert)
+ Markus is working on it (awesome)
* Changing .uno:InsertRows / .uno:InsertColumns action names (Kendy)
+ https://gerrit.libreoffice.org/#/c/16007/ renames:
+ .uno:InsertRows -> .uno:InsertRowsBefore
+ .uno:InsertColumns -> .uno:InsertColumnsBefore
+ reasoning is that:
+ in calc, InsertRows means "before", in writer it means "after", and
a later patch adds .uno:Insert*After which makes it hard to create icons
that represent the difference
+ the problem is that .uno:Something can be used from extensions or macros,
+ do an incompatible change?
+ change the meaning in Writer, not rename the current ones, and just add the *After variants?
+ anything else?
* Whatever we do here - we will be incompatible ...
+ no profile change (Norbert)
+ can we maintain the profile with this ?
+ macros can call these 'insertrows' via UNO (Kendy)
+ or just change behavior in writer ?
* can we not leave it as-is and add 4x new variants (Michael)
+ in writer same command, different functionality (Jay)
+ less controversial approach to retain old versions (Miklos)
* decision: stick old names for odd behavior & add 4x new specific consistent ones
+ add new ones to the toolbar & hope the old ones disappear over time.
* concerns putting it into 5.0 (Eike)
+ unsure if we have some assumption that rows are always inserted above.
+ can we punt that to 5.1
+ adding the UNO commands fine - but don't use in calc by default yet (Eike)
* Crashtest update (Caolán)
+ killed all the recent spikes
+ DrawingML recovered
+ missing namespace assert is fixed.
+ ~zero import issues
+ fluctuates between zero and one
+ something non-repeatable & odd.
+ calling via scripts, OOM etc. ?
+ slightly different idle layout issues too.
+ odd ordering / fonts etc.
+ need to look into that.
+ word-count forced on export for document stats (Michael)
+ export crashers:
+ last ODT export crashes fixed by Phil Jung
+ pushed a whole load of hyperlink fixes for docx
+ expert next set of results to be ~30 in total.
+ doing what it's supposed to do:
+ finding regressions quickly.
+ make sure students are on-track
+ mid-term is not so far out - June 26th, deadline July 3rd.
* Cambridge Hackfest survey (Andras)
+ Anonymous post-hackfest survey results.
+ 10x questions - got 9 resposnes of 14 participants
+ everyone liked scheduling: Thur -> Sat.
+ liked the timing midday -> midday
+ all but 1 - easy to get to venue
+ liked activities, beer festival & punting
+ partially / mostly achieved what they wanted to
+ most controversial - noise level
+ some happy, others v. noisy;
+ several comments eg. (ono)
"no loud off topic talks should be allowed"
"good to have separate rooms for chat, and to
communicate at start so no offense"
"shame to see only 1x new contributor Ben"
be good to have lighting talk slot ?
* Hackfests (Bjoern)
+ huge gapo from now -> next hackfest
+ but 5x hackfest per year too much last year.
+ a 'Virtual Hackfest' ?
+ don't we have one every day (Michael)
+ yes; add more outreach to newbies ? (Bjoern)
+ in the planing phase.
+ Hamburg, Germany: Oct 23rd -> 25th
+ draft page; to be filled out
+ great to see everyone there.
+ next venues / suggestions
+ Belgium another candidate ...
* ownCloud integration for Android (Jacobo)
+ working on the missing bits; now ready to be re-based and integrated
+ should it be back-ported to some branch ?
+ a couple of viewer branches in the repo ?
+ put it on master, and people will find it (Michael)
+ only a single feature-branch touching tiled-editing (Miklos)
+ GSOC, tiled editing foo.
+ managed to fix all the library oddness
+ had binary java libraries inside
+ in the end - built a package with all the deps in the source-code
+ it works & builds nicely.
* l10n Update (Sophie)
+ nothing new.
