[ubuntu-x] Tools for triaging -nvidia bugs

[Bryce Harrington]

Hi Alberto,

I promised to share some tools and ideas on tools for triaging -nvidia
bugs.

I've been accumulating my scripts in the arsenal project here:

  http://code.launchpad.net/arsenal/

You need to get and install python-launchpadlib-toolkit from the above
link.  Arsenal itself is just a collection of scripts that use this
toolkit.  See the README for installation directions.

(Don't laugh at my unskillful python packaging work here, I know it's
bad.  Patches or suggestions are of course most welcome.)

The contrib/ directory has some tools that may be of use for triaging
-nvidia bugs:

 * process-tagging.py:  This has some heuristics for identifying common
   X bug symptoms and applying tags accordingly.  I have only used this
   on -ati and -intel, so turn opt_dryrun on and run it a few times
   before turning it loose on nvidia in general.

 * ls-tag.py: This is a really simple script for listing bugs with a
   given tag.  Hint:  This is also a good starting point for writing
   your own scripts since it's so simple.

 * mark-nvidia-bugs.py: This is a script I used to ask people to re-test
   after we introduced the -180 driver.  You could repurpose this script
   with a new message and ask everyone to re-test again.  Also see
   retest.py which is a generalized version of this script.

 * pci-extract.py:  I use this to extract the pci id's from the lspci
   files people attach to the bugs.  I don't do this for nvidia; since
   we don't upstream nvidia bugs it seems pointless to do it, but YMMV.
   Note this script takes a while to run.

 * wontfix-all-bugs.py:  Closes all the bugs in a given source package.
   Don't use this one in anger!

 * expire-bugs.py:  This script gets run automatically every day
   already, including on -nvidia.  But it might be worth looking at for
   ideas on how to be more aggressive at expiring older -nvidia bugs.
   E.g. you can see the types of bugs this one leaves untouched, and
   selectively disable those checks to do a more aggressive expiration.


Now for some ideas of more tools that would be nice to have:

Xorg.0.log grepper:  Sarvatt had this idea.  Give it some string and
have the script search all Xorg.0.log files attached to bugs in a given
source package, and list out the bug #'s.  pci-extract.py could be a
good starting point for such a script since it is parsing through
attachments for info.

Check for manually-installed-nvidia: Use the Xorg.0.log grepper to look
for nvidia package versions that were not shipped in Ubuntu, or other
evidence that the reporter has installed nvidia manually from the web.
Close the bug report with a kindly word about how we don't support
configurations with externally obtained nvidia drivers, and point them
at the nvidia forums.

Nvidia crash traces:  We collect backtraces on nvidia bugs but in a lot
of cases it's clear from the trace that we'd never be able to debug
them.  It seems these ought to be detectable.  The reports should be
closed and the reporter redirected to report the issue directly to
nvidia via the forums.

Bryce