Wednesday 09 Feb 2011

People Present

People Present

  • JamieBennett

  • dmart
  • hrw
  • jcrigby
  • ppearse
  • wookey
  • tgall_foo
  • kunal
  • aviksil

Regrets

  • slangasek

Agenda

  • Review action items from last meeting
  • Thanks for testing alpha-2
  • Blueprint status
  • Blueprint

    Assignee

    Status

    other-linaro-n-profiling-tools

    aviksil

    performance-linaro-n-memory-footprint

    dmart

    packageselection-foundations-n-multiarch-support

    slangasek

    other-linaro-n-cross-compilers

    hrw

    other-linaro-n-u-boot-upstreaming

    jcrigby

    other-linaro-n-cross-compilation-environment

    wookey

    other-linaro-n-lttng

    aviksil

    multimedia-linaro-package-meego

    tom-gall

    multimedia-linaro-standardize-omx-core

    goelkunal

    other-linaro-n-gdbserver-integration

    tom-gall

    multimedia-linaro-seeds-1105-for-multimedia

    tom-gall

    multimedia-linaro-seeds-1105-for-graphics

    tom-gall

    graphics-linaro-package-kwin

    tom-gall

Action Items from this Meeting

  • JamieBennett to come up with test assignments for the new developer image

  • wookey to work with tgall_foo to get the linaro-nano image building
  • tgall_foo and JamieBennett to get linaro-desktop building in Offspring

  • slangasek and fturgis to decide whether or not to move to more recent version of systemtap
  • ppearse to investigate how libtool does ldopen for GObject Introspection work

Action Items from Previous Meeting

  • JamieBennett to come up with test assignments for the new developer image

  • jcrigby to write up landing team u-boot, kernel workflows and send for review after first draft is done: carried over
  • slangasek to review https://wiki.linaro.org/Platform/DevPlatform/Specs/MemoryFootprintMethodology for dmart: DONE

  • ppearse to figure out what version number we want for 'helios' to be in natty
  • dmart to send a beagle kernel config to aviksil1
  • aviksil1 to send pointer to lttng-enabled tree to slangasek: DONE

Status Reports

Avik Sil

  • Worked on building and running LTTng patched Linaro kernel. Successfully built and boot the kernel with help from Dave Martin. Build of lttng-modules and ltt-control yet to be done.
  • Verified functionality of oprofile package in Linaro.
  • Tested Linaro 11.05 natty alpha-2 Developer image on Beagle Board.

Peter Pearse

Progress

  • arm-m-ael-alip-evaluation
    • Two merge proposals raised
    • Wrote notes on wiki re gobject introspection crossing esp possible problems
  • other-linaro-n-eclipse-cdt
    • Looked at version differences between natty & upstream

Issues

  • eclipse packages in natty do not support cross building

Plans

  • arm-m-ael-alip-evaluation
    • More Natty cross builds & propose merges

    • Assist ARM internal cross build of LAMP stack
  • other-linaro-n-eclipse-cdt
    • Agree new work items with Steve
    • Import existing projects
    • Build eclipse with xdeb

Absences

  • None planned

Kunal Goel

  • Merged changes for auto test and register binary to libomxil-bellagio in natty
  • libomxil-components also added to natty
  • Working on libmeegochat but build failing due to change in version of libtelepathy-qt4-dev in natty

Dave Martin

Activity

  • hardware-n-kernel-standard-architecture
    • Upstream link-time discard versus SMP_ON_UP problem still not fixed.
      • The x86 solution to solve this problem is --er-- not to solve it, i.e., .exit code is linked in and discarded at runtime. Which suggests that the current proposal for arm is at least no worse.
      • Hacked gas to make it possible to put fixups in sensibly-named sections, which would enable this problem to be fixed properly. Posted to the binutils mailing list, but no response so far.
    • omap3/4 Thumb-2 compatiblity patches re-posted. Generally positive feedback, and some positive test results.
    • SMP_ON_UP doesn't work with THUMB2_KERNEL on Beagle xM in some trees, such as tmlind and khliman's upstream omap trees. Cause unknown so far.

Plans

  • Figure out what's going wrong with SMP_ON_UP on Beagle xM.
  • Tidy up the proposal for link-time section discard avoidance and report.
  • Discuss prioritisation for outstanding kernel tasks.
  • Write up the output from the Freescale i.MX BSP review discussion, and post for comment.

Absences

None planned.

