olpc software meeting notes for Tue Nov 28th, 2006

[Christopher Blizzard]

Software Meeting 11/28/2006

Attending:

Richard Smith
Lillian Walter
Jordan Crouse
John Palmieri
Mitch Bradley
Ted + Victor from Quanta
Marcelo Tosatti
Ivan and Chris Ball

Mitch running the meeting
Chris taking notes

Action Items:

AI: Victor will talk to Mark about quantity for Beta 2
AI: Blizzard to see where Beta 2 units are going (developers, countries)
AI: Mitch to own power project for Beta 2
AI: Chris Ball to own performance
AI: Chris Blizzard to own stability
AI: BIOS team to deliver working BIOS bits tomorrow (nov 29)
AI: John Palmieri to work on image updater with new BIOS (nov 29)
AI: Blizzard to find out from Lindsay how many we're sending out
AI: Jordan to work on getting DCON and X hacks out and get real fixes in 
now that Beta 1 is done
AI: Blizzard to talk to Andres about new kernel feature (.19)
AI: Victor to tell Mitch what chips they are using and send samples to 
Mitch for ethernet on the line
AI: Lots of bug triage for everyone

o Schedule for B2
- Integrated Verification Test by Jan 15 minus a week for Christmas vacation

o Should use trac for btest-2

o What outstanding items will we have by beta2?
- Victor: Mark says 500 units for beta2
- Victor will discuss with Mark next week to double check
- Not sure that they will have units to us by jan 15th, that's when the 
quanta testing team gets first units
- Those units will go where?
   - No distribution list yet - not sure where they are going to go
   - Chris: guessing some refreshes for developers, but mostly for countries

o Power Management for beta2 - want to have something to demo
o Performance, stability and power management - focus points for the 
next 6 weeks
o Beta2 builds will get faster python, which means faster sugar
o But with half the memory
- Python 2.5 has better memory management, if we go that route

o Preview of document stuff in place by beta2?
- Should have a working eBook reader by the end of this week?  John is 
coming into the office and has a working reader.  Will require 2-5 of 
integration work before it's ready to be put into the build.

o ROM security is not a priority for the next build, should target at beta3
   - Prototyping as available, but not pushing for beta2
   - Should have crypto bits for the bios by Nov 30th (code to see if an 
image is properly signed.)
   - Last bit is EC changes to clear the write protect bits?  Not 
tested, but Mitch thinks it is in.

o Blessed software bits?
- Richard has built what is supposed to be the right stuff, still has 
the brick.  Waiting on that.
- Software?  Any testing time on 187?
   - Haven't done any testing on that yet.
   - Chris Ball has a keyboard problem of some kind?
   - John thinks it's fine and is running it on his box.
   - It checks for a-test it doesn't load the module and uses a 
different xorg.conf.
   - Only differences between 185 + 187 is the module loading and the 
xorg.conf
   - What other testing do we need to do?
   - Load them on a few more machines and make sure that the i80* driver 
is loaded and the right xorg.conf is copied over.
   - Need to do some functional testing + tinderboxing + boot it on an 
a-test
   - Make sure it works with the new buildrom update and make sure that 
it leaves the manufacturing data is intact
     - at the OK prompt and there's a one line check for it

- Also need to generate an auto-update image that flashes the BIOS and 
push the right build onto the nand
   - Doesn't have to be the same build
   - Might need a > 1G key - can use the little seagate drives
   - John has done it in the past with a manual step, but we can hook 
that up to an init script
   - Get a minimum build with a 512 key if we want
   - Want to make sure
   - Need to talk to lindsay about how many go out tomorrow
   - Maybe pull olpcflash out of the build?  Maybe not
   - Blizzard won't be in the office, John volunteers to sign up to run 
this process  WE HAVE A WINNER!

o Ramping back up?
- Get update stuff out of the way and then get serious about power 
management - Mitch will drive that once we're done
- Chris will be driving Performance goals
   - Baselines for performance need to be established
   - Track progress - Chris will be driving some of that with new tbox stuff
- Stability - Chris Blizzard will own
   - Hard to measure
   - Memory Management (also can measure)
   - Process-wise making sure that things don't break over time
   - Smoketesting Procedures - running through a set of tests
   - Over time testing - 100 most popular pages and see how long we last
   - Need to automate sugar somehow
- Milestones for progress for build time and when we will do feature freezes

o Cleanup from btest1
- Unscrew A-test
   - Need to work on an updated a-test image at some point
   - Not in bad shape
- Wireless on A-test?
   - Wireless was having problems (from Richard?)
   - Needs more investigation
   - don't know what boot firmware was on it, need to investigate
   - Richard will test
- Cleaning up a bunch of apps
- X needs to be fixed
   - Lots of fixes for btest-1 that need to get cleaned up post-btest
   - Some DCON fixes as well

o Major pieces to include
- New kernel? (2.6.19?)
   - Job for Andres!
- New Python (2.5?)
- Let's document in trac or spreadsheet
- AI: Blizzard to talk to Andres for new kernel, etc

o Network support for manufacturing?
- High priority for beta test 2?
   - Would be good to test for beta2
   - What about the slow ethernet?
     - AI: Victor will tell Mitch the chips that they use
     - AI: Victor needs to send specific parts to Mitch for building/testing

o Ivan: looking at trac we have a number of bugs that are still 
targetted at old test builds
- Need to make sure that they are retargetted or closed at some point

o Bug triage

27: stress tester: cjb did an image for this and gave it to tests at quanta
   - Will close and will continue to enhance the testing
   - Quanta does have manufacturing testing and their own procedures
   - AI: Ivan: Will close with a comment that it was delivered to Quanta

43: wireless mesh software: we do have this and there's driver support
   - AI: blizzard: need to add a comment that it needs testing

270: USB high speed issues: engineers are still looking at those on 
pre-b boards
   - AI: David Woodhouse: See if the problems still exist on btest2
   - AI: Chris will retarget a btest2

272: 273: 274: 275: all same problem, chris will retarget at btest2
   - 273 should be reassigned to jordan

280: delay in dcon -> gpu transition: hasn't been touched since nov 6
   - Have a fix at amd but didn't want to add right before
   - AI: Jordan will check with David Woodhouse to see if we need to get 
the right fix into the right place
   - Will affect power management as well, need check (turning off panel 
earlier rather than later)

298: Wireless device appears + disappears: from comments mfoster 
believes this is a problem with the known power issues in beta1?
   - AI: need to ask david woodhouse and retarget at beta2

409: high number buttons as primary: owned by warp: jg says this is 
solved by new evdev that's in 172 needs testing by someone before closing
   - Needs testing: AI: retarget to beta2 and need to verify fix

454: wireless oops: dan williams:
   - AI: change owner to marcelo and change to beta2

455: OF failed to set device address: 1: new firmware might solve this, 
but solve this
   - AI: Chris will retarget to beta2 and ask marcelo to test with 
hardware and new firmware

331: sort out keymap: getting keymaps into X
   - Keymaps should all be in, need to be tested on the actual keyboards
   - AI: retarget to beta2 and add a comment to test this

461: overwriting manfacturing data
   - fixed in new image
   - olpcflash will be fixed after we get the update image out the door
   - move to beta2

213: hacking the kernel
   - beta2, move to andres

420-424 for various keymaps
   - Close these for 331 - we do have all the keyboards

512: beta test 1b auto update image
   - Should be done by tomorrow

446: sugar keys not working
   - retarget for beta2 with comment about only camera not working

476: manfacturing data values are in wrong order
   - Close it, it's fixed