software call agenda notes 02/06/2007

[Christopher Blizzard]

Software Call 02/06/2007

Attending:

o Jordan
o Chris Ball
o Chris Blizzard
o Marcelo
o Zephania
o Mitch
o John
o Andres (late)
o Richard (late)

Agenda:

o Process/policy by which things get into the kernel?
- Probably needs Jim around
- Need to pull in the experimental branch from time to time
- Need more communication about what needs to go in and when

o Build timeline updates
- Mitch wasn't there for the B2 build, so not sure
- Still have outstanding touchpad issue?  Not sure how it's to be resolved
   at this time?  (time-based fencing of non-existing ops?)
   - Two issues - erratic finger sensing?
   - Do they jump when they forget to send the Up packet?
   - Zeph - thinks that it's been fixed, but there's no specific list?
   - Mitch thinks that the mylar pad helps with the electical issues and 
there's a separate issue with the EC where it misses some packets
   - Status of the workaround in the kernel?  Warp thinks it's in the 
olpc-2.6 kernel tree (cjb confirms)
   - More issues with communications
   - More problems with propagating information
   - commits-kernel not working anymore
   - more discussion of stable vs. not stable
   - very confusing about how something makes it into the world
   - andres: kernels - being automatically pulled into the nightlies is 
fine?
   - b2 is already done, basically
   - the touchpad jumping issue isn't fixed
   - need to tell J5 that a new kernel is available? don't need that.
   - What else goes into the build?
     - OS components that change
     - both stable and unstable?
     - master == stable, experimental == unstable
     - andres, file trac bug for url for j5 to pull into the tree from 
master
   - any risk here?  need to inform devel at laptop.org that they commit to 
the right branch

o BIOS issues
   - Don't have the newest BIOS in an auto-updated image?
   - Bless 239 for auto-updating image?
   - Maybe 239 + new kernel + new BIOS?
   - New battery stuff + new kernel
   - Need to do some testing with the firmware?  Nothing _that_ important
   - New EC bits fix some problems but not a lot of testing outside of 
b2s in china
   - Do we have a list of burning issues for b2 right now?  Not really.
   - Overcharging battery problem?  Touchpad issue?  (Jumpy fix is in 
the OS, not in the EC anyway)
   - Going to include -61 in the b2 build in quanta - need to make an 
auto update image that includes the new BIOS
   - OS 239 + BIOS 61

o EC Code + Plan
- David has a copy of the code
- Mitch is still clean
- Need a decision here

o Power Management
- Mitch: VSAles code fails to do power off - it's solved now (missing GPIO)
- No feature that is missing from the fast boot VSAless firmware
- In a position to do resume from RAM
- Jordan and Mitch working on end to end PCI + power - committed to 
experimental
- What are we missing now?
   - Mitch thinks that we're ready to do it now, need a VSAless kernel
   - These guys are ready to go?  Sounds like it.
- Marcelo: one more issue: want drivers to shut down devices when possible
   - Each driver needs a timeout right now i.e. Marvell won't shut 
itself down
   - Need to discuss and define
   - dynticks?  Jordan hasn't been, really.  Patches are out there and 
basically ready to go.  Kind of waiting on Linus to avoid a later merge. 
  Andrew morton says that he thinks they are ready to go.
   - Power rails + management + measurement in place?
     - Richard has been working on it in between things.
     - Resistors need to be swapped out, trying to find the right parts
     - Hard to find the low ohm devices, making some progress
     - Probably need a solid day to get this all done (mili-ohm.)
     - Lots of discussion, etc.
     - Target is the end of the week

o Next steps for the UI
- Journal is the most important

o X depth and performance issues
- Needs more discussion
- 4444 or 5551, 24 bit depth is not the right to go in Jordan's opinion
   - no improvements in screen to screen copy
   - 24 bit accel engine can't convert while it's blending
   - animations might look awesome, but won't help other problems
   - can't animate things right now, maybe 3-4 a second
   - 4444
   - Jordan will make the driver work if he can get pictures in the 
right format

o Next steps for performance
- -fPIC is a really large slowdown in shared libraries (40% faster for 
python, as an example)
- measurement infrastructure
- power rail measurement into the tinderbox, but nothing new there

o Remaining wireless issues
- firmware driver for the wireless from lillian
- new firmware release next week some time
   - copy of the firmware in OFW - could export to the kernel?
   - would be good to unify?
   - boot up is fast, won't init wireless by default
   - 100k for the wireless firmware and compresses well
- richard will try to reproduce with router in #841
   - needs to be openwrt - RC5
   - and has to be the right firmware
   - tracy should have already bought one
   - richard needs to make sure it's the right version