Jamie Bennett

John Rigby

  • U-Boot
    • Finished a new version based on latest upstream rc with packaging in bzr.
  • Kernel
    • Finished a new version based on latest linux-linaro-2.6.37. This will probably be the last one based on 2.6.37. Need to ping Nico about moving to 2.6.38 as he said he would in mid February.
    • This new kernel has a u8500 flavour that builds but is untested. I tried adding samsung but it failed to build with CONFIG_PM on and that is now an enforced config option in the Linaro kernel as requested by PM working group. I may need to remove the enforce.

Tom Gall

  • kwin - Jammy reports GLES support has been merged, requesting that packaging begin
  • libjpeg-turbo, libjpeg62-neon - still pending PPA discussion (I think?)
  • linaro-nano - have been hacking a new variation in the busybox build that is more suited to what we'll want in nano. nano and headless has forked and headless is no more as a stand alone build
  • linaro-alip - working on fixing the --hostname option in livebuilder (it's hardcoded currently) (712031)
  • linaro-netbook part II (aka linaro-desktop) - created as a fork of the upstream ubuntu desktop that uses unity. Works, needs a little linaroization yet. (Missing auto login or account creation currently)
  • Tested all images on my panda and beagle XM boards. Lack of gfx drivers (707038) and not being able to use all 1G hurt (633227). Lack of EDID support (660604) in all builds defect has existed since maverick and it's marked low priority yet it affects usability. (netbook on a 1280x1024 LCD is unusable as a for instance and there's no reasonable mechanism to adjust the kernel params)
  • meego - working on last set of feedback from slangasek. Really would like to get a set of pkgs around the compositor to jammy before end of this week.
  • chromOS window manager - no action

Frederic Turgis

  • identified kernel files required for native usage of systemtap (kernel instrumentation only, not userspace):
    • .config: systemtap checks if mechanisms like Kprobes are present -> link it from /boot

    • several files like Module.symvers for kernel module compilation -> available in kernel headers package

    • vmlinux: used to check kernel functions probed in systemtap scripts -> available in ddeb package (but steve indicated ddeb compilation seems broken, Loic requested it to be fixed)

  • done some kernel recompilation (help request sent on linaro-dev + IRC were probably not clear enough about my needs):
    • native: too long, I stopped :-)

    • cross-compilation with debuild: worked OK but I cheated a bit so that systemtap can compile kernel module on target.
  • systemtap tests:
    • v1.3: usage of global variables in scripts is broken since kernel 2.36rc2. Known issue (http://sourceware.org/bugzilla/show_bug.cgi?id=11940), fix available. Tested OK with cross-compiled version of v1.3 (as it was already available on my desktop)

    • v1.4: translation to C code generates 64 bits values, which are truncated to 32bits. systemtap passes option to compiler to handle warnings like errors so script does not compile
  • Next:
    • recompile v1.3 with fix.
    • find v1.4 option to handle warnings as warnings and not errors. In the long run, translation to C code has to be fixed. Issue has already been seen and corrected in the past

some intro to systemtap:

  • systemtap allows basically dynamic instrumentation of kernel and userspace functions by leveraging existing mechanisms (Kprobe, tracepoints, utrace, ...). It performs following tasks:
    • identify instrumented functions from script
    • translate script to C code
    • compile C code as kernel module
    • run script (userspace uses relayfs and in the future ring buffer to get trace from kernel module)

Meeting Log

Meeting opened by JamieBennett at 16:02

https://wiki.linaro.org/Platform/DevPlatform/2011-02-09

  • <JamieBennett> So, a quick meeting I suspect, lets go over the past week action itesm

    <JamieBennett> oopps items

    <JamieBennett> [TOPIC] Action Item Review

Action Item Review

  • <JamieBennett> JamieBennett to come up with test assignments for the new developer image

    <JamieBennett> my bad, still need to do this, c/o

    <JamieBennett> jcrigby to write up landing team u-boot, kernel workflows and send for review after first draft is done

    <JamieBennett> jcrigby: ?

    <JamieBennett> OK, lets move on

    <JamieBennett> slangasek to review https://wiki.linaro.org/Platform/DevPlatform/Specs/MemoryFootprintMethodology for dmart

    <JamieBennett> dmart: did this happen?

    <tgall_foo> JamieBennett, wrt the developer image, qa tracker does show that 4 variations were tested .. so I think it was actually done :-)

    <dmart> JamieBennett: yes

    <JamieBennett> tgall_foo: yes, I'll assign 'official testers' though as well ;)

    <JamieBennett> dmart: thanks

    <JamieBennett> [ACTION] JamieBennett to come up with test assignments for the new developer image

JamieBennett to come up with test assignments for the new developer image

  • <dmart> JamieBennett: he was happy with it, apart from a couple of minor suggestions

    <tgall_foo> JamieBennett, chalk it up to well executed mind control ;-)

    <JamieBennett> dmart: did you incorporate the suggestions or do you still have to do that?

