[apport] hwdb-kde crashed with IndexError in sendData()

[Ricardo]

Binary package hint: hwdb-client

fgg

ProblemType: Crash
Architecture: i386
Date: Fri Apr 13 14:55:43 2007
DistroRelease: Ubuntu 7.04
ExecutablePath: /usr/bin/hwdb-kde
InterpreterPath: /usr/bin/python2.5
Package: hwdb-client-kde 0.6.10
PackageArchitecture: all
ProcCmdline: /usr/bin/python /usr/bin/hwdb-kde
ProcCwd: /home/ricardo
ProcEnviron:
 LANGUAGE=pt_BR:pt:en
 PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games
 LANG=pt_BR.UTF-8
 SHELL=/bin/bash
PythonArgs: ['/usr/bin/hwdb-kde']
SourcePackage: hwdb-client
Uname: Linux Rick 2.6.20-14-generic #2 SMP Thu Apr 12 22:53:19 UTC 2007 i686 GNU/Linux

[Dan O'Huiginn (daniel-ohuiginn)]

I can trigger this bug by clicking on the (greyed-out) 'next' button during the 'preparing hardware data' stage

[agrabah (agrabah)]

I tried to trigger that bug again. This time I didn't write any non-ASCII characters (but this seemed to be another bug), and only pressed the (not-greyed out button) "next" during the "preparing hardware data" stage. At first it seemed to work as expected; but then I canceled at the stage where it asked me if I wanted to send data to the server. It quit normally, but after a few seconds the crash reporter popped up and said that the program in question quit unexpectedly.
A few secs later I tried to trigger it again (same procedure). No dice. I doesn't crash again. But I see another problem; the "next" button i described above seems to be working, and interrupts the "preparing hardware data" stage, which, I think, means that all the data is not collected. But after that it asks me if I want to send the data to the server, and quits with no problems.

Tine

[Dan O'Huiginn (daniel-ohuiginn)]

Attached is a patch to fix this. It could do with checking by somebody more familiar with hwdb than I am.

As far as I can tell, the bug is caused by:
1) clicking the next button during data assembly (as in comment 2)
2) anything else that stops /tmp/hwdb_data.xml being created (e.g. you can manually delete the file)

So the patch:
1) disables the next button during data assembly
2) catches errors in creating the md5 checksum

The patch *doesn't*:
 - fix the similar problem in gnome hwdb (bug #83267)
 - do anything about other causes and results of the data file not being created (e.g. bug #108337)

Also, it uses hashlib, which is new in python 2.5. I don't imagine that'll be an issue at all with gutsy, but the dependencies will need changing. [or I could rewrite to work with older pythons, of course]

If the patch seems OK, let me know and I'll do the same to the gnome version.

[Stéphane Loeuillet (leroutier)]

The very same error is here with hwdb-send when you try to submit a non-existing xml file.
If the file doesn't exist, it can't calculate its md5sum, it can't bzip2 it and it can't rename it

[Harald Sitter (apachelogger)]

I am closing this bug report because it is quite old and because the responsible developers dropped the package (it also isn't included in any release > 8.04).
Hopefully you can understand this decision.

Nonetheless I want to thank you for spending time on this bug report :)