[Libreoffice-qa] minutes of ESC call ...
michael.meeks at collabora.com
Thu Aug 27 07:56:47 PDT 2015
+ Caolan, Eike, Michael M, Norbert, Stephan, Sophie, David, Jan-Marek, Robinson,
Noel, Muthu, Cedric, Robinson, Thorsten, Michael S, Bjoern (late), Bubli
* Completed Action Items:
+ create under-the-hood wiki page to accumulate for 5.1 (DTardon)
[ page with categories at: https://wiki.documentfoundation.org/UnderTheHood/5.1 ]
* Pending Action Items:
+ look at 'make check' failures on win32 again if interested (Michael S)
+ have a look at the CppUnit to implement 'expected failure' (Jan-Marek)
[ investigated the current CPPUNIT version and Qt test suite,
will be able to hack on that the next week @DebCamp.
was busy at the conference. ]
+ 'make -k stagingcheck' prototype (Bjoern)
+ have a look at revert notification (author + committer + the person who reverted) (Norbert)
+ review Giuseppe's webdav work in gerrit (Kendy/?)
[ Merged a couple of patches (Michael) ]
+ re-arrange the help XML for the menu changes (Jay)
[ Jay was away for vacation, promised to have a look at
HelpAuthoring.oxt over the weekend (Kendy) ]
+ UserAgent - drop bundled-languages (Michael)
+ UserAgent - produce a patch for review (Michael)
+ promote Development Mentoring Lead Job Offer through all channels: Universities, Companies, Clubs, ex-GSoCers etc. (all)
[ need more applicants, help appreciated pwrt. finding experienced candidates.
application period is theoretically over. ]
+ add a conditional to disable google warable bits for freedroid (Cloph)
+ ask students to write their own wrap-ups to aggregate (Cedric)
* Release Engineering update (Robinson/Christian)
+ 4.4.6 (last of the 4.4.x line) - status
+ RC1 - week 40 (week starting 28th September)
+ 5.0.1 - RC2 status
+ Has been announced to public
+ just released it.
+ 5.0.2 - RC1 - freeze Tue/Wed
+ some translations cut off
+ some toolchain problem with some Linux distros (Rene?)
+ German help -> front-page is not translated eg. (Cloph)
+ easy to spot if it's an issue.
+ perhaps some po syntax errors tooling doesn't like.
+ Rene provided his logs, this complains about non-matching tags, but
the XML seems OK. He'll try with LO included tool chain to double-check (Christian)
+ Late OpenGL Status (Michael)
+ heading for review / merge as default for RC1
+ various bits to reduce impact
+ catching bad drivers; implemented:
+ user-friendly Windows / OpenGL disable implemented
+ just merge a .reg fragment that locks that down / off
+ in -5-0
+ detect SEGV's inside OpenGL code paths and disable GL
+ in -5-0
+ watchdog thread detects hanging OpenGL code paths & disables GL
also exits the app if dead for 5 seconds.
+ in -5-0
+ extensive legacy H/W black-list coming - or white-list ?
+ plenty more bugs in-progress.
+ OpenGL by default poke for 5.0.2
+ tripple review needed - re-visit it later.
+ lots of fixing still going on.
+ Android & iOS Remote
* UX Update (Kendy)
+ Kendy on vacation.
* Documentation (Olivier)
+ move to flat ODT progress / filter issues ?
* QA / Blocker name (Robinson)
+ come up with a sensible new name for 'blocker'
+ which doesn't imply blocking-ness.
+ decision time today.
+ ¨very-critical" suggested by Bjoern
+ fair enough -> could do this.
+ renaming right now.
+ restrict access to it.
+ do we need to rename & restrict access ? (Michael)
=> go for what you see fit (Robinson)
* Crashtest update (Caolán)
+ 2 import failures, 5 export failures, 6 coverity warnings
import failures fixed again
export failures not locally reprodible, all odb->odb
coverity warnings all fixed again
+ fftester fuzzing trajectory
the pattern is for each new filter, find most in the first hour
then still some more in first day, next to nothing on subsequent days
working through the filters & fixing them.
png & other graphic filter: 5-6 days of fuzzing, nothing.
