Custom Search
|
Date: September 30, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<046.c52eb28b6661d54dd3cad18a42cecdb5@xxxxxxxxxx>
References:
<83-trac@xxxxxxxxxxxxxx>
#83: Systemtap in the OLPC kernel.
----------------------------+-----------------------------------------------
Reporter: cjb | Owner: cjb
Type: enhancement | Status: assigned
Priority: normal | Milestone: BTest-1
Component: infrastructure | Resolution:
Keywords: |
----------------------------+-----------------------------------------------
Comment (by cjb):
Replying to [comment:7 cjb]:
> I'm confused -- CONFIG_KPROBES was added, but CONFIG_DEBUG_INFO doesn't
appear anywhere in config-olpc-generic. How come?
Ah, I see: it's in config-generic, since it was already turned on in
Fedora.
--
Ticket URL: <http://dev.laptop.org/ticket/83#comment:8>
One Laptop Per Child <http://laptop.org/>
Date: September 30, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<046.c52eb28b6661d54dd3cad18a42cecdb5@xxxxxxxxxx>
References:
<83-trac@xxxxxxxxxxxxxx>
#83: Systemtap in the OLPC kernel.
----------------------------+-----------------------------------------------
Reporter: cjb | Owner: cjb
Type: enhancement | Status: assigned
Priority: normal | Milestone: BTest-1
Component: infrastructure | Resolution:
Keywords: |
----------------------------+-----------------------------------------------
Comment (by cjb):
Hi Marcelo,
Replying to [comment:6 marcelo]:
> No problem, since vmlinux will grow to a size just below <30MiB (where
the debugging information is retained), while the actual running kernel is
stripped down to current size + CONFIG_KPROBES.
>
> Adding these options to the OLPC Fedora RPM now...
I'm confused -- CONFIG_KPROBES was added, but CONFIG_DEBUG_INFO doesn't
appear anywhere in config-olpc-generic. How come?
--
Ticket URL: <http://dev.laptop.org/ticket/83#comment:7>
One Laptop Per Child <http://laptop.org/>
Date: September 30, 2006
From: Yoshiki Ohshima <yoshiki@xxxxxxxxxxxxxx>
Hello, There is a bug item on the bug tracking system to which I promised to provide more information. However, I couldn't do it this week, and I'm going to be away from the test board next week. http://dev.laptop.org/ticket/68 Can somebody with a board try the code Marcelo wrote on the page with and without a USB network adaptor? Or, even better, can somebody do following (or something like that) in Squeak, err, in the OLPC eToys environment with and without network? before := (FileStream readOnlyFileNamed: '/proc/interrupts') next: 10000. time := 200 timesRepeat: [Display forceToScreen: (0@0 corner: 400@400)]. after := (FileStream readOnlyFileNamed: '/proc/interrupts') next: 10000. The above code basically reads /proc/interrupts twice, before and after big memcpy like operations. Thank you so much! -- Yoshiki
Date: September 29, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<037.c7b1e9babeb1b1e7b5ff08ae6395e2fc@xxxxxxxxxx>
References:
<65-trac@xxxxxxxxxxxxxx>
#65: Temporary filesystems should be in RAM
-------------------------------+--------------------------------------------
Reporter: wmb@xxxxxxxxxxxxx | Owner: marcelo
Type: enhancement | Status: new
Priority: high | Milestone: BTest-1
Component: kernel | Resolution:
Keywords: |
-------------------------------+--------------------------------------------
Comment (by bluefoxicy):
Permissions should be different. /var/run should be 0755 and OLPC seems
to have a /var/lock that's 775 and owned by root:lock.
It seems /var/lock and /var/run need some sort of initialization when
starting, otherwise files and directories won't exist. Particularly
mounting over /var/lock via fstab breaks iwconfig; mounting over /var/run
breaks dbus. Even with mode=01777.
none /tmp tmpfs mode=01777 0 0
none /var/run tmpfs mode=0755 0 0
none /var/lock tmpfs mode=01777 0 0
none /var/tmp tmpfs mode=01777 0 0
--
Ticket URL: <http://dev.laptop.org/ticket/65#comment:7>
One Laptop Per Child <http://laptop.org/>
Date: September 29, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<032.74bc4b6fbf9cb650f47397ef952f5394@xxxxxxxxxx>
References:
<123-trac@xxxxxxxxxxxxxx>
#123: FC build images need md5sums
--------------------+-------------------------------------------------------
Reporter: rsmith | Owner: J5
Type: task | Status: new
Priority: normal | Milestone: ATest
Component: distro | Resolution:
Keywords: |
--------------------+-------------------------------------------------------
Changes (by blizzard):
* owner: blizzard => J5
--
Ticket URL: <http://dev.laptop.org/ticket/123#comment:1>
One Laptop Per Child <http://laptop.org/>
Date: September 29, 2006
From: Ian Stirling <ian.stirling@xxxxxxxxxxxxxx>
In-reply-to:
<451D7D1A.1070002@xxxxxxxxxxxxxx>
References:
<451BFCC2.3070701@xxxxxxx> <1159471397.20227.181.camel@xxxxxxxxxxxxxxxxxxxxx> <451D7D1A.1070002@xxxxxxxxxxxxxx>
Ian Stirling wrote:
Jim Gettys wrote:We've been very concerned about the "green" of our laptop, as we know that it may not be recycled properly. Obviously, a smaller/lighter/more rugged/longer lived machine has a lot of advantages as well in this area<snip>The very low power consumption of the machines means way less global warming due to electricity usage.Beware, I'm bored, and have the Internet!Say average child flatulence is 1l/day - sources I found said 200ml-2l, I'm assuming 1l, as diets high in vegetables are more likely to cause an increase. Assuming that methane is 7%, this is about 150ml/day. (I found several sources saying 7-9%, though some dismissed it as 'trace'. No primary literature.)Assuming the '20 times more potent than CO2' is accurate - equivalent to 3l/day of CO2. If we're talking 10Wh/day, how much of the food is used to produce electricity? Well, if we're talking 1000Calories per day baseline food, that's 4MJ/day, or around 1% before taking into account human efficiencies. As I see 25% as a rough 'maximum food conversion efficiency, that means somewhere on the order of 4% more food, or 120g of CO2 equivalent per day.(I'm assuming that the food is being produced in a carbon neutral manner.)
Err. Spot the comedy three orders of magnitude error. Methane is slightly less dense than water.
Date: September 29, 2006
From: Ian Stirling <ian.stirling@xxxxxxxxxxxxxx>
In-reply-to:
<1159471397.20227.181.camel@xxxxxxxxxxxxxxxxxxxxx>
References:
<451BFCC2.3070701@xxxxxxx> <1159471397.20227.181.camel@xxxxxxxxxxxxxxxxxxxxx>
Jim Gettys wrote:
We've been very concerned about the "green" of our laptop, as we know that it may not be recycled properly. Obviously, a smaller/lighter/more rugged/longer lived machine has a lot of advantages as well in this area
<snip>
The very low power consumption of the machines means way less global warming due to electricity usage.
