Author Topic: Opie Won't Start Properly (Read 4277 times)

systemparadox · « **on:** January 20, 2005, 10:36:31 am »

OK,
I have just reflashed for the third time in the past 4 days because of this.
I can't work out what triggers it, but I have a sneaky suspicion that it's those expansion cards again.

If I decide to shut down, or if i have to do a hard reboot, opie sometimes cannot start properly on the next boot. I just see a 1 px high line at the bottom of the screen- the rest is blank. It seems to still be running- I can get in via wifi (and something flickers if I plug the card in), but it doesn't respond to any taps or key presses.
That's it. If this happens it seems to be time for a reflash. No amount of rebooting can fix it. I can still use the console fine by cancelling opie's startup, and this has proved very useful for backing up things on root. There have been times inbetween that I have rebooted (hard or soft) fine, so I can't work out what causes it to do this.

Has anyone else had this problem?
Any suggestions?

Thanks
Simon

samac · « **Reply #1 on:** January 20, 2005, 05:09:32 pm »

Please consider this answer with a pinch of salt.

I remember that happening to me, but I cannot remember exactly why it happened. I seem to remember it was something to do with running out of memory. When that happened package manager crashed in the process of installing a package, causing all sorts of grief.

But I could be very wrong.

I would be tempted to try elimination.

ie. try to start opie with "/etc/init.d/opie start"

then uninstall packages you don't use.

Then only install one package at a time, if the problem occurs after a package you may have found the cause, but then again you might not have.

All I know is that I have not had any problems since /dev/root has 3MB of free space.

systemparadox · « **Reply #2 on:** January 21, 2005, 11:17:36 am »

samac,
thanks for replying. I tried /etc/init.d/opie start, but same problem.

Quote

I seem to remember it was something to do with running out of memory. When that happened package manager crashed in the process of installing a package, causing all sorts of grief.

I am having a lot of trouble with the packagemanager crashing during an install, and it goes very slowly even if it doesn't crash. This was not always the cause of the problem mentioned here, but iirc it did cause it once.

Quote

All I know is that I have not had any problems since /dev/root has 3MB of free space.

Where do you usually install programs to? SD?
I install to SD, usually over network or another partition on the SD, and I have found that copying files between the two partitions is really slow, but if i copy the files to ram and then copy them to the other sd partition it's much faster (like the copying between two floppy disks using on cache issue). This doesn't really explain the slowness when installing over network, but it depends where the packagemanager unpacks the packages. If it's the root, then that might explain it due to running out of space, if it's the same place the package is going, then that would expain it due to it copying from sd to sd.
Does anyone where the packagemanager unpacks the packages?
I might try moving the whole opt partition to the sd. This should free up some space on root, and possibly make ipkg unpack there if it isn't already.

If all else fails i'm gonna rewrite ipkg in bash, it will be very limited, but it should do the job, just don't ask me to uninstall anything or check what's installed.

Simon

samac · « **Reply #3 on:** January 21, 2005, 03:37:16 pm »

I changed to aqpkg as it appears more stable.

Just remember to install aqpkg and check it is working before uninstalling packagemanager.

I install to SD, and rather than copying anything I just uninstalled the package that was taking up room, say something unimportant like the appearance package or city time, then re-installed it onto SD, quick as you like direct from the package feeds on the internet. If you do this you will not have to move /opt as the "package manager" that you choose will create the necessary symbolic links from /mnt/card/opt to /opt.

The only folder I have moved is /root to /mnt/card/root and then symlinked it to /root, and that was mainly to protect my settings and private files.

This may not cure your problem but it will make the system more useable once it is stable, my problems with OZ 3.5.2 are now down to getting individual programs to work, the system overall is pretty stable.

systemparadox · « **Reply #4 on:** January 27, 2005, 11:55:53 am »

ITS JUST HAPPENED AGAIN!!! FOR GOODNESS SAKE WHAT IS GOING ON HERE??!?!?!?? I AM FED UP WITH REFLASHING!!! ARRRGGGHHHH!!!!!!!

The question is whether it's a problem isolated to ipkg or not- ipkg crashing is the only reason i ever have to reboot so I really don't know.