+ file format issues we've recently had reported issues with are
in 'fast' filters, eg. 'hwp' is one of the fasted loaders we have
+ hopefully we're overtaking them now with faster loading harness
+ just catching hangs & segvs so far.
+ lots of hangs, by the boat-load
* GSoC (Cedric/Thorsten)
+ Next important dates:
+ Evaluations all done.
+ Aug. 28: final evaluations deadline
+ mentor summit: Cedric & Moggi confirmed going.
AI: + discuss the results next week.
* Hackfests (Bjoern)
+ Aarhus LibreOffice conference bits ?
+ Hackfest topic: "C++11 in LibreOffice"
+ look into easy, easy C++11 EasyHacks for that
+ like Stephans "init list ctors for UNO-Sequences", or
using type-alias etc.
+ !!!please file easy hacks if you have ideas!!! (Bjoern)
+ Hamburg, Germany: Oct 23rd -> 25th
+ would be great to see everyone there.
+ next venues / suggestions
+ Belgium another candidate ...
* l10n Update (Sophie)
+ nothing of major note.
* New Committers (Norbert)
+ Mark Hung - seems to be doing great work over a long time
+ lots of nice asian language / RTL fixes.
+ sounds good (Caolan)
+ some great patch commit msgs, beautiful bugs & unit tests.
AI: + mail him (Norbert)
* Developer Certification (Stephan/Bjoern/Kendy/Thorsten)
+ pending / shelved for now - wait 2 weeks.
* Jenkins update (Norbert
The forrest is hiding the tree somewhat, due to infra issue.
That being said, There seems to be few test misbeaving.. I should start to repport
the 'aborted build' the one jenkins kill after a (3 hours as of now) timeout.
+ not a great week overall.
+ more and more hangs (tests that don't finish - cancelled wrt. timeouts)
+ trying to identify those
+ this morning, a test was 800Mb of read & writes.
from:Thu Aug 20 12:56:27 2015
master linux rel jobs:154 ok:146 ko:8 fail ratio:5.19% break:6 broken duration:4.87%
master linux dbg jobs:90 ok:81 ko:9 fail ratio:9.99% break:9 broken duration:11.47%
master mac rel jobs:146 ok:140 ko:5 fail ratio:3.42% break:3 broken duration:6.83%
master mac dbg jobs:136 ok:128 ko:7 fail ratio:5.14% break:5 broken duration:9.61%
master win rel jobs:80 ok:66 ko:13 fail ratio:16.24% break:7 broken duration:39.72%
master win dbg jobs:133 ok:93 ko:40 fail ratio:30.07% break:17 broken duration:31.94%
master win64 dbg jobs:124 ok:67 ko:57 fail ratio:45.96% break:6 broken duration:32.57%
lo-4.4 mac jobs:19 ok:17 ko:2 fail ratio:10.52% break:1 broken duration:4.77%
gerrit win rel jobs:207 ok:100 ko:105 fail ratio:50.72%
gerrit lin rel jobs:213 ok:188 ko:25 fail ratio:11.73%
gerrit mac rel jobs:211 ok:167 ko:44 fail ratio:20.85%
gerrit all jobs:217 ok:87 ko:128 fail ratio:58.98%
+ New hardware updates
we have 2 new Windows online tb72 and tb73. both turn around a full
windows build in ~30 minutes (when the tests behaves)
16 logical cores each & 64Gb of RAM, some SSDs too.
I also brought extra linux capacilty in the form of an ec2 instance (tb74)
all concentrated on 1x physical box.
+ make check on win32 chestnut.
no time yet; some CLI thing - race condition opening file (Norbert)
tried to reproduce it locally, opened read-only, no locking (Michael S)
only occurs rather irregularly (Norbert)
process monitor ? (Michael S)
+ what is up when the error happens.
perhaps a different .Net ver uses different locking flags on files (Norbert)
atomic re-names in makefiles ? (Michael)
better to find the cause (Michael S)
+ Miklos' bibisect repo -> needs to move - swapping to death.
* Perf-testing (Norbert)
+ beautiful new bits:
the code written to handle it. a postgress db
managed by a bash program perfdbmng
the client is loperf_suite, that exploit a configuration
to run valgrind test and repport the result to the database via ssh
perfdbmng also have a set of command to generate the web-site above
work in progress to reorg/rename/classify and documents the existing set of test
that have been running in the callgrind vm so far.