Beware, I'm bored, and have the Internet!Say average child flatulence is 1l/day - sources I found said 200ml-2l, I'm assuming 1l, as diets high in vegetables are more likely to cause an increase. Assuming that methane is 7%, this is about 150ml/day. (I found several sources saying 7-9%, though some dismissed it as 'trace'. No primary literature.)
Assuming the '20 times more potent than CO2' is accurate - equivalent to 3l/day of CO2. If we're talking 10Wh/day, how much of the food is used to produce electricity? Well, if we're talking 1000Calories per day baseline food, that's 4MJ/day, or around 1% before taking into account human efficiencies. As I see 25% as a rough 'maximum food conversion efficiency, that means somewhere on the order of 4% more food, or 120g of CO2 equivalent per day.
(I'm assuming that the food is being produced in a carbon neutral manner.) Wow!If you use this 120g of CO2 budget, and burn it in a modern gas power station, you get about 300Wh. Using a coal power station, 100Wh. Using an figure for a poorly maintained diesel generator (and neglecting local emissions) it's still >90Wh.
I'd never have guessed.
Date: September 29, 2006
From: "Jaya Kumar" <jayakumar.lkml@xxxxxxxxx>
In-reply-to:
<BD412EFD-7E30-4F8E-B2FD-54BEAEF0F6D9@xxxxxxxxx>
References:
<451BFCC2.3070701@xxxxxxx> <BD412EFD-7E30-4F8E-B2FD-54BEAEF0F6D9@xxxxxxxxx>
On 9/28/06, Kevin Purcell <kevinpurcell@xxxxxxxxx> wrote:
A plan for recovering broken laptops might be useful but considering the deployment by the governments involved it seems responsibility probably should flow downhill though developing countries governments are not very good in this area.
You are sooo right. Let's put the responsibility for recovering broken laptops and handling associated environmental issues in the hands of our governments. As you know, my governent, and many other governments in developing countries, are greatly renowned for their efficiency, transparency and successful public services. ;-) Seriously, If OLPC wants to do unit recovery/refurbishment/environmentally safe disposal, then establishing a return program such that damaged or failed units could be returned for compensation or for new units ought to be considered. Doing so through a recovery service company (preferably one in the private sector rather than government owned/run) is what I would encourage OLPC to explore. Best regards, jayakumar
Date: September 29, 2006
From: "Jaya Kumar" <jayakumar.lkml@xxxxxxxxx>
In-reply-to:
<451BFCC2.3070701@xxxxxxx>
References:
<451BFCC2.3070701@xxxxxxx>
On 9/28/06, Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@xxxxxxx> wrote:
Speaking of media coverage, I have noticed that OLPC is hardly ever mentioned in German mainstream media. I don't know how well we're doing in other countries, but I fear it's not much better.
Subjectively speaking, OLPC garners quite a lot of attention in the Indian press. Objectively speaking, Google http://www.google.com/trends?q=OLPC&ctab=0&geo=all&date=all, suggests that the peoples most interested in OLPC are as follows: 1. Taiwan 2. Argentina 3. Brazil 4. India 5. Canada 6. United States 7. Australia 8. China 9. Italy 10. Spain Looking at languages, Chinese and Portugese are the top 2 used in searching for OLPC. jayakumar
Date: September 29, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<034.8a8e43004082f980fca0d34d3f777627@xxxxxxxxxx>
References:
<135-trac@xxxxxxxxxxxxxx>
#135: Web browser does not start up
--------------------+-------------------------------------------------------
Reporter: ianb | Owner: dcbw
Type: defect | Status: new
Priority: normal | Milestone:
Component: sugar | Resolution:
Keywords: |
--------------------+-------------------------------------------------------
Old description:
> When I try to start the web browser, I get this exception on the console:
>
> {{{
> sugar-presence-service (12847): Level 10 - Service Ian
> [04cd1d2deb45f4ca36c869455ec28eb63fd83d1d]._GroupChatActivity_Sugar_redhat_com._udp
> in domain local on 3.0 disappeared.
> sugar-presence-service (12847): Level 40 - Error resolving service Ian
> [04cd1d2deb45f4ca36c869455ec28eb63fd83d1d]._GroupChatActivity_Sugar_redhat_com._udp:
> org.freedesktop.Avahi.TimeoutError
> Fd 4 did not have the close-on-exec flag set!
> Fd 5 did not have the close-on-exec flag set!
> Traceback (most recent call last):
> File "/home/ianb/src/ext/sugar-
> jhbuild/build/share/sugar/shell/view/frame/BottomPanel.py", line 54, in
> __activity_clicked_cb
> self._shell.start_activity(icon.get_bundle_id())
> File "/home/ianb/src/ext/sugar-
> jhbuild/build/share/sugar/shell/view/Shell.py", line 127, in
> start_activity
> activity = ActivityFactory.create(activity_type)
> File "/home/ianb/src/ext/sugar-jhbuild/build/lib/python2.4/site-
> packages/sugar/activity/ActivityFactory.py", line 56, in create
> xid = factory.create()
> File "/usr/lib/python2.4/site-packages/dbus/proxies.py", line 79, in
> __call__
> reply_message = self._connection.send_with_reply_and_block(message,
> timeout)
> File "dbus_bindings.pyx", line 458, in
> dbus_bindings.Connection.send_with_reply_and_block
> dbus_bindings.DBusException: Message did not receive a reply (timeout by
> message bus)
> }}}
New description:
When I try to start the web browser, I get this exception on the console:
{{{
sugar-presence-service (12847): Level 10 - Service Ian
[04cd1d2deb45f4ca36c869455ec28eb63fd83d1d]._GroupChatActivity_Sugar_redhat_com._udp
in domain local on 3.0 disappeared.
sugar-presence-service (12847): Level 40 - Error resolving service Ian
[04cd1d2deb45f4ca36c869455ec28eb63fd83d1d]._GroupChatActivity_Sugar_redhat_com._udp:
org.freedesktop.Avahi.TimeoutError
Fd 4 did not have the close-on-exec flag set!
Fd 5 did not have the close-on-exec flag set!
Traceback (most recent call last):
File "/home/ianb/src/ext/sugar-
jhbuild/build/share/sugar/shell/view/frame/BottomPanel.py", line 54, in
__activity_clicked_cb
self._shell.start_activity(icon.get_bundle_id())
File "/home/ianb/src/ext/sugar-
jhbuild/build/share/sugar/shell/view/Shell.py", line 127, in
start_activity
activity = ActivityFactory.create(activity_type)
File "/home/ianb/src/ext/sugar-jhbuild/build/lib/python2.4/site-
packages/sugar/activity/ActivityFactory.py", line 56, in create
xid = factory.create()
File "/usr/lib/python2.4/site-packages/dbus/proxies.py", line 79, in
__call__
reply_message = self._connection.send_with_reply_and_block(message,
timeout)
File "dbus_bindings.pyx", line 458, in
dbus_bindings.Connection.send_with_reply_and_block
dbus_bindings.DBusException: Message did not receive a reply (timeout by
message bus)
}}}
And then the system doesn't start.
