On Friday 30 September 2011 09:44:54 Scott Kitterman wrote: > Having fewer users or fewer bugs would help, so it may be that making the > software either better or worse would reduce the bug triage backlog. Sorry to disagree, in KWin we see the opposite. The better we become the bigger the (useless) backlog :-)
I just want to point out my experience as I read each and every comment on a KWin bug (about 20 new bugs per week, 10 to 20 mails per day): the most time I spend on KDE development is reading and managing bugs. If there would not be Thomas supporting keeping the bug tracker clean, I would have to give up. From time to time I go through and triage new bugs. To reduce the number of open bug reports by ten (that is setting to duplicate, asking for more info, etc) I need about one hour. In comparison it took me about 20 hours to add the initial Wayland support to KWin. If I triage 10 bugs (spending one hour) I have not fixed a single bug, in fact I hardly fix any bug shown from bugzilla. In the complete month September I did one commit to the 4.7 branch. The bugtracker in it's current state does not help to increase the quality of the software. In fact the quality decreases as the developers have to spend time on managing garbage. Yes what we get is garbage. Most of the reports in KWin are either a duplicate of a driver bug (which has so many duplicates that you cannot miss it) or is a duplicate which is already fixed. Most bugs we currently get are either version 4.6.2 (latest version in Kubuntu) or 4.6.0 (latest version in openSUSE). If at all there are useful bugs they come from Arch users with a recent version. Non-Crash reports are mostly user support issues: that means you have to iterate to get the information (that's what Anne mentioned). This is great that we in KWin provide this help but it's not what we developers should do and it does not help to improve the software. If we have such a report it is completely useless if you want to fix a bug afterwards. First you have to go through some pages of discussion before finding the real data. Another issue is that multiple users report to the same bug with issues which look similar, but aren't. It clutters the report and makes it impossible to get what the bug was about. In summary: useless. Another issue is that the users have a different language than we have. What they talk about is not what we developers need. We need clearly reproducable bug reports. The summary has to be in a way that I can understand what the user's issue is and that I can map it directly to where in the source the incorrect behavior is triggered. Hardly any bug report is in that state: they are all useless. So of 20 open bugs there might be one which is in a state that I could fix it. This means I have to read through 20 bugs just to prepare to fix it. If I have one hour of spare time to fix bugs, I cannot succeed because I won't find a bug to fix. A bugtracker is a great tool to manage issues and to distribute and plan the work. But we don't use the bugtracker in that way. We seriously need a layer in between the developers and the users. No company would allow users to communicate with the developers directly. There would be a user support channel in between. And only valid and good prepared issues would be passed to the developers. This is what we need. We need to have a developer only bug tracker. It may be readable for users, but only developers may write it. And with that I mean developers, not like it's now that if I open a bug report against KMail it's set to "New". Last week I contacted the KDE Forums team to try an experiment of using the forum to manage the KWin bugs to get only bugs in the bugtracker without the support issues and without the duplicate issues (most duplicates can be removed by just mentioning the workaround). I hope this will get started and will work out. Obviously we need way more users helping on the forums to have this idea being successful. I hope it's easier to attrackt developers to the forum than to the bugtracker. Executive summary: our bugtracking system is broken and it makes no sense to try to fix it with the current solution of hoping there will be users triaging it. Users and developers may not communicate in the same issue tracker. Sorry for the long mail Martin
signature.asc
Description: This is a digitally signed message part.
>> Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<