It would be much less annoying if I could fix it without reflashing, has anyone got any suggestions on this? (I suppose the problem with that is that I can't really try any fixes until the next time it happens, which I try to avoid)

I think I shall try using aqpkg instead.

lardman · « **Reply #5 on:** January 28, 2005, 05:16:35 am »

Start syslogd, then look at the output of logread.

-or if that doesn't produce the results-

Edit the /etc/init.d/opie script so that it outputs to the screen, then try running it via ssh to see what's failing.

Si

systemparadox · « **Reply #6 on:** February 01, 2005, 10:39:59 am »

It just happened again, so...

running 'syslogd', then '/etc/init.d/opie start', then 'logread' gives only one new line:

Code: [Select]

Feb 1 15:31:56 collie user.warn qpe: qt_init() [1344] - successfully entered daemon mode
Running '/etc/init.d/opie start' from a wireless ssh connection gives:

Code: [Select]

root@collie:/etc# /etc/init.d/opie start
Starting Opie in 5 seconds... press key to interrupt.
You seem to already have a /root/Applications directory.
Assuming it is the Opie Applications directory. Exiting.
Starting Opie....
QDir::readDirEntries: Cannot read the directory: /tmp
ODevice() - found 'Hardware     : Sharp-Collie'
qt_init() - starting in daemon mode...

And in both cases, 'ps -A' shows nothing relevant to opie or qpe. The screen is still frozen black with a 1 px bar across the bottom, looking like part of the taskbar. For some reason I often get this 1 px bar, even when everything is running fine.

EDIT: Opie appears to still be running, even though it is not listed in ps -A, when I insert my wifi card the screen flickers a bit, looking like it might be trying to produce the insert notification popup (but semi hidden behind the blackness).

It is not restricted to the packagemanager crashing.
I have a strong suspiction it's something to do with the sd.

Thanks for the help
Simon

systemparadox · « **Reply #7 on:** February 01, 2005, 10:48:46 am »

I have tried starting qpe manually on terminal 3 (not 2), but that gives nothing different.

How do I edit /etc/init.d/opie to make it give more output? Or have i already got all the output I'm going to get?

Possibly a problem with graphics device?
How do I recreate the devices? (none of them appear to be missing, i just want to try remaking them)

Thanks
Simon

lardman · « **Reply #8 on:** February 01, 2005, 12:20:33 pm »

/etc/init.d/opie is a script. Use vi (or whatever you like) to edit it to remove some of the &> /dev/null or add some descriptive echo statements etc.

Probably best to make a copy though and edit that one.

I'd also remove all of the storage cards to begin with to try to narrow it down.

The one pixel line also happens on mine, something to do with the kernel, not a problem and low priority to fix (as it doesn't do any harm, except slightly reducing the available screen space).

Si

systemparadox · « **Reply #9 on:** February 01, 2005, 12:33:22 pm »

AAHHH!!! URGENT HELP REQUIRED!!!
I was messing around trying to get it to work without reflashing (stupidly, before backing up), and I tried unlinking the packages on /mnt/card. Unfortunately, it appears that some vital library had to be upgraded, and was consequently installed onto the sd, rather than root where the old version was (but isn't anymore). I kept getting libm.so.6 errors whenever i tried to run a command (like 'ls'!), so I rebooted and now I can't even log in! Before i get to the login prompt i see some more libm.so.6 errors, concerning /bin/sh.
The prompt is just:
"(none) login: "
and it won't let me log in as root, which had no password (iirc there are no other logginable users setup on the system).

There are still files on root which I need- does anyone know a way around this or have I basically lost those files already (i suppose without logging in or a shell there isn't really much that can be done)?

Unfortunately I need to reflash and get the Z operational again within the next couple of hours, regardless of whether I have managed to retrieve those files.

Thanks in advance
Simon

I suppose the moral of the story is "backup BEFORE messing!"

systemparadox · « **Reply #10 on:** February 01, 2005, 12:55:54 pm »

Is it possible to boot from a cf or sd card?

News:

Author Topic: Opie Won't Start Properly (Read 4277 times)