Comment (by ianb):
Well, I did:
{{{
./sugar-jhbuild update
./sugar-jhbuild build --clean --autogen
}}}
And that went fine, until I run:
{{{
$ ./sugar-jhbuild run
Running the installed sugar...
Traceback (most recent call last):
File "/home/ianb/src/ext/sugar-jhbuild/build/bin/sugar", line 22, in ?
from sugar.session.Emulator import Emulator
File "/home/ianb/src/ext/sugar-jhbuild/build/lib/python2.4/site-
packages/sugar/session/Emulator.py", line 5, in ?
from sugar.session.Process import Process
File "/home/ianb/src/ext/sugar-jhbuild/build/lib/python2.4/site-
packages/sugar/session/Process.py", line 3, in ?
import gobject
File "/home/ianb/src/ext/sugar-jhbuild/build/lib/python2.4/site-
packages/gtk-2.0/gobject/__init__.py", line 30, in ?
from _gobject import *
ImportError: /home/ianb/src/ext/sugar-jhbuild/build/lib/python2.4/site-
packages/gtk-2.0/gobject/_gobject.so: undefined symbol:
PyUnicodeUCS2_FromUnicode
}}}
--
Ticket URL: <http://dev.laptop.org/ticket/135#comment:2>
One Laptop Per Child <http://laptop.org/>
Date: September 28, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<034.8a8e43004082f980fca0d34d3f777627@xxxxxxxxxx>
References:
<135-trac@xxxxxxxxxxxxxx>
#135: Web browser does not start up
--------------------+-------------------------------------------------------
Reporter: ianb | Owner: dcbw
Type: defect | Status: new
Priority: normal | Milestone:
Component: sugar | Resolution:
Keywords: |
--------------------+-------------------------------------------------------
Comment (by marco):
I think this is the gtkmozembed segfaulting. Can you try to do a full,
clean sugar-jhbuild and see if that helps. The browser build changed a lot
recently and you might have leftovers of the old builds.
--
Ticket URL: <http://dev.laptop.org/ticket/135#comment:1>
One Laptop Per Child <http://laptop.org/>
Date: September 28, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<046.d15f1878264cfda698dab8a001ece246@xxxxxxxxxx>
References:
<130-trac@xxxxxxxxxxxxxx>
#130: Sugar browser crash when opening popups
--------------------+-------------------------------------------------------
Reporter: marco | Owner: marco
Type: defect | Status: closed
Priority: normal | Milestone: BTest-1
Component: sugar | Resolution: fixed
Keywords: |
--------------------+-------------------------------------------------------
Changes (by marco):
* status: assigned => closed
* resolution: => fixed
--
Ticket URL: <http://dev.laptop.org/ticket/130#comment:3>
One Laptop Per Child <http://laptop.org/>
Date: September 28, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<046.d943a02d94d3dcafd74a6013465e344c@xxxxxxxxxx>
References:
<87-trac@xxxxxxxxxxxxxx>
#87: Firefox freezes when domain not found
--------------------+-------------------------------------------------------
Reporter: ianb | Owner: marco
Type: defect | Status: closed
Priority: normal | Milestone: BTest-1
Component: sugar | Resolution: fixed
Keywords: |
--------------------+-------------------------------------------------------
Changes (by marco):
* status: assigned => closed
* resolution: => fixed
--
Ticket URL: <http://dev.laptop.org/ticket/87#comment:2>
One Laptop Per Child <http://laptop.org/>
Date: September 28, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<046.46848ad8ea1cca2c1ac22fd3ac116071@xxxxxxxxxx>
References:
<65-trac@xxxxxxxxxxxxxx>
#65: Temporary filesystems should be in RAM
-------------------------------+--------------------------------------------
Reporter: wmb@xxxxxxxxxxxxx | Owner: marcelo
Type: enhancement | Status: new
Priority: high | Milestone: BTest-1
Component: kernel | Resolution:
Keywords: |
-------------------------------+--------------------------------------------
Changes (by marcelo):
* owner: blizzard => marcelo
--
Ticket URL: <http://dev.laptop.org/ticket/65#comment:6>
One Laptop Per Child <http://laptop.org/>
Date: September 28, 2006
From: "Chris Kavanagh" <librestreamkav@xxxxxxxxxxx>
Hello,I'm working on an entirely different platform both hardware and software. However, it seems that OLPC and myself are both using Marvell based WiFi radios. I have been battling with a problem for sometime now and haven't received a whole lot of support from Marvell. I noticed on a Libertas group that the OLPC driver was recieving a status code of 4 on associations - I am also seeing this with my system.
I've been leaning towards the idea that this might be more then just an incorrect status code but rather serious issue in the firmware. In fact if I dump the entire contents of my association response buffer recieved from the firmware via the AP, I actually see data from my association request buffer that was just previously downloaded to the card. As if one buffer is being copied over the other in the firmware.
It took me a while to see this because the buffer recieved from the firmware almost looks good. For example:
The first 4 ushorts seem to look fine (correct response code, reasonable size, sequence number and command result), so the driver begins processing the rest of the data (the association data). The Cap Info looks reasonable, the associationID is fine (same as what's "sniffed" in the air) but the status code is 0x0004. It wasn't until I started looking at the rest of the buffer that I noticed my WEP key in place text was now part of my association response IE elements. Obviously something that wasn't transmitted by the AP :)
This concerns me because on the edge of a network with pour signal strength this seems to happen often as per retry's and occationally the firmware stops responding to commands altogether.
Has anyone had any luck in tracking down this "StatusCode == 4" issue? In advance, thank you very much for you help, Chris _________________________________________________________________Buy what you want when you want it on Sympatico / MSN Shopping http://shopping.sympatico.msn.ca/content/shp/?ctId=2,ptnrid=176,ptnrdata=081805
Date: September 28, 2006
From: Jim Gettys <jg@xxxxxxxxxx>
In-reply-to:
<451BFCC2.3070701@xxxxxxx>
References:
<451BFCC2.3070701@xxxxxxx>
We've been very concerned about the "green" of our laptop, as we know
that it may not be recycled properly. Obviously, a smaller/lighter/more
rugged/longer lived machine has a lot of advantages as well in this area
than a conventional machine.
It is lead free. The batteries are NiMH, which do not require recycling
as they don't have nasty chemicals in them (Lithium ions should be
recycled, if they don't burst into flames first ;-)). We've been
careful on materials selection.
The very low power consumption of the machines means way less global
warming due to electricity usage.
I expect we stack up extremely well on this front. And we've been
worrying about it from the beginning.
Regards,
- Jim
On Thu, 2006-09-28 at 18:48 +0200, Carl-Daniel Hailfinger wrote:
> Hi,
>
> Greenpeace has been making news lately with their campaign for
> "greener" laptops.
> http://www.greenpeace.org/international/campaigns/toxics/electronics/how-the-companies-line-up
>
> We probably do not want to be slapped by them with a negative
> label, so it might be sensible to publish more on our strategy
> for environmentally responsible manufacturing and deployment.
