minutes of ESC call ...

* Present:
    + Stefan W. (smux, muted), jan i, Sophie, Lionel, Norbert, Miklos, Stephan,
      Caolan, Michael M, Eike, Noel, Jan-Marek, Markus, Thorsten, Michael S, Olivier,
      Heiko, Andras, Bjoern (10 minutes in)
 
* Completed Action Items:
    + mark no-code-pointer Easy Hacks as NEEDINFO & NeedsDevEval (Heiko)
 
* Pending Action Items:
    + connect Cloph + Norbert wrt. Mac & Windows machines for base (Lionel)
         [ unable to do anything, can't find student on IRC at the same to
           give access to remote Mac / Windows machines:
               some good progress ongoing in Firebird (Lionel)
               need an SSH key and desired user-name to access a Win firewall (Cloph)
                  + idle Swiss Mac Mini ? (Norbert)
                  + JanI happy to help with Win issues. ]
    + investigate a cron job that queries & auto-merges (Norbert)
         + if +2 by author and +1 by jenkins -> auto-push (on-going)
            + SQL query done.. but yield interesting result...
                  https://gerrit.libreoffice.org/#/c/26347/  which 1/ is marked
                  'not mergeable' but I can't yet detect that
              and really should not be merged anyway.. based on comments.
                  + Not found a way to detect this yet.
              + That's probably a typo, he must have meant -1/-2 (Lionel)
                  + in the same click foo.
              + hope is that if it gets merged this will stop (Michael)
              + says "can't merge" but not reflected in SQL table (Norbert)
              + what happens in the chron job +2 by author, but -2 by other ? (Lionel)
                  + query - condition - is: +2 by author, and verified +1 by ci (Norbert)
                      + has to be a committer - no non-committers can't +2
                      + wise to check no -1's by anyone else.
                      + also only valid on master, not elsewhere.
                  + on other branches: could automate the rules (Lionel)
    + tweak UI and get LiveConnect API key / build case for board (Christian)
    + attempt to re-build a recent gstreamer 1.0 / core spec file
      on our CentOS6 base (tdf#94508) (Christian)
 
* Release Engineering update (Christian)
    + 5.2 RC2 - due July 5th (Tue next week)
        + will create libreoffice 5-2-0 branch
    + 5.1.5 rc1 : July 12th (week 28)
    + Android & iOS Remote
        + still crashes on startup.
 
* Documentation (Olivier)
    + Delivered chapter on Base for revision
    + To close the book on 5.1 - need to update base & impress
    + next is 5.2
    + Discussion with staff on how to increase Doc' visibility & Download rates (WIP)
        + More downloads, more public, more reads, more interest,
          better feedback/contrib/volunteers
        + a 2nd product - books; can give beside the software download.
           + can we have links in the help menu ?
           + such as books, ask etc. ?
           + easy to do - poke UX guys too (Michael)
    + Blog post
        + asking for new people to show up:
            
https://blog.documentfoundation.org/blog/2016/06/29/libreoffice-documentation-join-us-in-this-adventure/
        + two new people showed up already
 
* GSoC 2016 (Thorsten)
    + Everyone passed at Mid-Term
        + progress looks good.
        + next deadline is mid/end August for final evaluation.
    + poke & ask/encourage students to submit & register for LibOCon
        + particularly for foreign students.
        => will be re-imbursed up to Eur 1k each, for more ask mentor.
    + blogging encouraged too ! with screenshot would be good.
    + talked to UX students (Heiko)
        + not having time to come to the conference.
AI: + send E-mail to ask & encourage (Thorsten)
 
