Bug 11928 - r5497 does not see or connect to any wireless network
: r5497 does not see or connect to any wireless network
Status: CLOSED FIXED
Product: SB Controller
Classification: Unclassified
Component: SqueezeOS
: unspecified
: PC Windows XP
: P1 blocker (vote)
: 7.3.3
Assigned To: Matt Wise
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2009-04-29 15:29 UTC by James Richardson
Modified: 2009-09-08 09:19 UTC (History)
6 users (show)

See Also:
Category: ---


Attachments
serial capture (24.38 KB, text/plain)
2009-04-29 15:29 UTC, James Richardson
Details

Note You need to log in before you can comment on or make changes to this bug.
Description James Richardson 2009-04-29 15:29:32 UTC
Created attachment 5170 [details]
serial capture

Load 5497 to a device already on the network, notice that it fails to recoonect to the Player or Network after reboot.

Factory reset a jive with 5497 on it, walk through the setup, at network screen no APs are listed.

See attached log
Comment 1 James Richardson 2009-04-29 15:54:46 UTC
With 5497 on the controller, menu's fail as well, I see the following errors when access some menu's
---------------------
Loading: jive.JiveMain
ifconfig: eth0: error fetching interface information: Device not found
error in event function:
        /usr/share/jive/jive/utils/locale.lua:253: bad argument #2 to 'format' (
string expected, got nil)
stack traceback:
        [C 0x47878]: in function 'format'
        /usr/share/jive/jive/utils/locale.lua:253: in function </usr/share/jive/
jive/utils/locale.lua:248>
        (tail call): ?
        ...usr/share/jive/applets/AboutJive/AboutJiveApplet.lua:40: in function
'settingsShow'
        /usr/share/jive/applets/AboutJive/AboutJiveMeta.lua:20: in function </us
r/share/jive/applets/AboutJive/AboutJiveMeta.lua:20>
        (tail call): ?
        /usr/share/jive/jive/ui/SimpleMenu.lua:152: in function 'itemListener'
        /usr/share/jive/jive/ui/Menu.lua:122: in function </usr/share/jive/jive/
ui/Menu.lua:117>
        (tail call): ?
        (tail call): ?
        /usr/share/jive/jive/ui/Widget.lua:607: in function </usr/share/jive/jiv
e/ui/Widget.lua:583>
        [C 0x13984]: in function 'dispatchEvent'
        /usr/share/jive/jive/ui/Widget.lua:483: in function 'dispatchNewEvent'
        /usr/share/jive/jive/ui/Menu.lua:185: in function </usr/share/jive/jive/
ui/Menu.lua:133>
        (tail call): ?
        /usr/share/jive/jive/ui/Widget.lua:607: in function </usr/share/jive/jiv
e/ui/Widget.lua:583>
        [C 0x1f4fc]: in function '_eventHandler'
        /usr/share/jive/jive/ui/Window.lua:1295: in function </usr/share/jive/ji
ve/ui/Window.lua:1294>
        [C 0x13984]: ?
        [C 0x13dd8]: in function 'processEvents'
        /usr/share/jive/jive/ui/Framework.lua:247: in function </usr/share/jive/
jive/ui/Framework.lua:245>
Comment 2 James Richardson 2009-04-29 16:31:03 UTC
FYI: up to 5492 no {ifconfig: eth0: error fetching interface information: Device not found} error messages are observed and jive connects to network with no issues.
Comment 3 Felix Mueller 2009-04-29 16:57:41 UTC
The issue is that for whatever reason the wlan driver modules (gspi.ko and gspi8xxx.ko) got installed into the wrong directory.

They should be in: /lib/modules/2.6.22-P7/

but they are in: /lib/modules/

Because of that the wlan init script cannot find them and fails:

root: Starting wlan
insmod: cannot insert '/lib/modules/2.6.22-P7/gspi.ko': unknown symbol in module, or unknown parameter
insmod: cannot insert '/lib/modules/2.6.22-P7/gspi8xxx.ko': unknown symbol in module, or unknown parameter

I checked the recipe that installs them, but for one thing that hasn't changed recently and it also looks ok to me.

My current explanation is that in the recipe ${KERNEL_VERSION} for whatever reason is evaluated to an empty string which would lead to the wrong path.