>
> http://wiki.laptop.org/go/Our_technology#Considering_the_millions_to_be_built.2C_will_it_be_recyclable.3F
> is unfortunately very vague. Greenpeace would probably complain
> that there is no published plan for recycling the laptops, but
> since our interest is permanent deployment and we hope none of
> the laptops fail, recycling and take-back should never be needed.
>
> If we manage to get a positive statement by Greenpeace, it would
> probably help getting more media coverage.
>
> Speaking of media coverage, I have noticed that OLPC is hardly
> ever mentioned in German mainstream media. I don't know how well
> we're doing in other countries, but I fear it's not much better.
>
>
> Regards,
> Carl-Daniel
> _______________________________________________
> Devel mailing list
> Devel@xxxxxxxxxx
> http://mailman.laptop.org/mailman/listinfo/devel
--
Jim Gettys
One Laptop Per Child
Date: September 28, 2006
From: Kevin Purcell <kevinpurcell@xxxxxxxxx>
In-reply-to:
<451BFCC2.3070701@xxxxxxx>
References:
<451BFCC2.3070701@xxxxxxx>
One should not that GP does this for it's own publicity.Their attacks on Apple (why Apple? because they're hot right now) have been debunked by several out there. Engaging them will not generate any useful publicity for OLPC.
<http://www.roughlydrafted.com/RD/Home/E83D58B3-10E0-4A9C-8847- BCE665EE235C.html> <http://www.roughlydrafted.com/RD/Home/29C5599A- FCD8-4E30-9AD5-5497999ABA1B.html> <http://www.roughlydrafted.com/RD/Home/ABC6DFDA-9DE9-4EA8- A269-65EAAB628676.html>
The OLPC (I presume) will conform to RoHS (I hope). It might be worthwhile subjecting final prototypes to EPEAT testing for independent confirmation and to head of GP style ranting.
<http://www.epeat.net/>A plan for recovering broken laptops might be useful but considering the deployment by the governments involved it seems responsibility probably should flow downhill though developing countries governments are not very good in this area.
Of course, GP will just use that stick to beat you with. That seems to be their only purpose or screwing up their own PR
"In the twenty years since the Chernobyl tragedy, the world's worst nuclear accident, there have been nearly [FILL IN ALARMIST AND ARMAGEDDONIST FACTOID HERE],"
<http://www.washingtonpost.com/wp-dyn/content/article/2006/06/01/ AR2006060101884_pf.html>
(BTW, I was a GP member in the 1980s. I left when I found out how they worked.).
On Sep 28, 2006, at 9:48 AM, Carl-Daniel Hailfinger wrote:
Hi, Greenpeace has been making news lately with their campaign for "greener" laptops.http://www.greenpeace.org/international/campaigns/toxics/ electronics/how-the-companies-line-upWe probably do not want to be slapped by them with a negative label, so it might be sensible to publish more on our strategy for environmentally responsible manufacturing and deployment.http://wiki.laptop.org/go/ Our_technology#Considering_the_millions_to_be_built. 2C_will_it_be_recyclable.3Fis unfortunately very vague. Greenpeace would probably complain that there is no published plan for recycling the laptops, but since our interest is permanent deployment and we hope none of the laptops fail, recycling and take-back should never be needed. If we manage to get a positive statement by Greenpeace, it would probably help getting more media coverage. Speaking of media coverage, I have noticed that OLPC is hardly ever mentioned in German mainstream media. I don't know how well we're doing in other countries, but I fear it's not much better. Regards, Carl-Daniel _______________________________________________ Devel mailing list Devel@xxxxxxxxxx http://mailman.laptop.org/mailman/listinfo/devel
-- Kevin Purcell kevinpurcell@xxxxxxxxx
Date: September 28, 2006
From: "Jordan Crouse" <jordan.crouse@xxxxxxx>
In-reply-to:
<57947bf80609280950l6c7cb03cx7659f0ee37109d37@xxxxxxxxxxxxxx>
References:
<1159448995.20227.106.camel@xxxxxxxxxxxxxxxxxxxxx> <57947bf80609280950l6c7cb03cx7659f0ee37109d37@xxxxxxxxxxxxxx>
On 28/09/06 12:50 -0400, Tom Sylla wrote: > Should this tree, built with the tools called out in the instructions, > create a binary matching the one currently cat'ed with LB? Has that been > checked? i.e., has this tree been verified to be the one used by SteveG to > make the last VSA binary in use? Yes - brave Martin Roth verified that this code matches the VSA binary that we are using. Jordan -- Jordan Crouse Senior Linux Engineer Advanced Micro Devices, Inc. <www.amd.com/embeddedprocessors>
Date: September 28, 2006
From: "Tom Sylla" <tsylla@xxxxxxxxx>
In-reply-to:
<1159448995.20227.106.camel@xxxxxxxxxxxxxxxxxxxxx>
References:
<1159448995.20227.106.camel@xxxxxxxxxxxxxxxxxxxxx>
1) the VSA source code is now in Git: thanks to Jordan for bird-dogging
this. http://dev.laptop.org/git?p=geode-vsa;a=summary Many of you know
this, but for those that don't: the VSA code can only be built currently
using very obsolete and no longer commercially available versions of
various Windows compilers and other tools. But it does mean you can see
how it works, and we know that if we really had to, we could fix
problems.
2) A test framework and some specific implementations of high use
routines (e.g. strcmp, memcopy, and the like), implemented by John
Zulauf. These need to be rerun on our boards to see if John's initial
results are correct, and where appropriate, integrated into glibc or
other appropriate places; I believe John did most of this work on an LX,
but the GX and LX are similar in most areas. I feel the test framework
is as valuable as the specific routines that John worked on optimizing.
Current work includes:
* Memcmp performance test -- improved 1.47x
* Memcpy performance test -- improved 1.24x
* Memset performance test -- improved 1.1x
* Strcmp performance test -- improved 2.32x
* strcpy performance test -- improved 1.65x
* Strlen performance test -- improved 1.2x
Note that I checked the whole kit and caboodle into git, including
doxygen generated html files; it would be good if someone got this
properly set up on a web server on laptop.org and took it as a personal
crusade.
