SIGFPE in create_dpis() in QPaintDevice::~QPaintDevice

[Simon Law]

The fglrx binary driver has a bug in it when you suspend and resume or when you use XRandR. It will set your screen size to 0x0mm, which is an impossible situation.

In this nonsense case, we want to assume the screen is some reasonable default so that our applications will just work. This could be the Xft.dpi resource or if that doesn't exist, then fallback to 96dpi which is the KDE standard.

[Simon Law (sfllaw)]

I'll hack on this.

[Daniel Holbach (dholbach)]

Simon: how's this going?

[Ken Phillis Jr (kphillisjr)]

i can confirm this bug, it occurs when your using TV output.

[Ken Phillis Jr (kphillisjr)]

i got the information that is required, and it's not to hard.... just look at the xdpyinfo for each effected box, it should turn out somethin like this ( notice screen # 1 on my box )

xdpyinfo | grep -a2 dimen

screen #0:
  dimensions: 1280x1024 pixels (370x280 millimeters)
  resolution: 88x93 dots per inch
  depths (7): 24, 1, 4, 8, 15, 16, 32
--

screen #1:
  dimensions: 800x600 pixels (0x0 millimeters)
  resolution: -2147483648x-2147483648 dots per inch
  depths (7): 24, 1, 4, 8, 15, 16, 32

[Ken Phillis Jr (kphillisjr)]

got a patch that fixes the issues with the dpi's, but the problem might still occure with other apps, please resort to testing them as to figure this out, and so far i am certain that the new control panel for ati cards that is in the binary package is not fixed by this repair.

[Simon Law (sfllaw)]

The patch looks like it solves the problem, although you probably only want one test for (DisplayWidthMM(dpy,i) < 1).

Thanks!

[Ken Phillis Jr (kphillisjr)]

actually, i wanted to check the height and width, and if i am checking width twice, taht means that the second test needs to be for this statement here:

(DisplayHeightMM(dpy,i) < 1)

instead of what it is now, which is:

(DisplayWidthMM(dpy,i) < 1)

change that line to the appropriate check and it'll work even better, because it won't let the applications crash due to having a 0 for the height and/or width.

[Ken Phillis Jr (kphillisjr)]

just upgraded to feisty, and the pesky bug returned, but of course i didn't patch my install with my own patch so it is verified.

[Ken Phillis Jr (kphillisjr)]

I just updated the patch to feisty level, so here's the new patch.

[Jonathan Riddell (jr)]

Patch applied in qt-x11-free_3.3.8really3.3.7-0ubuntu4. Please try it once it has compiled and found its way to the mirrors and let me know if it fixes the problem.

[Ken Phillis Jr (kphillisjr)]

on my machine it works just fine, no word from others who had this issue.