Also I cannot reproduce this scenario on my local build. Maybe it only happens from a clean checkout?
Comment 4 Blackketter Dean 2009-04-29 17:01:39 UTC
Matt Wise: Can you look at this?
Comment 5 Matt Wise 2009-04-29 18:07:20 UTC
I'll take a look at this on the build servers, but even if this is a  
difference between a build server and Felix's home system -- Poky  
needs to build right on every system, so I would assume its a poky  
recipe problem.
Comment 6 Matt Wise 2009-04-29 19:01:05 UTC
One thing I noticed -- I think -- is that this build never actually succeeded on the production build system. It only succeeded on the beta build system, and was accidentally copied over. I'm re-running the build with a clean checkout on the production system to see if theres any difference. I still suspect Poky configuration issue though...
Comment 7 Matt Wise 2009-04-29 19:12:18 UTC
Although the jive_74 build hasnt finished yet, here's the output of a  
find from the production build host (this is the jive nightly build I  
believe)

./b24898co/a/u/t/o/7.4/trunk/squeezeos/poky/build/tmp/work/jive-poky- 
linux-gnueabi/marvell-gspi-module-bin-1.0-r2/install/marvell-gspi- 
module/lib/modules/2.6.22-P7/gspi.ko
./b24898co/a/u/t/o/7.4/trunk/squeezeos/poky/build/tmp/work/jive-poky- 
linux-gnueabi/marvell-gspi-module-bin-1.0-r2/image/lib/modules/2.6.22- 
P7/gspi.ko
./b24898co/a/u/t/o/7.4/trunk/squeezeos/poky/build/tmp/work/jive-poky- 
linux-gnueabi/marvell-gspi-module-bin-1.0-r2/gspi.ko


On the other hand here's the same find on the beta build host (jive_74  
build)

/opt/parabuild/etc/build/b0co/a/u/t/o/jive/7.4/trunk/squeezeos/poky/ 
build/tmp/work/jive-poky-linux-gnueabi/marvell-gspi-module-bin-1.0-r2/ 
image/lib/modules/gspi.ko
/opt/parabuild/etc/build/b0co/a/u/t/o/jive/7.4/trunk/squeezeos/poky/ 
build/tmp/work/jive-poky-linux-gnueabi/marvell-gspi-module-bin-1.0-r2/ 
gspi.ko
/opt/parabuild/etc/build/b0co/a/u/t/o/jive/7.4/trunk/squeezeos/poky/ 
build/tmp/work/jive-poky-linux-gnueabi/marvell-gspi-module-bin-1.0-r2/ 
install/marvell-gspi-module/lib/modules/gspi.ko


Sure implies that something is different about the two hosts, but I  
have no idea what yet. I think its very dangerous that we just keep  
building our code on Debian systems 'hoping' that they'll keep working  
without understanding the reasons why they might break if built on a  
RedHat (or other) distribution. When all the builds finish I'll start  
comparing log files...
Comment 8 Richard Titmuss 2009-04-30 01:45:51 UTC
(In reply to comment #7)
> I think its very dangerous that we just keep  
> building our code on Debian systems 'hoping' that they'll keep working  
> without understanding the reasons why they might break if built on a  
> RedHat (or other) distribution. When all the builds finish I'll start  
> comparing log files...

Matt, we don't hope they'll keep working. But like any bug we just need to find it and fix it. Cross-compiling is complex, and unfortunately it does depending to some extent on the host environment. Poky does a much better job at isolating this than the previous home grown build system, but it's still not perfect. We can debug multiple host OS if needed, but each time a new one is used it may introduce different build errors.
Comment 9 Matt Wise 2009-04-30 13:07:50 UTC
I believe this is fixed. Changes have been made to the build.sh parabuild script that force you to source a script that includes the SVN location for the private repos. Build 5362 is running right now with these changes and should be done in a few hours.
Comment 10 James Richardson 2009-05-04 07:23:44 UTC
r5577 appears to not have this issue
Comment 11 James Richardson 2009-06-17 09:37:27 UTC
This bug has been fixed in the 7.3.3 release version of SqueezeCenter!

If you haven't already. please download the new version from http://www.logitechsqueezebox.com/support/download-squeezecenter.html 

If you are still experiencing this problem, feel free to reopen the bug with your new comments and we'll have another look.