- Jim
--
Jim Gettys
One Laptop Per Child
_______________________________________________
Devel mailing list
Devel@xxxxxxxxxx
http://mailman.laptop.org/mailman/listinfo/devel
_______________________________________________ Devel mailing list Devel@xxxxxxxxxx http://mailman.laptop.org/mailman/listinfo/devel
Date: September 28, 2006
From: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@xxxxxxx>
Hi, Greenpeace has been making news lately with their campaign for "greener" laptops. http://www.greenpeace.org/international/campaigns/toxics/electronics/how-the-companies-line-up We probably do not want to be slapped by them with a negative label, so it might be sensible to publish more on our strategy for environmentally responsible manufacturing and deployment. http://wiki.laptop.org/go/Our_technology#Considering_the_millions_to_be_built.2C_will_it_be_recyclable.3F is unfortunately very vague. Greenpeace would probably complain that there is no published plan for recycling the laptops, but since our interest is permanent deployment and we hope none of the laptops fail, recycling and take-back should never be needed. If we manage to get a positive statement by Greenpeace, it would probably help getting more media coverage. Speaking of media coverage, I have noticed that OLPC is hardly ever mentioned in German mainstream media. I don't know how well we're doing in other countries, but I fear it's not much better. Regards, Carl-Daniel
Date: September 28, 2006
From: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@xxxxxxx>
In-reply-to:
<1159448995.20227.106.camel@xxxxxxxxxxxxxxxxxxxxx>
References:
<1159448995.20227.106.camel@xxxxxxxxxxxxxxxxxxxxx>
Jim Gettys wrote: > 1) the VSA source code is now in Git: thanks to Jordan for bird-dogging > this. http://dev.laptop.org/git?p=geode-vsa;a=summary Many of you know > this, but for those that don't: the VSA code can only be built currently > using very obsolete and no longer commercially available versions of > various Windows compilers and other tools. But it does mean you can see > how it works, and we know that if we really had to, we could fix > problems. Quoting from the README.txt in the VSA code: "It is under the GNU LGPL." That's really great! Thanks Jordan! On a related note, this means ALL code running on the host CPU is completely open source. (Probably worth a news item in the community news.) The wireless firmware is now the only closed source code in the ROM (and the next OLPC revision will hopefully have the wireless firmware built around an opensource kernel). Regards, Carl-Daniel
Date: September 28, 2006
From: Jim Gettys <jg@xxxxxxxxxx>
1) the VSA source code is now in Git: thanks to Jordan for bird-dogging this. http://dev.laptop.org/git?p=geode-vsa;a=summary Many of you know this, but for those that don't: the VSA code can only be built currently using very obsolete and no longer commercially available versions of various Windows compilers and other tools. But it does mean you can see how it works, and we know that if we really had to, we could fix problems. 2) A test framework and some specific implementations of high use routines (e.g. strcmp, memcopy, and the like), implemented by John Zulauf. These need to be rerun on our boards to see if John's initial results are correct, and where appropriate, integrated into glibc or other appropriate places; I believe John did most of this work on an LX, but the GX and LX are similar in most areas. I feel the test framework is as valuable as the specific routines that John worked on optimizing. Current work includes: * Memcmp performance test -- improved 1.47x * Memcpy performance test -- improved 1.24x * Memset performance test -- improved 1.1x * Strcmp performance test -- improved 2.32x * strcpy performance test -- improved 1.65x * Strlen performance test -- improved 1.2x Note that I checked the whole kit and caboodle into git, including doxygen generated html files; it would be good if someone got this properly set up on a web server on laptop.org and took it as a personal crusade. - Jim -- Jim Gettys One Laptop Per Child
Date: September 28, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<038.38594231cc67b11bfc740e5cc244a7e9@xxxxxxxxxx>
References:
<120-trac@xxxxxxxxxxxxxx>
#120: buildrom fails on fc5
-----------------------+----------------------------------------------------
Reporter: dilinger | Owner: rminnich
Type: defect | Status: new
Priority: normal | Milestone:
Component: linuxbios | Resolution:
Keywords: |
-----------------------+----------------------------------------------------
Comment (by rsmith):
Replying to [comment:4 rminnich]:
> It's not really lzma that is the issue, this is still some kind of weird
toolchain issue. I can build on fc5 just fine with lzma.
> I need a login on a failing machine.
Ron is correct. I spend a lot of time looking at this tonight. Lzma is
just the straw that broke the camels back.
The linuxbios size rom size is set in the config file at 32k. A lzma
build on my Debian Testing machine ends up with about 0x99 bytes free in
that area.
When you build with the FC5 toolchain the resulting image is just a wee
bit larger this extra size causes a section overlap and breaks the build.
When I increase the linuxbios rom image size from 32k to 33k things build
fine for my on learn.laptop.org where previously they would break.
Heres and objdump between the stock build on my Debian Testing machine and
sucessful build on learn.
Richard's Debian box:
{{{
./fallback/linuxbios: file format elf32-i386
Sections:
Idx Name Size VMA LMA File off Algn
0 .ram 00007106 ffff8000 ffff8000 00001000 2**0
CONTENTS, ALLOC, LOAD, DATA
1 .rom 00000e3a fffff106 fffff106 00008106 2**4
CONTENTS, ALLOC, LOAD, READONLY, CODE
2 .reset 00000010 fffffff0 fffffff0 00008ff0 2**0
CONTENTS, ALLOC, LOAD, READONLY, DATA
3 .id 00000017 ffffffd9 ffffffd9 00008fd9 2**0
CONTENTS, ALLOC, LOAD, READONLY, DATA
}}}
Learn.laptop.org with a 33k linuxbios size.
{{{
./fallback/linuxbios: file format elf32-i386
Sections:
Idx Name Size VMA LMA File off Algn
0 .ram 0000719f ffff7c00 ffff7c00 00000c00 2**0
CONTENTS, ALLOC, LOAD, DATA
1 .rom 00000e41 ffffed9f ffffed9f 00007d9f 2**4
CONTENTS, ALLOC, LOAD, READONLY, CODE
2 .reset 00000010 fffffff0 fffffff0 00008ff0 2**0
CONTENTS, ALLOC, LOAD, READONLY, DATA
3 .id 00000017 ffffffd9 ffffffd9 00008fd9 2**0
CONTENTS, ALLOC, LOAD, READONLY, DATA
}}}
So you can see that the .ram and .rom size on learn is 0xa0 bytes bigger
than my Debian image. This was blowing the 0x99 free bytes and
overlapping segments.
With the 33k linuxbios size I did not receive any of the wierd objcopy
problems.
--
Ticket URL: <http://dev.laptop.org/ticket/120#comment:5>
One Laptop Per Child <http://laptop.org/>
Date: September 27, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<038.7ae5d9f6f8dbcf29d186715dcc29e535@xxxxxxxxxx>
References:
<124-trac@xxxxxxxxxxxxxx>
#124: --hash-style for decreased load times
-------------------------+--------------------------------------------------
Reporter: bluefoxicy | Owner: blizzard
Type: enhancement | Status: closed
Priority: normal | Milestone: gamma
Component: distro | Resolution: fixed
Keywords: |
-------------------------+--------------------------------------------------
Changes (by cjb):
* status: new => closed
* resolution: => fixed
Comment:
Verified present in our build86:
bash-3.1# readelf -a /bin/echo | grep GNU_HASH
[ 3] .gnu.hash GNU_HASH 08048168 000168 00003c 04 A 4
0 4
0x6ffffef5 (GNU_HASH) 0x8048168
bash-3.1#
Closing.
--
Ticket URL: <http://dev.laptop.org/ticket/124#comment:4>
One Laptop Per Child <http://laptop.org/>
Date: September 27, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<038.7ae5d9f6f8dbcf29d186715dcc29e535@xxxxxxxxxx>
References:
<124-trac@xxxxxxxxxxxxxx>
#124: --hash-style for decreased load times
-------------------------+--------------------------------------------------
Reporter: bluefoxicy | Owner: blizzard
Type: enhancement | Status: new
Priority: normal | Milestone: gamma
Component: distro | Resolution:
Keywords: |
-------------------------+--------------------------------------------------
Comment (by cjb):
> Performance tools are your friends; we'll measure, and then optimize;
but don't overlook this technique.
