[Community-news] OLPC News 2007-09-01

[Walter Bender]

1. Schedules/testing: This week was "feature freeze" and by the end of
the week we were very close to finalizing the feature set for the
Trial-3 software release. Since there were a lot of new features
getting checked in, we saw numerous builds and wrote up and fixed
many, many blocking and regression bugs. Over the next week we will be
focused on stability, through bug fixing and testing. There was some
good progress on some of the biggest (and loudest) bugs related to
suspend and resume, which is great to see. No more features for
Trial-3. If you think there is an exception to this rule, please
contact Jim Gettys and Kim Quirk.

2. Sugar: The Collabora team continued to work on final items before
the upcoming software release.  This included adding support for
mutable activity properties (name, tags, colors, etc), invitation
support, porting of many of the activities over to the new tubes
specification and cleaning up a lot of the base system elements.
Morgan Collett updated the Connect and Chat activities to the new
interfaces.

If no school server is chosen, a presence server in the MIT
collocation center is being used to enable individual developers to
share.

A new version of TamTam from Jean Piché's team is included in the new builds.

Pango (to enable support for languages like Amharic) and Cairo were
updated to their latest versions.

Chris Ball updated Pippy to add journal integration, sound support from
Nathanaël Lécaudé and the TamTam team, and new examples from Madeleine
Ball, Mel Chua, and Rafael Ortiz.

Simon Schamijer updated the Memorize activity for the new tubes API.

Simon also has been working on a the web browser (See the new visual
design at http://wiki.laptop.org/go/Web_Browser). He added a sharable
link tray and fixed bugs in the shared browser session.

Marc Maurer worked on the Abiword (Write Activity) collaboration
backend. He also added a new format toolbar to Write and added some
additional style options (headings, numbering, bullets, etc.)  He also
added the ability to insert images into Write directly from the
Journal.

3. Upgrades and multi-boot: Scott Ananian installed Debian Linux
alongside Fedora Sugar on an XO as a demonstration of the new upgrade
mechanism, which allows you to keep the old version around and boot
into if the upgrade goes awry; and the P_SF_RUN Bitfrost security
mechanism, which allows a child to poke around the root filesystem and
muck with things, while still being able to revert to the "pristine"
OS image if things go wrong (See preliminary instructions at
http://wiki.laptop.org/go/Installing_Debian_as_an_upgrade). This will
get much easier next week as the rest of the upgrade infrastructure is
rolled out: you will be able to just subscribe to the "debian" stream
to get Debian installed (for example). There will be "stable" and
"devel" streams for Sugar releases—the real point of this work.

4. X Window System: Bernardo Innocenti merged more xkb changes for our
existing keymaps, defined a few missing keysyms and updated the olpc
patch in
response to Sergey Udaltsov 's reviews. He is currently testing new
RPMs. A new patch submission is due soon.

Bernie and Walter Bender have also been finalizing the keyboard
layouts for mass production (By way of example, see
http://wiki.laptop.org/go/Image:Keyboard_english.png).

5. Embedded controller: Richard worked on the
"battery-always-charging" bug. It does not appear to be just an EC
problem as Open Firmware shows the correct status but the kernel does
not. They both use the same EC commands. It is still unclear what is
wrong.

6. The Wireless Bug (Trac #1835): Richard is also working on this bug.
 The more he gets into it, the more it smells like hardware. On a
resume, after some number of cycles, the machine will hang. It is
likely that we are not getting good data from the reset-vector fetch.
The next step is to decode the LPC (low pin-count) bus and see what
data is coming back in the fail case.

Javier Cardona and Chris Ball also continued work on the
wireless-resume bug. Javier added debugging code that is providing
useful information on each crash, and is continuing to try to iron out
the bugs as we find them.

Richard and Chris found a bug where the CPU occasionally on resume
appears to be stalling instead of executing instructions. Since this
happens during our extreme-traffic wireless-resume testing, the
problem could be as simple as us not giving the power rails enough
time to quiesce. Richard is working on it.

7. School server: The school server software continues to improve; a
new release with several bug fixes and the laptop registration service
is expected by the beginning of next  week. To aid in server
development in crowded work areas (such as Cambridge), John Watlington
tested and documented the mesh blinding tables (See
http://wiki.laptop.org/go/Mesh_Debug). Several new school servers came
online this week: one for a trial in India, one for content
development, and several for testing in Cambridge.

8. Kernel: This week Andres Salomon worked on vserver patches and
fixing sound bugs. The screeching-upon-resume bug has been fixed,
along with a number of other sound bugs. The patches been pushed
upstream.  A few bugs still remain, but no known major ones.

Marcelo Tosatti added code to our kernels that makes it possible to
trace how long suspend/resume cycles take. This should allow us to
pinpoint where we spend a lot of our time. Currently, the largest
consumer of time turns out to be writing to the serial console during
resume.

9. Firmware: Mitch Bradley Released Q2C26 firmware with OS security
and activation support, some bug fixes, improved NAND FLASH bad-block
management, and created a kit for creating signed OS images and
leases. Lilian Walter put into place the code to solicit stateless
DHCPv6
information. She is finishing up IPv6 fragmentation and reassembly.

10. Wireless: Michail borrowed an Anritsu Network analyzer and did
some antenna performance measurements on the C- and B- Build laptops.
He is happy to report that the C Build units are the first ones with
completely functioning antennas. Quanta has implemented grounding of
the antenna cable's shield in the C-Build machine and that seems to
have made the right antenna perform properly. Michail's only comment
is that the insulation on the left antenna's wire is being stripped
too short: the braid deforms from the mechanical stress imposed on it
during the antenna's rotation. The insulation has to be left intact
all the way to the grounding sticker as it the case with the right
antenna.

Note: the C-Build laptops have metalization all around the plastic parts.
Because of that, when the antennas are closed, the plastic below them
acts as a ground plane and diminishes the antenna's performance
completely compared to the previous builds. The antennas work much
better when in the up position on the C-Build laptops.

11. Build 542.3: John (J5) Palmieri reports that a new spin of the
stable build fixes a bug in Sugar that had prevented the installation
of translation files; it also fixes a Journal bug for the Spanish
locale.

http://olpc.download.redhat.com/olpc/streams/development/build542.3/

When installing this image, it should be named 'os5423.img' (along
with the corresponding 'os5423.crc'). The FAT file system would
otherwise be unhappy with the double dots in the names and the
auto-reinstallation script would thus refuse to install them.

-walter
-- 
Walter Bender
One Laptop per Child
http://laptop.org