* Disable OpenCL by default in Calc (Markus, Eike)
    + Quite a lot of nasty bugs with OpenCL (Markus)
        + hard to reproduce them
        + three different code-paths to calculate
        + Hard to reproduce the bugs
    + Tracker bug (Michael)
        + https://bugs.documentfoundation.org/showdependencytree.cgi?id=97391&hide_resolved=1
        + 23 bugs over all time, 6 open currently, 4 needinfo 2 un-confirmed.
    + Crashes:
        + bad - but mis-calculation worse (Eike, Michael)
            + not spotted mis-calculations are worst.
            + need to avoid them.
    + Concern (Markus)
        + quality, and rate of bug fixing.
    + OpenCL - suggested improvements: (Michael)
        + test it when we 1st ever start - before we do real calcs.
        + OpenCLZone - cf. GL stuff to turn off on crash.
        + more active maintenance.
    + currently 3 different code-paths (Markus)
        + don't know which one we're using.
        + don't like the S/W interpreter.
    + Sw/Interp. (Michael)
        + only used for subset of ocAcc, ocSub, ocMul, ocDiv, ocSum, ocProduct
            + much faster.
    + Tor just accelerated the CPU S/W vs. OCL testing (Michael)
    + don't see value of it for the project at its current state (Markus)
        + would not use it himself.
    + Crash reporting:
        + e.g. 
http://crashreport.libreoffice.org/stats/crash_details/4aafce7f-71cc-42c0-893b-6aeebec158ac
        + there are some issues here - but there are plenty of crashes to go around (Michael)
    + Why is it useful for LibreOffice, given that few people would see the perf improvement ? 
(Thorsten)
        + the default is very important to Collabora & our calc investment (Michael)
    + how much data do we need (Heiko)
        + 100 rows currently by default (Michael)
        + is that worthwhile ? (Eike)
            + consider 100x of a SUMIF operating on big data (Michael)
            + could improve the algorithm to find big N^2 etc. (Michael)
        + bug reports - put as needinfo / worksforme (Markus)
    + read through all bugs containing 'opencl' today (Michael)
        + some QA burden of asking for status - but lots independent
    + concern mostly around the accuracy (Norbert)
        + screwed calculating taxes etc. PR issue.
            + prefer calc is slow to calc is inaccurate.
    + tinderboxes running OCL tests ? (JMux)
        + run whatever is setup to run.
    + the problem [ via IRC ] (Moggi)
        + is that normally calculation bugs and even more regressions are fixed quickly
            + true.
    + how many benefit ? (Norbert)
        + presumably only huge sheets.
        + can reduce the amount of usage to just those (Michael)
        + defaults change-able by packaging (Bjoern)
        + prefer on/off vs tweaking threshold (Norbert)
            + need some 'is work big enough to be worthwhile' heuristic (Michael)
                + for testing we turn it right down.
        + if significant concern by those doing work - be prudent.
    + like to see it in the about box (Bjoern)
        + helps QA if we add to Help->About => to paste it easily into bugs.
    => propose come back with a proposal next week on it (Michael)
        + if we can't resource maintaining it properly, disable it by default.
 
* UX Update (Heiko)
     + GSoC projects are in progress
         + mix of progress across students.
     + Discussion about go-to dialog (tdf#83054) and how to
       realize unicode emojis (tdf#100100)
         + use navigator, interaction-like gallery or
           special-character like dialog
     + crunching data on what is done with draw
         + really creative with the tool
             + holes in docs, etc.
         + draw is great from a functionality perspective.
 
* Crashtest update (Caolan)
    + ? import failures, ? export failure, ? coverity
       + no results back from this week.
       + expect it stable in the next day.
    + coverity - runs not working at the moment.
 
* Security (Bjoern)
    + CVE went out Mon/Tues
       + v. small fix; but softpedia made an article about it.
       + worth pushing fixes
       + 5.1 and 5.0 back-ports the same (Caolan)
       + a read / dereference of an empty stack
 
* Hackfests (Bjoern)
    + next hackfests: nothing confirmed yet.
    + Hamburg Home-hacking setup ? (Bjoern)
        [ various people on vacation now ]
    + next venues / suggestions
        + Chaos event - good for outreach; at the same time as Munich.
        + Infra Meetup: July 24th to 30th, close to Munich (kaufbeuren)
        + Akademy / FSFE summit - in the week before LibOCon ? (Thorsten)
        + more suggestions welcome
    + Autumn / week-long Home-Hacking in Hamburg with students? (Bjoern)
        + advertise at university ? C++ training ? etc.
 
* LibreOffice Conference in Brno - CfP open (Sophie)
    + Kendy & Thorsten on program committee:
    + 
https://blog.documentfoundation.org/blog/2016/04/08/libreoffice-brno-conference-call-for-paper/
        + deadline is mid July.
        => think what to submit / chase people ...
    + travelling to Brno -> fly to Prague / Vienna unlikely to go direct.
 
* Mentoring/easyhack update (janI)
   + legend: contributors are not-yet-committers, numbers in () are +/- compared to last week
   + summary:
       17143(-214) commits in a year of which 8419(-262) were made directly, without use of gerrit
       295(2) people made commits in a year, 68(68) in a month. 44(-3) people to be emailed
   + gerrit:   committer open  62(-1)      contributor open  42(-9)
               updated merged    abandoned   updated  merged    abandoned
        year:   62(-1)  7261(39)   565(7)     42(-9)   1463(9)   478(0)
        month:  61(-1)   848(-77)   41(-2)    42(-9)    192(-3)   37(-4)
        week:   44(-1)   135(-42)   13(4)     19(-11)    46(-11)  10(-2)
   + easyHacks:
        total 228(-2) open 200(-2) assigned 18(0)
        needinfo 10(0) needsDevEval 473(2) needsUXEval 112(11)
        missing cleanup due to many comments 183(-1)
   + top 5 contributors (based on merged patches last month):
         Zdenek Crhonek got 62 patches merged, and have a total of 79 merged patches
         Muhammet Kara got 21 patches merged, and have a total of 37 merged patches
         krishna keshav got 14 patches merged, and have a total of 14 merged patches
         Jakub Trzebiatowski got 11 patches merged, and have a total of 21 merged patches
         Susobhan Ghosh got 10 patches merged, and have a total of 12 merged patches
   + new easyHacks:
         None
 