Understood, Jim, thanks. I've looked at this now, and am told that the
binutils and glibc patches have recently gone into FC6, all the FC6 binary
RPMs have been rebuilt with --hash-style enabled, and we don't have to do
anything to enable it when building our own binaries under FC6. Will
leave this bug open until I've verified that it's present in our build,
but it looks like we don't need to do anything to get the benefit here.
--
Ticket URL: <http://dev.laptop.org/ticket/124#comment:3>
One Laptop Per Child <http://laptop.org/>
Date: September 27, 2006
From: Ivan Krstić <krstic@xxxxxxxxxxxxxxxxxxxxxxxxx>
In-reply-to:
<055.2d4216141b9609eb969244dbd63571af@xxxxxxxxxx>
References:
<120-trac@xxxxxxxxxxxxxx> <055.2d4216141b9609eb969244dbd63571af@xxxxxxxxxx>
Ron Minnich wrote: > I need a login on a failing machine. Please send me your SSH2 key. -- Ivan Krstić <krstic@xxxxxxxxxxxxxxxxxxxxxxxxx> | GPG: 0x147C722D
Date: September 27, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<046.6f504da1bed47bee0b14cd550d5f8a24@xxxxxxxxxx>
References:
<120-trac@xxxxxxxxxxxxxx>
#120: buildrom fails on fc5
-----------------------+----------------------------------------------------
Reporter: dilinger | Owner: rminnich
Type: defect | Status: new
Priority: normal | Milestone:
Component: linuxbios | Resolution:
Keywords: |
-----------------------+----------------------------------------------------
Comment (by rminnich):
It's not really lzma that is the issue, this is still some kind of weird
toolchain issue. I can build on fc5 just fine with lzma. I need a login on
a failing machine.
ron
--
Ticket URL: <http://dev.laptop.org/ticket/120#comment:4>
One Laptop Per Child <http://laptop.org/>
Date: September 27, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<038.38594231cc67b11bfc740e5cc244a7e9@xxxxxxxxxx>
References:
<120-trac@xxxxxxxxxxxxxx>
#120: buildrom fails on fc5
-----------------------+----------------------------------------------------
Reporter: dilinger | Owner: rminnich
Type: defect | Status: new
Priority: normal | Milestone:
Component: linuxbios | Resolution:
Keywords: |
-----------------------+----------------------------------------------------
Comment (by cjb):
Building with LZMA disabled doesn't give this error.
I tried with and without LZMA on learn.laptop.org; with-LZMA failed as in
the bug, without-LZMA completed successfully.
--
Ticket URL: <http://dev.laptop.org/ticket/120#comment:3>
One Laptop Per Child <http://laptop.org/>
Date: September 27, 2006
From: "Zarro Boogs per Child" <bugtracker@xxxxxxxxxx>
In-reply-to:
<038.7ae5d9f6f8dbcf29d186715dcc29e535@xxxxxxxxxx>
References:
<124-trac@xxxxxxxxxxxxxx>
#124: --hash-style for decreased load times
-------------------------+--------------------------------------------------
Reporter: bluefoxicy | Owner: blizzard
Type: enhancement | Status: new
Priority: normal | Milestone: gamma
Component: distro | Resolution:
Keywords: |
-------------------------+--------------------------------------------------
Changes (by jg):
* priority: low => normal
Comment:
Replying to [comment:1 cjb]:
> Since this technique only works for ELF binaries, and most of the code
we run during use of the laptop (I assume) will be native Python code,
this isn't much more than a boot-time optimisation for us. Nice to have,
but we have bigger performance issues to deal with first.
Hmmm...
Chris; think a bit further out: once we know where the cycles really go,
any serious CPU burner will need to be rewritten into something other than
an interpreted language, and building our key applications and libraries
well is important.
One of the failures of the Newton was that it burned its batteries, by
only using an interpreter.
Performance tools are your friends; we'll measure, and then optimize; but
don't overlook this technique.
--
Ticket URL: <http://dev.laptop.org/ticket/124#comment:2>
One Laptop Per Child <http://laptop.org/>
Date: September 27, 2006
From: Jim Gettys <jg@xxxxxxxxxx>
In-reply-to:
<1159385274.4966.19.camel@xxxxxxxxxxxxxxxxxxxxx>
References:
<uslidnoub.wl%yoshiki@xxxxxxxxxxxxxx> <1159384139.5997.284.camel@xxxxxxxxxxxxxxxxxxxxx> <1159385274.4966.19.camel@xxxxxxxxxxxxxxxxxxxxx>
Andres, everyone!
We do not encourage people to build their own LinuxBIOS, unless they are
involved in LinuxBIOS development.
Given that we have a problem with which toolchains can correctly build
the BIOS image, and the failure results so bad (unless the person has a
PLCC flash part), only *tested* bios images we "bless" should be
installed.
>From last night's status call, I believe the next build (87) should have
versions of LinuxBIOS which should set the PS/2 port properly.
Yoshiki, please wait until the next build is available.
Regards,
- Jim
On Wed, 2006-09-27 at 15:27 -0400, Andres Salomon wrote:
> On Wed, 2006-09-27 at 15:08 -0400, Jim Gettys wrote:
> > Andres has managed to get it working here; exactly what incantations are
> > required, I'm not sure.
> >
> > Zephaniah Hull (who has been working on this), is currently quite sick,
> > so progress on that end on hold pending him feeling better.
> >
> > Andres, what did you end up having to do?
> >
> > You certainly need a Bios upgrade: the symptoms you describe are those
> > we had here until the right registers were set for the PS/2 port.
> > Regards,
> > - Jim
> >
>
> First, you have to build a new linuxbios image using buildrom. That's
> described here:
> http://wiki.laptop.org/go/Building_LinuxBIOS
> You definitely want to use revision 2432 or newer (that's what I used).
> Edit Config.mk, setting LINUXBIOS_VER to that. LinuxBios doesn't
> currently build under Fedora, but it builds just fine under Ubuntu
> (dapper).
>
> If you'd like, I can just send you the rom image that I used.
>
> Once you've flashed the rom image onto your board or PLCC chip
> (http://wiki.laptop.org/go/Flashing_LinuxBIOS), the build86 Fedora image
> should work fairly well. Make sure the various kernel modules are
> loaded (psmouse, evdev), and you should be okay. By default, xorg.conf
> will use /dev/input/mice. This should work okay w/ the touchpad, but
> isn't really desired in the long run (the evdev driver is what we want).
> However, that requires upgrading to xorg 7.2, and doesn't seem to work
> all that well yet (at least not for me).
