Details added in paragraph “Modifications”
Thanks for the super-detailed report and investigation @ric9k. I’ve created an internal bug report about this and tagged it as “tracked”.
Can I please double-check that you ruled out PatchManager as a factor? That’s what I understood from reading through, but there’s still a fragment of doubt in my mind that it would be helpful to clear up.
One thing to add:
There have been versions of MyBackup which have a bug in the RPM uninstall script where it literally does a rm -rf ~
.
While I can’t see how this would trigger through a user switch, have any of you that were affected had MyBackup installed?
Unfortunately, I don’t remember much more than what I wrote.
But when it happened, I didn’t touch anything in PM before.
Perhaps interresting for you:
I still have the disk image of the system just before nuke happens.
I can restore it and be able to reproduce.
If you indicate me a way to log everything happening into the phone, I could try and report.
Indeed, I have MyBackup installed. (but finally never used)
Also here, I could delete the concerned file (the one containing rm -f
) and try to reproduce.
Or, to avoid changing too much things, I could temporarly rename the rm command, prior to try reproducing…
For general debugging, I guess the Jolla team must answer that.
For the patchmanager preload library there is the enviromnent variable PM_PRELOAD_DEBUG
which makes it verbose (but will log to stdout/stderr which would need to be captured).
For the daemon there is /var/lib/environment/patchmanager/10-dbus.conf
where you can set the debug environment to true and restart the service.
Enabling persistent journal would be the first step for sure.
Re: MyBackup you can’t edit the script because it’s in the rpm file.
And busybox rm
is a symlink to busybox itself, or a builtin, so no moving it out of the way either.
But if it is uninstalled and triggers the rm
there should be some output in /var/log/zypp/history
, I expect lots of permission denied messages.
But actually the MyBackup bug should nuke the whole system, not just home, so it’s probably unrelated.
You can however create this script and temporarily point the /usr/bin/rm
symlink to it:
#!/bin/sh
echo "rm called: " $0 $* > /dev/stderr
echo -n "rm parent: "$PPID ": " > /dev/stderr
tr "\0" " " < /proc/$PPID/cmdline > /dev/stderr
echo > /dev/stderr
exec /usr/bin/busybox rm "$@"
This issue happened to me yesterday, when switching between default user and guest (Access disk of device not starting properly). I did have patchmanager installed, so I haven’t been able to rule that cause out.
What I did find was that when I tried to start Jolla-settings from icon it failed to start. But when starting from terminal it successfully launched, though with warnings that “manager” couldn’t be created (can’t remember the exact wording) and that the database was locked. Will try to recreate the behavior when I have more time.
Did you start or stop Android Support just before changing user?
I never manually start or stop Android service. It may have been OOM killed, before I tried to switch to guest user, or implicitly started by me launching an Andoid app.
Unfortunately I can’t recall the exact details, as I was focusing on the user switching rather than Android support
I have also experiencing and user data deletion after switching to a secondary user.
I have Android support, but noting special do with it. Although Android data are not lost, in contrary to other sailfish user data.
I don’t tried to re produce it. But what I identified as having done is :
- from admin user (defaultuser) ask switching to a secondary user (not used for month) from interface
- rapidly after switched, ask to switch again to admin user
- when logged to admin user, the profiled was reset to new user, and “first startup setup” was launched.
This is very annoying, If I have a save, this one is over a month old.
Sorry it happened to you too.
Do you remember if Android Support was On while you did the described steps?
Yes, I supposed y than Android support was On.
I never stop it explicitly, and have application than start in background (like messaging app Element)
Same experience for me this morning. I need to reflah the smartphone !
I loose few datas but i loose some
And reflash is very easy, it’s after : you need one day to restore the smartphone as before the bug.
Jolla has the chance that i’m an afficionado
Same.
Happened again today, all data lost (secondary user is my main user) , after failing to start android in defaultuser.
No way to use recovery since the new (and appreciated) alphanumeric user as recovery is not accepting the password.
It happend to me today. When I checked my phone (XA2 plus, 4.4.0.72) guest user was activated. I don’t know how it happend. So I switched back to device owner/admin and all data was lost, \nemo was almost empty and only an android data folder (don’t remember the excact name) was there.
Is there any chance to restore the data? My last backup is some weeks old. I think the whisperfish database (used mybackup) changed in the meantime and the backup would not be functional for it.
Thanks for any help.
Hi, very sorry. This prob is a big annoyance.
Nemo? I guess coming from an old SF version.
Anyway. I don’t remember if data subsists. Seems once I switched off immediately and could find some data through recovery mode. Not sure though. Maybe worth it exploring the filesystem from there.
Strange that it happened all of a sudden. Scarry!
Generally happens after fiddling Android start/stop button and changing user.
About WF, sorry, I don’t know. Maybe ask on the WF thread.
Good luck.