    <dmart> JamieBennett: I made the changes; I don't think there's outstanding work on me for that.

    <JamieBennett> OK

    <dmart> (unless slangasek can think of something else)

    <JamieBennett> :)

    <JamieBennett> ppearse to figure out what version number we want for 'helios' to be in natty

    <ppearse> https://spreadsheets.google.com/a/linaro.org/ccc?key=0AnPR4S1Uev7KdFV2STV3VF9DUF9YQWI1TS1xTmx0d2c&authkey=CO_4s78P&hl=en&ndplr=1#gid=0

    <ppearse> hopefully shared

    <JamieBennett> yep, I see it

    <tgall_foo> yup works for me

    <JamieBennett> thanks ppearse

    <JamieBennett> dmart to send a beagle kernel config to aviksil

    <aviksil> done

    <JamieBennett> awesome

    <JamieBennett> aviksil to send pointer to lttng-enabled tree to slangasek

    <aviksil> this too is done :)

    <JamieBennett> just me and jcrigby letting the team down then ;)

    <JamieBennett> A quick thank you to all that tested Alpha-2, please keep up the testing. I'll be setting up qatracker tomorrow for weekly testing if you can spare an hour or so to go over images

    <JamieBennett> [TOPIC] Blueprint status

Blueprint status

  • <hrw> JamieBennett: ordered new sd cards to be able to test images ;D

    <JamieBennett> So ordering of status

    <JamieBennett> tgall_foo jcrigby dmart fturgis ppearse JamieBennett kunal aviksil wookey slangasek hrw

    <JamieBennett> tgall_foo: your up

  • tgall_foo snaps to attention

    <tgall_foo> kwin - GLES support has been merged, requesting that packaging begin

    <tgall_foo> (from Jammy)

    <tgall_foo> linaro-nano - have been hacking a new variation in the busybox build that is more suited to what we'll want in nano. nano and headless has forked and headless is no more as a stand alone build

    <tgall_foo> linaro-alip - working on fixing the --hostname option in livebuilder (it's hardcoded currently) (712031)

    <tgall_foo> linaro-netbook part II (aka linaro-desktop) - created as a fork of the upstream ubuntu desktop that uses unity. Works, needs a little linaroization yet. (Missing auto login or account creation currently)

    <tgall_foo> Tested all images on my panda and beagle XM boards. Lack of gfx drivers (707038) and not being able to use all 1G hurt (633227). Lack of EDID support (660604) in all builds defect has existed since maverick and it's marked low priority yet it affects usability. (netbook on a 1280x1024 LCD is unusable as a for instance and there's no reasonable mechanism to adjust the kernel params)

    <tgall_foo> meego - working on last set of feedback from slangasek. Really would like to get a set of pkgs around the compositor to jammy before end of this week.

    <tgall_foo> chromOS window manager - no action

    <tgall_foo> libjpeg-turbo, libjpeg62-neon - still pending PPA discussion (I think?)

    <tgall_foo> questions ?

    <wookey> is linaro-nano an ubuntu on busybox base?

    <tgall_foo> wookey, yes I'm coming at it such that it'll be busybox based

    <JamieBennett> tgall_foo: any eta on the -destop image and similarly -nano?

    <wookey> wahey. We have a lot of patches in emdebian-crush that you might find helpful

    <tgall_foo> wookey, tho the other side of the problem is we really need to get the hwpack (and all the kernel modules) on a diet

    <wookey> there is an awful lot of stuff that breaks if coreutils is replaced

    <tgall_foo> JamieBennett, I have desktop successful built and bootting ... tho the unity desktop isn't running right yet ... tracking down setup to make sure that's correct

    <wookey> very happy to help make that stuff get done though

    <JamieBennett> [ACTION] wookey to work with tgall_foo to get the linaro-nano image building

wookey to work with tgall_foo to get the linaro-nano image building

  • <tgall_foo> wookey, that'd be great

    <wookey> getting it so that busybox is overlayable with necessary full packages is tricky'

    <hrw> wookey: you need huge amount of update-alternatives

    <hrw> wookey: we did that this way in openembedded

    <wookey> that's one way.