>
>
>
--
Jim Gettys
One Laptop Per Child
Date: September 27, 2006
From: "Richard Smith" <smithbone@xxxxxxxxx>
In-reply-to:
<yd38xk5nkxx.fsf@xxxxxxxxxxxxxxxxxxxxxx>
References:
<uslidnoub.wl%yoshiki@xxxxxxxxxxxxxx> <1159384139.5997.284.camel@xxxxxxxxxxxxxxxxxxxxx> <ur6xxnl6j.wl%yoshiki@xxxxxxxxxxxxxx> <yd38xk5nkxx.fsf@xxxxxxxxxxxxxxxxxxxxxx>
1. * wait for build87, which will hopefully be released today, and use
the ( (ram!=infineon) ? linuxbios.rom : linuxbios.rom.cas25 )
2. * take a git checkout of buildrom and build from that manually 3. * ask Andres for a copy of his recent linuxbios.rom Regardless of what path you chose you first should consult http://wiki.laptop.org/go/Upgrading_to_LinuxBIOS#Checking_your_DRAM_type If you have Infineon DRAM options 2 & 3 need extra care. Option 1 should be covered in the instructions. Let me know if you have any feedback on the instructions. -- Richard A. Smith
Date: September 27, 2006
From: Chris Ball <cjb@xxxxxxxxxxxxxx>
In-reply-to:
<ur6xxnl6j.wl%yoshiki@xxxxxxxxxxxxxx> (Yoshiki Ohshima's message of "Wed, 27 Sep 2006 12:50:28 -0700")
References:
<uslidnoub.wl%yoshiki@xxxxxxxxxxxxxx> <1159384139.5997.284.camel@xxxxxxxxxxxxxxxxxxxxx> <ur6xxnl6j.wl%yoshiki@xxxxxxxxxxxxxx>
Yoshiki,
> Thank you. I wasn' aware that there are different versions of
> LinuxBIOS. I've upgraded to the one (linuxbios.rom) that comes
> with the build86 devel disk image, but the symptoms are the same.
> I'll wait for more info...
The change you require was made after build86. To summarise the options
in the replies so far, you can:
* wait for build87, which will hopefully be released today, and use
the linuxbios.rom in that
* take a git checkout of buildrom and build from that manually
* ask Andres for a copy of his recent linuxbios.rom
> (The disk image I use is *not*
> olpc-redhat-stream-development-ext3.img but
> olpc-redhat-stream-development-build-86-20060922_1506-devel_ext3.img.
> Does that matter?)
These are the same image, just renamed.
- Chris.
--
Chris Ball <cjb@xxxxxxxxxxxxxx> <http://blog.printf.net/>
Date: September 27, 2006
From: Yoshiki Ohshima <yoshiki@xxxxxxxxxxxxxx>
In-reply-to:
<8a0c36780609271235y752356dfyfab6e0819b4e793a@xxxxxxxxxxxxxx>
References:
<uslidnoub.wl%yoshiki@xxxxxxxxxxxxxx> <1159384139.5997.284.camel@xxxxxxxxxxxxxxxxxxxxx> <8a0c36780609271235y752356dfyfab6e0819b4e793a@xxxxxxxxxxxxxx>
Andres, Richard and Jordan, I sent response without reading these messages. Sorry. > Can you hold off for a bit? build 87 should have the necessary bios > fixes and it would be good to have someone else test the new upgrade > instructions. > > Build is scheduled for today sometime. If it doesn't show up then we > can discuss building your own or someone can build you a rom file. Sure. It makes more sense to everybody that the code and files I have are in sync. Thank you! -- Yoshiki
Date: September 27, 2006
From: Yoshiki Ohshima <yoshiki@xxxxxxxxxxxxxx>
In-reply-to:
<1159384139.5997.284.camel@xxxxxxxxxxxxxxxxxxxxx>
References:
<uslidnoub.wl%yoshiki@xxxxxxxxxxxxxx> <1159384139.5997.284.camel@xxxxxxxxxxxxxxxxxxxxx>
Jim, > You certainly need a Bios upgrade: the symptoms you describe are those > we had here until the right registers were set for the PS/2 port. Thank you. I wasn' aware that there are different versions of LinuxBIOS. I've upgraded to the one (linuxbios.rom) that comes with the build86 devel disk image, but the symptoms are the same. I'll wait for more info... (The disk image I use is *not* olpc-redhat-stream-development-ext3.img but olpc-redhat-stream-development-build-86-20060922_1506-devel_ext3.img. Does that matter?) Thank you again, -- Yoshiki
Date: September 27, 2006
From: "Jordan Crouse" <jordan.crouse@xxxxxxx>
In-reply-to:
<1159385274.4966.19.camel@xxxxxxxxxxxxxxxxxxxxx>
References:
<uslidnoub.wl%yoshiki@xxxxxxxxxxxxxx> <1159384139.5997.284.camel@xxxxxxxxxxxxxxxxxxxxx> <1159385274.4966.19.camel@xxxxxxxxxxxxxxxxxxxxx>
On 27/09/06 15:27 -0400, Andres Salomon wrote: > On Wed, 2006-09-27 at 15:08 -0400, Jim Gettys wrote: > > Andres has managed to get it working here; exactly what incantations are > > required, I'm not sure. > > > > Zephaniah Hull (who has been working on this), is currently quite sick, > > so progress on that end on hold pending him feeling better. > > > > Andres, what did you end up having to do? > > > > You certainly need a Bios upgrade: the symptoms you describe are those > > we had here until the right registers were set for the PS/2 port. > > Regards, > > - Jim > > > > First, you have to build a new linuxbios image using buildrom. That's > described here: > http://wiki.laptop.org/go/Building_LinuxBIOS > You definitely want to use revision 2432 or newer (that's what I used). > Edit Config.mk, setting LINUXBIOS_VER to that. LinuxBios doesn't > currently build under Fedora, but it builds just fine under Ubuntu > (dapper). Buildrom HEAD is set for the correct revision now, as is the tagged version 20060926-1. Jordan -- Jordan Crouse Senior Linux Engineer Advanced Micro Devices, Inc. <www.amd.com/embeddedprocessors>
Date: September 27, 2006
From: "Richard Smith" <smithbone@xxxxxxxxx>
In-reply-to:
<1159384139.5997.284.camel@xxxxxxxxxxxxxxxxxxxxx>
References:
<uslidnoub.wl%yoshiki@xxxxxxxxxxxxxx> <1159384139.5997.284.camel@xxxxxxxxxxxxxxxxxxxxx>
You certainly need a Bios upgrade: the symptoms you describe are those
we had here until the right registers were set for the PS/2 port.
