System Software Meeting Minutes, 2006-10-25

[Chris Ball]

System SW Meeting Minutes
2006-10-25

Chris Blizzard, Mitch (moderator), Jim, Chris Ball (scribe),
Zephaniah, Richard, Jordan, Ray, Ted, Marcelo.


*** New action items:

AI: Chris Ball to tell Mitch which build to use on DCON
AI: Chris Ball to get new LB image to Mitch for 256M testing
AI: Chris Ball to backup DCON mode-changing software
AI: Chris Ball to announce new BIOS to software-team at +Quanta for testing
AI: Blizzard to push a new build tomorrow, test and pronounce stable
AI: Blizzard to test csound packaging with Barry Vercoe
AI: Blizzard to find out dependencies on mplayer, use it to demo camera
    * if too many dependencies, build ourself and put in our own repo
AI: Blizzard to talk to AbiWord guys about non-GNOME AbiWord build
AI: Everyone with a DCON (Mitch, Cambridge and Jordan) to test BIOS.
AI: Jordan to put Jim in touch with Geode team.
AI: Marcelo to integrate SDHCI driver with Fedora build.


*** Existing action items:

Blizzard:
 - sysprof module has been added to kernel, is untested
 - If the kernel module works, we can `yum install sysprof`
 - Haven't played with memprof
 - oprofile may come for free as a result of sysprof
 - Adam Jackson found a workaround for 16bpp, it's in the new build, so
   we're back running at 16bpp

Mitch:
 - Mitch to send bad board to Quanta?  (old AI)
   * done
 - Mitch to deliver a new LB release with fixes
   * done
 - Mitch: supply Open Firmware for people to start testing
   * done, waiting testing
 - Mitch and Chris Ball to do a PRS test that runs in tinderbox
   * not done

Andres:
 - add verified state to Trac
   * not done

Ray:
 - tell Mitch how to detect a 3700 vs a 3920:
   * done
 - release EC version 20 next week with quiet mode, battery and AUX
    multiplexing
   * done

Everyone:
 - mail Ivan to get an internal wiki account if necessary
   * still open

Richard:
 - 113: LinuxBIOS flashing procedure needs additional step
 - Richard also needs to fix something with olpcflash
   * not done, will do tonight

Zephaniah:
 - Input: we have a functional kernel driver that needs minor tweaks
   before sending upstream.
   * not upstream, will happen this week.
 - Not quite ready because of btest hardware stuff
   * touchpad has stopped working, is investigating

Jordan:
 - make trac entries for any PRS issues that require LinuxBIOS changes.
   * Made Wiki page
   * Still need to file some trac tickets for things that need to be
     changed
 - open a trac bug about the text rendering speed problem.
   * No progress
 - DCON:  Preserve picture when graphics engine, gamma support
   * Sent gamma patch to Mark
   * done, is in kernel tree


Do we have another build for widespread deployment?
 - build122 is good, many new features
 - but old buildrom
 - and touchpad isn't working
 - plan on pushing build124 tomorrow and testing that, marking stable if
   good

BIOS stuff:
 - we now have our own linuxbios repository, which buildrom git pulls from
   * cafe_nand is working now in Linux-as-Bootloader.
 - Chris Ball to get new LB image to Mitch for 256M testing
 - Mitch tested OFW on DCON, seems to work but his DCON has problems in
   both OFW and Linux.  Sounds like bad display.

Testing:
 - run Mitch's bits on Cambridge's DCON?
 - The LB on Mitch's DCON doesn't seem to give boot messages
   * need to investigate
   * possibility of the console going to the serial only, not the screen.
 - We need a lot more DCON Testing with no serial port connected:
   * Mitch, Cambridge and Jordan
 - mail LB image to software-team plus Quanta for configuration testing.

Driver integration:
 - battery driver is commited to olpc-2.6 and fedora kernel trees.
 - SD needs to be turned on in Fedora builds.
 - camera is included.
 - Marcelo working on libertas.

Report from Jim on Marvell wireless:
 Talked with Marvell on the phone, There was a change for the boot
 code by Marvell, to solve a "warm reboot" problem.  That change was
 broken, when libertas tried to touch the device it didn't
 acknowledge.  Not clear for existing B-Test boards with that boot
 code whether we can reprogram them; Marvell has patches to do this,
 Marcelo is trying to do this.  If works, the low-risk thing to do is
 to use same boot code as we used on A-Test, instead of this new
 (broken) code.  It's not clear that we can talk to the chip well
 enough on any B-Test board to reprogram it; we would need to use a
 soldering iron, unless we are lucky and the reprogramming technique
 works.  Chicken and egg problem.

 The "USB reset" wireless problem that we're seeing is not a problem,
 it is a workaround for the warm reboot problem.  Is easier to upgrade
 to newer boot code when we have the "B" boot code.  The reset
 workaround doesn't work on B-Test; the device is just dead.

 Jordan had Marvell send him a USB dongle, but it had the old firmware,
 so it appeared to be working but B-Test still wasn't.

Priorities:
 - give BIOS to Ray at Quanta.
 - user-level:
   * Barry Vercoe's stuff is packaged?
   * csound is packaged but untested, etoys and squeak are packaged
     * AI:  Blizzard to test csound with Barry.
   * AI:  find out dependencies on mplayer, use it to demo camera
   * AI:  if too many dependencies, build and put in our own repo.
   * need an editor: non-GNOME version of AbiWord?
     * AI: Blizzard to talk to AbiWord guys.

BIOS can slip to 26th/27th Cambridge time without affecting SMT build.
If we can release on the 25th, Quanta can test it and possibly release
new EC code; if we can't, we'll have to test ourselves.

Quanta needs the NAND filesystem on November 1st Taiwan, Halloween
US-time.

Marvell may be sending new FPGA bits, Quanta hasn't heard anything.

Jordan to get Geode team in touch with Jim.

-- 
Chris Ball   <cjb at laptop.org>