Cursory analysis of a batch of ~900 Ubuntu bugs

[Matt Zimmerman]

On Mon, Mar 30, 2009 at 02:38:48PM +0100, Henrik Nilsen Omma wrote:
> [2] http://qa.ubuntu.com/reports/bugnumbers/bugs-since-2009-03-26.html

I found this report very interesting as a (representative?) sample of a
large batch of bugs.  I noticed the following patterns:

 * Nearly 25% had no package specified (filed on "Ubuntu" only).  Presumably
   we can infer from this that they were filed by non-expert bug reporters
   going directly to Launchpad, and will require help from a human just to
   get filed in the proper place

 * The next 10% of them are program crashes captured by apport

 * Another 10% are suspend/resume failures captured by apport

   Notably, there is at least one known issue here which causes false
   positive reports

 * A significant number are package install/upgrade failures (3-4%)

   About 1/3 of these, or 1% of the total sample of bugs, were package
   install/uprgade failures for kernel packages alone!  This is consistent
   with my casual browsing of the kernel bug list, which turned up a huge
   number of these bugs.  Bug 292159 is particularly notable as a cause of
   these bugs.

 * About 2% were Ubiquity bugs

Together, the above categories account for fully *half* of the bug reports.

(I couldn't trivially measure the number of duplicates using this report,
but noticed quite a few in browsing the list.  That would be a useful
category to pull out.)

-- 
 - mdz