    <hrw> but in OE busybox is default

    <tgall_foo> hrw, yeah gentoo has a similar approach

    <wookey> simply overlaying with real packahge is nicer if you can make it work. But I guess this should be offline discussion

    <hrw> never used gentoo

    <tgall_foo> wookey, yup agreed

    <tgall_foo> anything else ? if not, EOL

    <JamieBennett> so tgall_foo, linaro-desktop today/tomorrow?

    <tgall_foo> JamieBennett, I think that's in the realm of the possible

    <JamieBennett> OK, let me know and I'll add it to Offsping

    <JamieBennett> [ACTION] tgall_foo and JamieBennett to get linaro-desktop building in Offspring

tgall_foo and JamieBennett to get linaro-desktop building in Offspring

  • <JamieBennett> thanks tgall_foo

    <JamieBennett> jcrigby: Your next

    <JamieBennett> no jcrigby ?

    <JamieBennett> OK, we can loop back around later

    <JamieBennett> dmart: Shoot

    <dmart> JamieBennett: for dev platform, the memory footprint methodology writeup was the main bit of progress

    <dmart> Other than that, I've been continuing to hack Thumb-2 patches for the omap kernel

    <JamieBennett> hows that going?

    <dmart> Mostly good. Finally, I'm able to test suspend/resume properly, and it turns out that that doesn't work correctly in Thumb-2 ... which is actually not too surprising

    <dmart> But I think fixing that should be fairly routing

    <dmart> routine

    <JamieBennett> great

    <JamieBennett> anything else dmart ?

    <JamieBennett> guess not

    <JamieBennett> jcrigby: Your next

    <JamieBennett> doh

    <dmart> JamieBennett: ^ no, that's all from me

    <JamieBennett> fturgis: Your next

    <JamieBennett> :)

    <aviksil> dmart: do you have any update on the SMP_ON_UP fix, as you said it was not a conclusive fix?

    <fturgis> letting aviksil and dmart finishing

    <JamieBennett> fturgis: carry on, they can inline

    <fturgis> OK. had to recompile kernel because of missing kernel ddeb (vmlinux need) :-(

    <fturgis> tested systemtap v1.3 with internal scripts like tracking scheduler context switches

    <dmart> aviksil: no update yet ... I've been working on the omap stuff

    <fturgis> systemtap v1.3 has known bug with kernels > 2.6.26rc2 if using global variable in script -> fix available in redhat git tree and tested OK

    <aviksil> dmart: OK

    <fturgis> I also tested systemtap v1.4 from Redhat git. Because of known bug, shall we move to more recent version or patch v1.3 ?

    <JamieBennett> That will need some thought

    <JamieBennett> [ACTION] slangasek and fturgis to decide whether or not to move to more recent version of systemtap

slangasek and fturgis to decide whether or not to move to more recent version of systemtap

  • <JamieBennett> lets discuss that later

    <fturgis> that's all (kernel compilation took time)

    <JamieBennett> OK, thanks fturgis

    <JamieBennett> ppearse

    <ppearse> arm-m-ael-alip-evaluation

    <ppearse> Polishing patches & submitting merge proposals

    <ppearse> Convinced nyseld we cant cross build gobject introspection into packages

    <ppearse> s/nyseld/myself/

    <wookey> ppearse: I notice libtool has some special magic for doing ldopen on foreign things

    <wookey> I wonder if that can be made to help with the GI thing

    <ppearse> OK - I'll have a look

    <wookey> need a good hard think and really hot cup of tea that one

    <ppearse> That'll be tomorrow then

    <JamieBennett> [ACTION] ppearse to investigate how libtool does ldopen for GObject Introspection work

ppearse to investigate how libtool does ldopen for GObject Introspection work

  • <wookey> oops, giving you extra actions!

    <JamieBennett> :)

    <ppearse> other-linaro-n-eclipse-cdt

    <ppearse> Spreadsheet as per link

    <ppearse> There's a debian bug that I'll track http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=587742

    <ubot2> Debian bug 587742 in eclipse "eclipse: Please package new version of Eclipse 3.6" [Wishlist,Open]

    <ppearse> And enquiring on Java Maintainers about e.g. eclipse-cdt

    <ppearse> Installed Helios DS-5 in my natty VMWare & started watching/following the videos

    <ppearse> Should be able to spam efika about it next week

    <ppearse> "DS-5 with efika as target"

    <hrw> ppearse: need debugboard for efika to use it with ds-5?