* Developer Certification (Stephan/Bjoern/Kendy/Thorsten)
+ at the beginning
* New Committers
+ Philip Jung - no concerns.
AI: + add and send welcome mail (Norbert)
* Jenkins update (Norbert)
from:Thu May 28 13:37:25 2015
master linux rel jobs:155 ok:154 ko:1 fail ratio:0.64% break:1 broken duration:0.04%
master linux dbg jobs:107 ok:80 ko:23 fail ratio:21.49% break:2 broken duration:29.43%
master mac rel jobs:130 ok:124 ko:4 fail ratio:3.07% break:2 broken duration:0.64%
master mac dbg jobs:123 ok:118 ko:4 fail ratio:3.25% break:2 broken duration:0.65%
master win rel jobs:94 ok:77 ko:17 fail ratio:18.08% break:10 broken duration:18.12%
master win dbg jobs:91 ok:81 ko:10 fail ratio:10.98% break:3 broken duration:13.93%
lo-4.4 mac jobs:10 ok:10 ko:0 fail ratio:0% break:0 broken duration:0%
gerrit win rel jobs:208 ok:112 ko:95 fail ratio:45.67%
gerrit lin rel jobs:211 ok:181 ko:29 fail ratio:13.74%
gerrit mac rel jobs:213 ok:180 ko:32 fail ratio:15.02%
gerrit all jobs:220 ok:106 ko:113 fail ratio:51.36%
+ backs are somewhat normal (for us) again.
+ gerrit numbers lag week before master state.
+ an issue with Jenkins
+ some builds may have dropped
+ bumped the RAM, upgraded a bit.
+ last 12x hours ok.
+ spent TDF cash to buy new hardware to get faster Windows CI
+ should improve response time.
+ during the day - things stock-pile; 10-20 builds pile up.
+ time from gerrit commit -> build, 5-6 hours.
+ usually overnight it drains and
+ the revenge of success (Michael)
* Easy Hacks:
+ new easy-hacks thanks to:
+ Bjoern Michaelsen
+ create your own by setting whiteboard to eg.
"easyHack difficultyBeginner skillCpp topicCleanup"
* Updating ESC composition
+ New proposal for refreshing that by E-mail
+ Propose a tweak:
+ as before; plus Sophie & Kendy
+ no objections
AI: + wait a couple of days & send -> board (Michael)
* QA (Robinson)
+ UNCONFIRMED at 378
+ its that time in the cycle ...
+ needAdvice at 11
+ Bibisect 43max and 50max repos available
+ mjayfrancis - more accurate / granular repos.
+ Bibisect / win32 (Norbert)
+ been incrementally up-loading it, windows is about 1100 builds behind
+ started at 9000
+ git repo - physically same machine as gerrit - can just clone it.
+ is it documnted on the bibisect wiki page ? (Bjoern)
+ need to write it up (Norbert)
+ Mac one is complete & up-there;
+ commit format msg is a bit different; it lists all
the source commits between.
+ git log --grep <source-sha> - shows the commit that includes it.
+ for the 5.1 epoch -> setup Jenkins to build bibisect as we go.
+ will have during the dev period, not just at the end.
+ intend to do it for Windows as/when we can afford it.
+ Win64 builds
+ hopefully coming with the new H/W.
+ Windows 10 (aka "Threshold") - Release date: July 29th (54 days away)
+ coming at a similar time; be great to have testing on it.
+ call for interested smoke-testers ...
+ using Windows Server 2012 R2 (Norbert)
+ 2008 server on old box, and 2012 R2 & new H/W the same
+ much more convenient & reasonably robust too ...
+ EC instances that stay up.
+ Regression count is too high (Michael)
+ clearly the best way out & approach is not process, but fixing them.
+ ... or catching and not introducing them in the first place (Bjoern)
+ bit optimistic (Michael)
+ good way - stop coding altogether (Norbert)
+ in general - would be good to categorized regressions into bit-sized chunks (Caolan)
+ developer motivation vs. the cliff-face
+ list of 3x easier to fix than list of 600
+ anything re-opens in this category becomes a crisis etc.
