Archive for December, 2011

Krita bug statistics

December 21, 2011

A few days ago  Martin Gräßlin posted some statistics about bugs in KWin. That got me interested in doing some statistics about Krita.

The first thing I wanted to find out was how long bugs need to get fixed. To do that I wrote a script that extracted the data from the Bugzilla csv export, which is quite limited. There is some data available like the creation date of a bug or the date of the last change, but most details are not available directly. So to get the the data to calculate the lifetime of a bug, I approximated the closing date of a bug with the last change.

From the bug data I generated the following picture that shows all Krita bugs from 2004 till now. Each row shows one line whose length represents the time from report creation to it’s closing. Closed bugs are shown as black, new open bugs are green and unconfirmed bugs red. In the background you can see the length of the years and the blue vertical lines show major releases.

Although the diagramm is very simple, there are lots of things that can be discovered. First thing you might notice, is that it looks like there are some blue lines missing in the middle. That not a mistake, but the big gap were we were porting to KDE 4 (Surprised me a bit, as I thought that it was much longer than it looks).

For the bugs themself it shows that many bugs are closed near their creation, but there are also many bugs with evil long lines. After the release of Version 2.0 the curve get much steeper, which shows that more bugs got reported. It might sound strange, but as the software got more stable more bug reports were submitted. I think it just shows that Krita is used much more now and users work more with it and by that find more bugs. It also shows once again that the raw number of bug reports doesn’t mean anything.

Next I plotted how many bugs get opened and closed every year. Both have increased quite a lot of the last few years. As you can see we usually fix slightly less bugs than are reported (slightly above 90%), so the total number of bugs is still growing. No idea if that is better or worse than other projects.

Last thing I wanted to find out was how the bugs are resolved e.g. if they are really fixed in the classical sense or are just duplicates etc. So I extract the resolution from the data and plotted the four resolutions that are by far the most common for Krita. Not surprising of course is that “fixed” and “duplicate” are the most common. Although I was under the impression that we got a lot more duplicates in the last time, the actual number of duplicates isn’t that big (I also did that for Plasma to compare and it really has a crazy amount of duplicates).

One last detail: I know developers are often criticized for closing many bugs as wontfix and it turns out that this at least doesn’t apply for Krita (yet), as only about one percent of all bugs are closed as wontfix.