+ each view is the cestimate - using the algorithm
more detail when you drill-down.
overall geomean possible ? (Michael)
* 4.3 MABs - what's the plan with them ? (Robinson)
+ Re-check and prioritize -> will be prioritized.
AI: + get new queries for MAB-like bits for each release (Robinson)
+ Little tricky to capture all; just highest-prio for release series doable
+ MABs were hard-partitioned into groups for each release series.
were partitioned out into each release series.
+ please do send a description, and some sample queries.
* Easy Hacks:
+ new easy-hacks thanks to:
+ Michael Meeks x3
+ create your own by setting whiteboard to eg.
"easyHack difficultyBeginner skillCpp topicCleanup"
* QA update (Robinson)
+ UNCONFIRMED at 485
+ needAdvice at 15
+ Most Pressing Bugs:
+ can we filter NEEDINFO out of this ? (Michael)
+ Yep: http://ur1.ca/nkp0t
+ tdf#92478 - Scanning crashes on 64bit Windows 10 build
+ Still present
+ tdf#93600 - Condition Formatting: Entering "=" in comparison value crashes program
+ Another crash, Bubli looked into it.
+ tdf#93437 - Once the Function Wizard has been Displayed, it is
not Possible to Select Columns or Rows or to Open
the Context Menu of Columns and Rows
+ fixed. Thanks to Bubli.
+ tdf#91882 - UI: Dialogue for Advanced filter is transparent.
+ persists with 5.x builds
=> poking (Michael)
* Bibisected regressions (Bjoern)
+ 25 open 5.0 bibisected regressions
+ dug through the writer ones
+ couple of Miklos / formatting bits.
+ many other rendercontext issues
+ not like that in other apps.
+ are they marked to the tracker (Michael)
+ please mark as blocking tdf#91488.
+ 1000+ regressions bibisected since the start
+ a bit more - if count those 'bisected'
AI: + find & reward / promote whomever did the 1000'th (Robinson)
+ volunteer at Texas Linux Fest (Ryan Nowakowski)
+ interested in writing the windows bibisect frontend: awesome.
* QA stats
+140 -100 (+40 overall)
many thanks to the top bug squashers:
Xisco Faulí 18
Maxim Monastirsky 13
V Stuart Foote 7
Joel Madero 6
Michael Meeks 4
Katarina Behrens (CIB) 3
Matthew Francis 3
Cor Nouws 3
Alex Thurgood 2
David Fischer 2
Pierre Fortin 2
Julien Nabet 2
* Open 5.0 MAB
+ 1/12 1/12 1/12 1/12 1/12 1/12 1/12 1/12 2/12 2/12 3/12 5/11
* Open 4.4 MAB
+ 11/60 11/60 11/60 11/60 10/60 10/60 10/60 10/60 12/60 14/61 16/61
* Open 4.3 MAB
+ 59/168 59/168 59/168 61/168 61/168 61/168 62/168 62/168 64/169 65/169
* Bibisected bugs open: whiteboard 'bibisected'
+ 345/1020 338/1006 328/991 314/964 309/956 311/951 305/936 305/931
* all bugs tagged with 'regression'
+ 616(+5) bugs open of 4137(+25) total 40(-2) high prio.
* ~Component count net * high severity regressions
Calc - 16 (-1)
Writer - 9 (+0)
LibreOffice - 7 (+0)
Impress - 5 (+0)
UI - 1 (+0)
filters and storage - 1 (+0)
Base - 1 (+0)
Draw - 0 (-1)
* ~Component count net * all regressions
Writer - 208 (+1)
Calc - 124 (+5)
Impress - 47 (-1)
Crashes - 43 (+2)
LibreOffice - 38 (+1)
UI - 37 (-4)
Base - 26 (+0)
graphics stack - 24 (+0)
Borders - 23 (+0)
Draw - 21 (+0)
filters and storage - 13 (+0)
Chart - 11 (+1)
Printing and PDF export- 10 (+0)
BASIC - 7 (+0)
framework - 5 (+0)
Linguistic - 4 (+0)
Formula Editor - 4 (+0)
Extensions - 2 (+0)
michael.meeks at collabora.com <><, Pseudo Engineer, itinerant idiot
More information about the Libreoffice-qa