During the Akonadi meeting in Berlin we decided that Akonadi needs a separate crash handler to handle Akonadi crashes. The reason is that nobody ‘owns’ the Akonadi instance. It could be accessed by Mailody, KMail, KNode or the plasma applet. Besides, Akonadi isn’t a real KDE application as it is cross-desktop and cross-platform, but that means the Akonadi crash it self can not be handled by the default KDE crash handler.

So, let me hereby present to you our new crash handler:

Some of the things in there, I want to discuss a bit more in detail. The first thing you notice is that we will be using clear error codes and error numbers. This is done because up front we don’t know which errors will occur. Before the release we will create a spot on our Akonadi website which lists those errors and the solutions so the user will know exactly what is going on. In a later version we will probably add a simple ‘feedback’ button so the user does not have to write it down and go to the website manually.

We had a rather large discussion about allowing the user to see the the backtrace. In a close vote we decided against it. People thought that the vast amount of output generated by a backtrace would just confuse the users and that the error number has proved to be enough.

As our approach has many similarities to the Microsofts way of reporting errors, we decided to give it the same look as well. As KDE starts to be true platform independend, it is no longer a problem to grab to good parts of Windows and introduce that to the KDE desktop. It has been acknowledged by our usability experts that this form of error report is the most approriate one since it has decades of wide scale testing across all ranges of target audiences.

Maybe other applications want to use this framework as well? Comments are welcome as usual in #akonadi