   + create your own by updating Keywords to eg. "easyHack, difficultyBeginner, skillCpp, 
topicCleanup"
   + please remember, a codepointer is mandatory, and typically a look from an experienced developer
   + create new easy hacks from these ones or remove "needsDevEval" if not feasible):
       https://bugs.documentfoundation.org/buglist.cgi?keywords=needsDevEval
 
   + Wiki cleanup (BrokenRedirects, DeadendPages, DoubleRedirects, WantedTemplates, 
UncatagorizedPages, Redirects all done)
       + now ready to be moved into the new structure.
   + needsUXEval grows quite rapidly
       + tagging every enhancement as needsUXEval (Heiko)
           + eg. improve performance of file load (Michael)
           + if it is tagged enhancement and is UX related - should have it.
       + benefit from having the keyword set (Heiko)
           + goes through the ML.
       + doesn't mean blocking until needsUXEval is gone ? (Bjoern)
           + definitely not blocking (Heiko)
               + just a tool to see what's going on.
    + no new Easy Hacks in 3 weeks
       + would prefer to look at NEEDINFO ones & add code pointers:
           
https://bugs.documentfoundation.org/buglist.cgi?bug_status=NEEDINFO&keywords=easyHack&list_id=619842
 (EASYHACK & NEEDINFO)
       + any NEEDINFO easy-hacks that have no clear author ? (Bjoern)
           + can be closed automatically by mistake (JanI)
               + will check them then
               
https://bugs.documentfoundation.org/buglist.cgi?bug_status=NEEDINFO&keywords=easyHack%2C%20&keywords_type=allwords&list_id=619840&query_format=advanced&resolution=---
 
* Commit Access
    + FYI - Jaskaran Singh (GSOC) added at the mentor (moggi)'s request
 
* Developer Certification (Stephan/Bjoern/Kendy/Thorsten)
     => timeout for 9 weeks.
  
* Squeaky-clean master vs. work-in-progress changes (Thorsten, Stephan)
    + loplugin removals interfering with ongoing work / fine-granular committing
    + a bit late (Stephan)
        + sometimes code committed - that is not directly used
            + many toolchains warn about unused features, code, parameters etc.
            + plugins quite aggressive.
        + then people get build breakers
        + sometimes you tread on toes ...
        + this week happened badly for Mike
        + for plugins - would help if gerrit / jenkins-bots build with them enabled
        + plan is WIP here (Norbert)
            + back-ported them so they build vs. that machine's clang.
            + a couple of issues remaining - Jenkins re-build etc.
            + need to do 2x Linux builds instead of one.
                 + on my list for the weekend
            + 2 builds: one gcc-release and another clang+dbgutil+plugins.
                 + may be a need to annotate in source: to silence the plugin
                 + should unit test or code use be in the same code ? (Michael)
                 + stub method - plugin complain it should be static method (Miklos)
                     + if know a next patch - suggest part of review: (void)this;
            + do ping Noel if your patch need re-basing (Noel)
     + concern wrt. discouragement to newer people (Thorsten)
        + commit fine & granular to master, and then get issues.
        + concerned to fix social contract, being friendly & welcoming etc.
 
* Jenkins / CI update (Norbert)
    from:Thu Jun 23 13:40:18 2016
    master linux rel  jobs: 209 ok: 174 ko:  33 fail ratio: 15.79 % break:   9 broken 
duration:13.85%
    master linux dbg  jobs: 168 ok: 145 ko:  21 fail ratio: 12.50 % break:   8 broken duration: 
9.73%
    master mac rel    jobs: 201 ok: 201 ko:   0 fail ratio:  0.00 % break:   0 broken duration: 
0.00%
    master mac dbg    jobs: 205 ok: 205 ko:   0 fail ratio:  0.00 % break:   0 broken duration: 
0.00%
    master win rel    jobs: 173 ok: 171 ko:   2 fail ratio:  1.16 % break:   2 broken duration: 
0.68%
    master win dbg    jobs: 162 ok: 160 ko:   2 fail ratio:  1.23 % break:   2 broken duration: 
0.61%
    master win64 dbg  jobs: 171 ok: 170 ko:   1 fail ratio:  0.58 % break:   1 broken duration: 
0.39%
    lo-5.2 mac        jobs:  12 ok:  12 ko:   0 fail ratio:  0.00 % break:   0 broken duration: 
0.00%
    lo-5.1 mac        jobs:  12 ok:  12 ko:   0 fail ratio:  0.00 % break:   0 broken duration: 
0.00%
    should break:
    gerrit win rel    jobs: 377 ok: 307 ko:  69 fail ratio:18.30%
    gerrit lin rel    jobs: 378 ok: 290 ko:  87 fail ratio:23.02%
    gerrit mac rel    jobs: 379 ok: 320 ko:  58 fail ratio:15.30%
    gerrit all        jobs: 377 ok: 261 ko: 115 fail ratio:30.50%
    + odd - Linux is the bad platform this week
        + only 2x breaks on windows - amazing.
        + sadly Linux broken 10% of the time
            - used to see that for Windows.
        + success ratio getting much better for test builds
             function of a better master.
    + interesting to see what Linux numbers are caused by ? (Bjoern)
        + is this windows developers arriving ?
    + when doing plugin tests - perhaps transient Linux / base test failures (Norbert)
 