+ writer 130 regressions:
+ needs to be sub-divided down further.
+ writer eg. comments by Michael S - doesn't support floating tables (Bjoern)
+ eg. make 1x floating-table meta-bug and close rest as dups
+ it only worked by accident in the past.
+ can we do a minimal change - still seem to work (Miklos)
+ this week fixd a regression around floating tables in headers
+ not really related to floating tables; just not throwing up hands
in the filter: got back to old situation.
+ in many cases - still something we can do.
+ went through all writer regressions recently (Bjoern)
+ 10% not really broken existing fuctionality
+ proper error handling on XML in some context ...
+ stuff broke: we noticed there was an error ...
+ some degree of support required for de-regression-izing these ? (Michael)
+ recommendation - bugs in this category - mark it differently & filter out (Bjoern)
+ enhancement, priority low.
+ help to reduces the subset of nasties.
+ catching regressions early (Bjoern)
+ ring change in writer; lots of minimal changes, each change 5x lines.
+ caused a regression; by having a small commit -> helped find it.
+ starting to look at calc regressions @ Collabora (Michael)
+ improving test infrastructure (on-list)
+ please reply to mail on dev-list with ideas.
+ subsequentcheck parallelism (Kendy)
+ when run with icecream -> out of memory.
+ doesn't it use icerun (Michael)
+ how does that work (Bjoern)
+ parallelism is passed to make -j (Kendy)
+ can we use icerun (Michael)
+ building bibisect for windows (Norbert)
+ out of 7000 build so far, 1700 did not build (compile or link)
+ a build without test & -Werror.
+ not good.
+ CI infrastructure; more important (Bjoern)
+ if you break windows, you break CI.
+ wider discussion before pushing automated code / bugzilla changes: good:
+ Stephan & Noel - punt it again ?
* QA stats:
+169 -120 (+49 overall)
many thanks to the top bug squashers:
Adolfo Jayme 13
Caolán McNamara 12
Michael Meeks 11
Yousuf (Jay) Philips 7
V Stuart Foote 5
Terrence Enger 4
Alex Thurgood 4
Katarina Behrens (CIB) 4
Cor Nouws 3
Julien Nabet 3
Andras Timar 2
Markus Mohrhard 2
* Open 5.0 MAB
+ 6/12 5/11 6/11 5/9 6/9 5/8 5/8 4/7 2/5 0/3 1/3 1/3 1/3 0/2 0/2
* Open 4.4 MAB
+ 17/61 16/60 19/60 15/56 16/56 15/53 15/53 14/52 15/52 14/51 14/50 13/49
* Open 4.3 MAB
+ 75/170 78/170 77/169 77/169 78/167 79/166 79/166 80/165 79/163 78/163
* Bibisected bugs open: whiteboard 'bibisected'
+ 329/898 335/895 333/874 337/869 334/849 334/839 327/831 330/820
* all bugs tagged with 'regression'
+ 599(+2) bugs open of 3881(+29) total
* ~Component count net *
Writer - 211 (+1)
Calc - 118 (+0)
Crashes - 51 (+5)
Impress - 48 (+2)
LibreOffice - 39 (+0)
UI - 32 (-2)
Draw - 21 (-1)
graphics stack - 26 (+1)
Base - 19 (+1)
Borders - 18 (+0)
filters and storage - 12 (-2)
Printing / PDF export - 10 (-1)
Chart - 10 (+0)
framework - 4 (+0)
Extensions - 4 (+0)
BASIC - 4 (-1)
Formula Editor - 4 (-1)
Linguistic - 2 (+0)
sdk - 1 (+0)
Installation - 1 (+0)
Impress Remote - 0 (+0)
michael.meeks at collabora.com <><, Pseudo Engineer, itinerant idiot
More information about the Libreoffice-qa