    <ppearse> No - I'll use gdbserver/rse

    <ppearse> Wan't a story for low cost development

    <hrw> ppearse: discuss with PaweB Moll - he is using ds-5 with efikasb

    <ppearse> hrw: Ack

    <ppearse> Thats all from me

    <ppearse> (Before I get more actions)

    <JamieBennett> :) OK, thanks ppearse

    <JamieBennett> skipping me as there is nothing to report for DevPlatform work really

    <JamieBennett> kunal:

    <kunal> yes

    <kunal> merge proposal for libomxil accepted

    <kunal> changes for dh_auto_test and binary for registering components added in natty

    <kunal> libomxil -components accepted by natty

    <kunal> worked on libmeegochat review changes

    <kunal> but it's failing due to change in version of libtelepathy-qt4-dev in natty

    <kunal> from the time it was added to ppa

    <kunal> some changes in class structure, will try to fix it

    <kunal> that's all from my side

    <JamieBennett> OK, thanks kunal

    <JamieBennett> aviksil

    <aviksil> Worked on building and running LTTng patched Linaro kernel

    <aviksil> Successfully built and boot the kernel with help from dmart

    <aviksil> built and installed lttng-modules and ltt-control on ARM. Got oops on running lttctl - issue reported to Mathieu Desnoyers

    <aviksil> built ltt-bin deb package for armel from ltt-control source package in ubuntu

    <aviksil> Tested Linaro 11.05 natty alpha-2 Developer image on Beagle Board

    <aviksil> that's all from me

    <JamieBennett> OK, thanks aviksil

    <JamieBennett> wookey:

    <wookey> xdeb cyclic dependency analysis - appears to include too much, but not clear why - will post patches as-is for now and worry about bugs later

    <wookey> xdeb patches still not properly sorted. Still trying to get that done but people keep pestering me about other stuff

    <wookey> Some emdebian/arm sprint planning to get pre-discussions right (especially for people not present)

    <wookey> FOSDEM. Gave talk on armhf port (at short notice when holger was poorly)

    <wookey> Lots of interesting stuff as ever

    <wookey> Refined xdeb behaviour spec on train - will circulate for review.

    <wookey> Still thinking about how the bootstrapping tool should work

    <wookey> Found major cross-build problem in libtool - almost finished big write-up for debian-devel and linaro-dev

    <wookey> I appear to be the 6th person to find it :-)

    <JamieBennett> :)

    <wookey> OK, that's all. Any questions?

    <ppearse> wookey: What area?

    <wookey> modew=relink

    <ppearse> ack

    <wookey> when installing packages in debian/tmp/foo

    <wookey> s/packages/libraries/

    <JamieBennett> sounds like a busy week wookey, thanks

    <JamieBennett> hrw:

    <wookey> they all are :-)

  • hrw present

    <JamieBennett> :)

    <hrw> armel-cross-toolchain-base hacked badly for debian and nearly builds

    <hrw> - nscd/others needs patching to build (I know how to build by hand it when it fails)

    <hrw> gcc-4.5-armel-cross just builds under debian - no changes needed

    <hrw> ia64-cross-toolchain-base gave me some fixes due to libc6.1-dev not libc6-dev

    <hrw> but it fails anyway cause bootstrap is more complicated - gcc-4.5 needs libuwind :(

    <hrw> will try to get hppa cross built instead

    <hrw> those non-armel targets are just to test does my package works properly ;D

    <wookey> :-)

    <hrw> working on 'script to provide tarballs with toolchain for unpack to /opt/linaro

    <wookey> youneed to explain to me why binutils cross stuff isn;t built separately from rest of toolchain sometime.

    <hrw> binutils + gcc-4.5 are ok, eglibc is now building and will need changes (longterm) or hack (shortterm)

    <hrw> thats all from me

    <hrw> questions?

    <JamieBennett> OK, thanks hrw

    <JamieBennett> [TOPIC] AOB

AOB

  • <wookey> superfast meetings from Mr bennett :-)

    <JamieBennett> :)

    <hrw> wookey: ubuntu binutils lacks sysroot support which I need during bootstrap. so armel-cross-toolchain-base builds binutils/sysroot + binutils/ubuntu

    <JamieBennett> Anything else to discuss before I close?

    <JamieBennett> going ...

    <JamieBennett> going ...

    <ppearse> Not from me

    <wookey> hrw: you don't need sysroot to build cross-binutils

    <wookey> no, ignore us :-)

    <JamieBennett> #endmeeting


Meeting closed at 16:52

Platform/DevPlatform/Meetings/2011-02-09 (last modified 2011-03-11 11:35:49)