* l10n (Sophie)
    + all normal, nothing to say.
 
* QA update (Robinson)
 
    + missing Robinson
 
    + Bug Hunting Session results (Sophie)
        + rather few people, as normal - under 10 joined.
 
    + UNCONFIRMED: ???
        https://bugs.documentfoundation.org/reports.cgi?product=LibreOffice&datasets=UNCONFIRMED
        + enhancements: ??
        + in ux-advise: ??
    + needAdvice at ??
 
    + Most Pressing Bugs:
        http://tdf.io/mostressingbugs
        + tdf#100295 - LibODev-5.2.0.0.b2 does not run in Windows_XP (32 or 64 bit)
            + still pending; moggi said he'd look at it.
        + tdf#92516 - "x" accelerator and mouse click not exiting StartCenter
          in "File -> Exit LibreOffice" menu on Windows.
            + Continue to see dupe reports, an annoyance
        + tdf#100458 - Data loss on loading Calc document with hidden 0 (Jan-Marek)
            + LHM has a lot of documents with this kind of formating
            + Proposed patch: https://gerrit.libreoffice.org/#/c/26435/
            + pending review from Markus.
        + Mail merge regressions (Bjoern)
            + http://tdf.io/mmregressions (7 open; 7 open last meeting)
 
* QA stats
 
  + https://bugs.documentfoundation.org/page.cgi?id=weekly-bug-summary.html
    +142    -114        (+28 overall)
    many thanks to the top bug squashers:
        Caolán McNamara       12
        V Stuart Foote         8
        Buovjaga               7
        Heiko Tietze           7
        Adolfo Jayme           6
        Julien Nabet           6
        Justin L               5
        Jan Holesovsky         4
        m.a.riosv              4
        Samuel Mehrbrodt       4
        Aron Budea             3
        Bartosz Kosiorek       3
        Andras Timar           3
        jan iversen            3
        Tor Lillqvist          3
 
* Highest-Priority bugs (aka "MABs"):
        5.2: 5/16   - 31%   +2
        5.1: 2/29   -  6%   +0
        5.0: 7/53   - 13%   +0
        4.4: 7/73   -  9%   +0
        4.3: 5/68   -  7%   +0
        4.2: 8/133  -  6%   +0
        4.1: 4/79   -  5%   +0
        4.0: 10/82  - 12%   +0
        old: 34/248 - 13%   +0
 
* Bibisected bugs open: whiteboard 'bibisected'
  + 362/1374 364/1371 367/1371 365/1363 369/1361 366/1354 366/1352
    + http://tdf.io/bibibugs
 
* all bugs tagged with 'regression'
        + 729(-2) bugs open of 5006(+12) total 11(+0) high prio.
 
        * ~Component   count net * high severity regressions
                  Calc -  4 (+0)
                Writer -  4 (+0)
               Impress -  1 (+0)
                  Base -  1 (+0)
           LibreOffice -  1 (+0)
                + http://tdf.io/regressionscritical
 
        * ~Component   count net * all regressions
                Writer - 223 (-4)
                  Calc - 138 (-0)
               Impress -  59 (+1)
           LibreOffice -  46 (-1)
        graphics stack -  38 (+0)
                    UI -  37 (-1)
               Crashes -  33 (+1)
               Borders -  32 (+0)
                  Base -  32 (+1)
                  Draw -  27 (+1)
   filters and storage -  22 (+0)
                 Chart -  15 (+0)
                 BASIC -  14 (+0)
Printing and PDF export-  11 (+0)
        Formula Editor -   5 (+0)
            Linguistic -   4 (+0)
            Extensions -   3 (+0)
             framework -   3 (+0)
          Installation -   3 (+0)
                   sdk -   0 (-2)
                + http://tdf.io/allregressions

-- 
 michael.meeks@collabora.com  <><, Pseudo Engineer, itinerant idiot