System Software Meeting minutes, 2007-01-09.

[Chris Ball]

System Software Meeting minutes, 2007-01-09:

Absent:  Mitch, Marcelo, Andres, Quanta folks, Dave, Jon Corbet.

Current action items:

* AI: cjb to relnote python performance
* AI: cjb to investigate delaying rw / mount, kill -STOP jffs2_gcd_mtd0
* AI: #601: cjb to check whether we have pango extents fix in
* AI: #668: jg to rewrite this bug to talk about rotation instead
* AI: jg and blizzard to talk with Lennart about sound server ideas
* AI: Jordan and ajax to investigate xrandr segfault
* AI: J5 to fix camera app
* AI: dcbw and marcopg to investigate mesh/avahi not working.
* AI: Richard to coordinate getting B2 boards out to the right people
      when they arrive in Cambridge next week.

Previous action items:

* AI: Mitch: send USB analyzer to Richard and get a board modified
> USB analyzer hasn't arrived in Cambridge, need to ask Mitch.

* AI: Richard to send a symptomatic battery to Quanta.
> Done, received at Quanta and problem duplicated there.

* AI: Jim: Work with Walter on touchpad decisions
> Touchpad changes probably won't make B2.

* AI: Jordan, Jim, et al, decide what to do about gamma for B3.
> There's more realistic gamma support in Xrandr 1.2, which we'll move to.

* AI: Mitch: upload the PCI spoofing patch
> Unknown.

* AI: Jordan: find PM code for VSA
> Done, patch sent out.  We can suspend now, and the vcore CPU rail does
  turn off.  We aren't jumping to the right resume address yet.  Progress!

Discussion:

Performance:  the jffs2_gcd_mtd0 thread hurts boot and post-boot
performance; we can't get rid of it completely, but we can stop 
it by sending SIGSTOP and restarting it once the boot is finished.
We can also look at delaying the remount read-write of the jffs2 /,
which has the same effect since the thread isn't started until the 
filesystem goes read-write; cjb will see how many of the writes
during boot we can get inside a tmpfs to delay going read-write.

Adam Jackson cut a large chunk out of our X load time out with an 
xkbcomp optimisation.

jg and blizzard will be at LCA next week, but contactable online.
jg has a useful Google spreadsheet with a testing checklist for release
QA:   http://spreadsheets.google.com/ccc?key=pN4reQKyQ_DQbGTkcbqYddw

B-2 boards give us many more places to measure current on the power
rails, we should see the first B-2 boards late this week or early next 
week.  We need to make sure people who need B2 boards get them, but jg 
and blizzard will be away next week.  Richard to coordinate with OLPC
admins.

Bug triage actions:

#42:  Needs testing, perhaps from dilinger.
#230: What's the state of Urdu in pango/etc?
#318, #340:  dcbw is looking at these at the moment.
#397: Need hardware before we can test this.
#658: dcbw/marco will work on this tomorrow.
#224: We should drop a note to dwmw2 to let him know we're happy with this.
#362: We think cupsd is gone, but should check.
#601: CJB to check whether we have pango extents fix.

-- 
Chris Ball   <cjb at laptop.org>