Packages to investigate

Brian Murray brian at ubuntu.com
Fri Dec 3 22:58:23 UTC 2010


On Tue, Nov 30, 2010 at 11:51:19AM +0100, Jean-Baptiste Lallement wrote:
> On 11/22/2010 10:22 PM, Brian Murray wrote:
> > As we all know there are a lot of packages and bugs to keep track of in
> > Ubuntu.  One idea that occurred to me recently, to help identify
> > packages in need of triage, was to calculate an average bug heat for the
> > particular package.  I've made a first pass at this using all the
> > packages in the ubuntu-desktop package set.  Here are the top 10
> > packages:
> > 
> > usb-modeswitch-data - median: 764, mode: 1448
> > language-pack-gnome-fa-base - median: 408, mode: 408
> > language-pack-gnome-pt - median: 318, mode: 318
> > couchdb-glib - median: 210, mode: 210
> > netbook-meta - median: 145, mode: 408
> > pyopenssl - median: 136, mode: 259
> > shotwell - median: 124, mode: 6
> > appmenu-gtk - median: 122, mode: 408
> > gnome-python-extras - median: 114, mode: 3
> > telepathy-gabble - median: 104, mode: 49
> > 
> > As I ran this last Friday there might be some variance in the numbers.
> > I'm curious whether or not you think the bug reports for these packages
> > really need reviewing.
> > 
> > Thanks,
> > 
> Thanks Brian. This is really interesting. If we apply this calculation
> to the packages installed by default in Natty, the top 10 looks like:
> shotwell - median: 189
> empathy - median: 91
> gcalctool - median: 86
> gbrainy - median: 86
> gwibber - median: 81
> simple-scan - median: 76
> software-center - median: 68
> pitivi - median: 68
> transmission - median: 60
> gnome-bluetooth - median: 59
> 
> Compared to the list based on the number of bugs:
> evolution
> firefox
> gdmsetup
> nautilus
> nm-connection-editor
> ooffice
> software-center
> totem
> update-manager
> 
> So I can not say that this is the list that I had in mind, but the
> result is expected.
> When we build the list based on volume of bugs, we get the most used
> applications. But when we use the bug heat we get the applications with
> fewer bugs but which are in need of triage (I'm not saying that the
> applications with a large number of bugs don't need triage)
> 
> The list needs a closer look though. For instance:
> - shotwell is in 1rst position, but when I look at the reports, there
> are 2 bugs with a heat of 3491, and I don't see anything specific that
> could explain such a value. Is it the weight of untriaged report being a
> bit too important or something else ? It doesn't seem to match the bug
> heat algorithm.

Looking closer at the bug heat algorithm it seems that a fair bit of it
is based on recent activity.  From some of the documentation the theory
follows:

Bug has been active within the past 24 hours
    Add 25% of the project's hottest bug's score divided by the number
of days since the first activity on the bug in question

Bug has not been active* in within the past 24 hours
    Subtract 1% of the bug heat score for every day of inactivity

So it seems to follow that the newest bugs will the highest heat.

I guess there is a larger question of what we think makes a bug hot.
Initially, I thought the number of users affected, number of subscribers
and number of duplicates (among some other things) was a good indicator.
Two out of those three things are now cached on the bug table so we
could recreate the bug heat without taking into account recent activity.
I feel like this might provide a more useful number.

> - gbrainy, I was surprised to find a game in the list. In fact, there is
> a very limited number of bugs filed against this package and some of
> them with high heat. I've triaged the report with the highest heat
> (>300) and it fell to a heat of 12. So the next run of your script
> should move it lower into the list.
> - gcalctool, same thing, the first bug had a heat of 237, and after
> asking for more information it fell to 8.
> 
> So, yes, this is useful to help us to direct our testing and triaging
> effort. But the calculation of the bug heat needs to be clarified.
> Maybe this could be added to the 'Opportunities' list in harvest ?

By this do you mean the hottest, those outside the standard deviation,
bugs for a package should be listed as an opportunity in harvest?

Thanks,
-- 
Brian Murray
Ubuntu Bug Master
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <https://lists.ubuntu.com/archives/ubuntu-quality/attachments/20101203/e6eca631/attachment.pgp>


More information about the Ubuntu-qa mailing list