Regards,
- Jim
Yoshinki: Can you hold off for a bit? build 87 should have the necessary bios fixes and it would be good to have someone else test the new upgrade instructions. Build is scheduled for today sometime. If it doesn't show up then we can discuss building your own or someone can build you a rom file. -- Richard A. Smith
Date: September 27, 2006
From: Andres Salomon <dilinger@xxxxxxxxxx>
In-reply-to:
<1159384139.5997.284.camel@xxxxxxxxxxxxxxxxxxxxx>
References:
<uslidnoub.wl%yoshiki@xxxxxxxxxxxxxx> <1159384139.5997.284.camel@xxxxxxxxxxxxxxxxxxxxx>
On Wed, 2006-09-27 at 15:08 -0400, Jim Gettys wrote: > Andres has managed to get it working here; exactly what incantations are > required, I'm not sure. > > Zephaniah Hull (who has been working on this), is currently quite sick, > so progress on that end on hold pending him feeling better. > > Andres, what did you end up having to do? > > You certainly need a Bios upgrade: the symptoms you describe are those > we had here until the right registers were set for the PS/2 port. > Regards, > - Jim > First, you have to build a new linuxbios image using buildrom. That's described here: http://wiki.laptop.org/go/Building_LinuxBIOS You definitely want to use revision 2432 or newer (that's what I used). Edit Config.mk, setting LINUXBIOS_VER to that. LinuxBios doesn't currently build under Fedora, but it builds just fine under Ubuntu (dapper). If you'd like, I can just send you the rom image that I used. Once you've flashed the rom image onto your board or PLCC chip (http://wiki.laptop.org/go/Flashing_LinuxBIOS), the build86 Fedora image should work fairly well. Make sure the various kernel modules are loaded (psmouse, evdev), and you should be okay. By default, xorg.conf will use /dev/input/mice. This should work okay w/ the touchpad, but isn't really desired in the long run (the evdev driver is what we want). However, that requires upgrading to xorg 7.2, and doesn't seem to work all that well yet (at least not for me).
Date: September 27, 2006
From: Jim Gettys <jg@xxxxxxxxxx>
In-reply-to:
<uslidnoub.wl%yoshiki@xxxxxxxxxxxxxx>
References:
<uslidnoub.wl%yoshiki@xxxxxxxxxxxxxx>
Andres has managed to get it working here; exactly what incantations are
required, I'm not sure.
Zephaniah Hull (who has been working on this), is currently quite sick,
so progress on that end on hold pending him feeling better.
Andres, what did you end up having to do?
You certainly need a Bios upgrade: the symptoms you describe are those
we had here until the right registers were set for the PS/2 port.
Regards,
- Jim
On Wed, 2006-09-27 at 11:31 -0700, Yoshiki Ohshima wrote:
> Jim,
>
> Ian brought back the touchpad prototype from Cambridge, and he did
> soldering of the board as suggested. But it doesn't work so far.
>
> We are on LinuxBIOS and build86 image (and booting from a USB
> memory). We booted the system with the touchpad connected (it doesn't
> smoke, so I'd assume soldering is done properly). X and sugar
> interface start up, but when I touch the pad, the mouse cursor
> sometimes moves randomly but most of the time it doesn't move. In the
> console, when I touch the pad, I get messages like:
>
> psmouse.c: PenTablet at isa0060/serio4/input0 lost sync at byte1
> psmouse.c: issuing reconnect request
> psmouse.c: Failed to reset mouse on isa0060/serio1
> psmouse.c: Failed to enable mouse on isa0060/serio1
>
> In the dmesg, there are lines like:
>
> input: PS/2 Generic Mouse as /class/input/input7
> ...
> input: PS/2 Generic Mouse as /class/input/input8
>
> etc. the last number seems to be incremented.
>
> In console, I get the messages only when I touched the pad, so the
> system is getting some signals from it, but seems to be failing to
> receive some bytes and interpret them.
>
> Do you have any suggestions to try next?
>
> -- Yoshiki
--
Jim Gettys
One Laptop Per Child
Date: September 27, 2006
From: "Jordan Crouse" <jordan.crouse@xxxxxxx>
In-reply-to:
<57947bf80609271043n2c1ac556i32ba15c410ffc5ba@xxxxxxxxxxxxxx>
References:
<4519EF8E.7030609@xxxxxxxxxxxxx> <e8ac1af10609270525r3d9ec801qb9caef2dd4eb84ec@xxxxxxxxxxxxxx> <20060927172628.GJ997@xxxxxxxxxx> <57947bf80609271043n2c1ac556i32ba15c410ffc5ba@xxxxxxxxxxxxxx>
On 27/09/06 13:43 -0400, Tom Sylla wrote: > Actually, GX has many performance counters, in some cases, more detailed > than pentium or A64 counters. The main problems are that no one has added > the mappings and support for them to oprofile. Note that the GX counters are > unable to cause interrupts like those in A64 or Pentium, so you would have > to run oprofile in RTC or timer mode (which isn't so horrible). Unfortunately, our performance counters don't have the ability to interrupt on rollover, which is what oprofile uses as its trigger. Timer mode does work OK, its just that your profiling isn't as detailed, and you miss out on those situations where interrupts are disabled. It should be enough for nominal userland monitoring though. Jordan -- Jordan Crouse Senior Linux Engineer Advanced Micro Devices, Inc. <www.amd.com/embeddedprocessors>
Date: September 27, 2006
From: Yoshiki Ohshima <yoshiki@xxxxxxxxxxxxxx>
Jim, Ian brought back the touchpad prototype from Cambridge, and he did soldering of the board as suggested. But it doesn't work so far. We are on LinuxBIOS and build86 image (and booting from a USB memory). We booted the system with the touchpad connected (it doesn't smoke, so I'd assume soldering is done properly). X and sugar interface start up, but when I touch the pad, the mouse cursor sometimes moves randomly but most of the time it doesn't move. In the console, when I touch the pad, I get messages like: psmouse.c: PenTablet at isa0060/serio4/input0 lost sync at byte1 psmouse.c: issuing reconnect request psmouse.c: Failed to reset mouse on isa0060/serio1 psmouse.c: Failed to enable mouse on isa0060/serio1 In the dmesg, there are lines like: input: PS/2 Generic Mouse as /class/input/input7 ... input: PS/2 Generic Mouse as /class/input/input8 etc. the last number seems to be incremented. In console, I get the messages only when I touched the pad, so the system is getting some signals from it, but seems to be failing to receive some bytes and interpret them. Do you have any suggestions to try next? -- Yoshiki
Date: September 27, 2006
From: "Tom Sylla" <tsylla@xxxxxxxxx>
In-reply-to:
<20060927172628.GJ997@xxxxxxxxxx>
References:
<4519EF8E.7030609@xxxxxxxxxxxxx> <e8ac1af10609270525r3d9ec801qb9caef2dd4eb84ec@xxxxxxxxxxxxxx> <20060927172628.GJ997@xxxxxxxxxx>
On Wed, Sep 27, 2006 at 05:55:30PM +0530, Tushar Adeshara wrote:
> > What do we need for performance analysis? Chris Ball: systemtap
> > needs no kernel support. Marcello added new config options.
> > We need strace, valgrind (no dependencies), sysprof (needs atk),
> > ltrace (no dependencies), memprof (needs bonobo, etc, so it loses
> > unless the dependencies can be reduced).
>
> oprofile is also a good system-wide profiler, it works out of box on
> FC4 and provides good results.
> http://oprofile.sourceforge.net
I don't think the geode has performance counting MSRs, which oprofile needs.
Dave
_______________________________________________
Devel mailing list
Devel@xxxxxxxxxx
http://mailman.laptop.org/mailman/listinfo/devel
_______________________________________________ Devel mailing list Devel@xxxxxxxxxx http://mailman.laptop.org/mailman/listinfo/devel
Date: September 27, 2006
From: Dave Jones <davej@xxxxxxxxxx>