February 10, 2013

Yet another KDE QA failure

by orzel
Categories: Gentoo, KDE
Tags: ,
Comments: 79 Comments

Just as lot of people here and elsewhere keep on denying KDE QA major problem, the 4.10 release is yet another proof of how bad the situation is.

KDE cares so much that they already closed the bug with a oh-so-easy “not our fault”. Which is probably true, but still I consider that

  • this shows a deep problem with how KDE tests  releases.
  • closing the bug is like telling their users “we dont care”.
  • declaring the bug solved is wrong, it is not.

For example Gentoo handles this better by not closing their own bug, and trying to fix it, working with upstream (here, the Qt Project). Some projects care about their users, some not.

The irony there is that based on comments from this thread I had installed KDE 4.10 to give it another try and check if it was usable compared to when i gave up (~15 months ago). I tried from an new/test/empty account for fairness.

It was such a mess! No startup dialog, empty default panel and plasma restartint (aka crashing) very often. I mean RELIABLY crashing, not even randomly. My favourite way to reproduce is to try to add a system tray applet to the panel. I couldn’t have such a major component as the systray.  Without systray, no way to get back half of my background applications, actually started but invisible.

But hey, having a working systray is so much less fun than having a webpage embedded in a plasmoid, dude.

Once I recompiled qt-core with -Os, which is the most reliable known workaround currently, things went a lot smoother, indeed, and all the obvious bugs/crashes from this 5minutes test were fixed.  But what for a fix is that ? Not everybody is using Gentoo and can do that easily.

Don’t get me wrong, the bug is a huge one and it is not KDE’s fault. It’s a Qt bug, acknowledged as such by the Qt project and it seems to find its roots very deeply in Qt. We are used to see Qt bugs closed quickly, and the fact that this bugs is still not fixed is a clear indication of how difficult the problem is.

But come on, my setup is such a common one. Linux kernel, amd64, no fancy compiler, no fancy compiler options, the most standard system ever.

The KDE ticket was opened long ago (december 15th, during beta2), confirmed by several people on several distributions, and the bug was clearly identified as linked to Qt, so it was kinda obvious every distribution will be hit. This has been confirmed since then.

And guess what KDE did ?

Nothing, they released it as is, their conscience clear with the bug being closed. They do not have the notion of showstopper. Nothing new, we’ve  known that since 4.0. Making things work is no fun, this should rather be distributions’ duty.