Re: Xquartz Crashes

Subject: Re: Xquartz Crashes
From: Ben Byer <email@hidden>
Date: Fri, 16 Nov 2007 15:52:38 -0800

On Nov 14, 2007, at 9:54 PM, Chip Griffin wrote:

I've been having fairly routine crashes since I upgraded to the latest lib and binary listed on the x.org/wiki site.

Below is the latest from a few moments ago. I was trying to resize an application window. The crash before that also happened when I was doing an interaction with a window (click and drag on a map which is supposed to trigger a zoom into that area). In each case it was Xquartz, not the other application, which crashed.

Below is the CrashReporter log in case it'll be helpful (I have no clue how to read it). I do notice some similarities with other crashes with respect to the stuff listed under the thread that crashed.

Yup, this is another Rootless / fb code crash. It is helpful, and let me see if I can help break it down a little bit (for you and anyone else who's curious):

When a program crashes, CrashReporter will take a snapshot of the state of all threads in the program, even ones that were not directly involved in the crash. Therefore, you can generally ignore every thread but the one that says "Thread <x> crashed":

Exception Type: EXC_BAD_ACCESS (SIGBUS)
Exception Codes: KERN_PROTECTION_FAILURE at 0x00000000000c9aa4
Crashed Thread: 1
[...]

Thread 1 Crashed:
0   Xquartz    0x000c094a fbBltOne + 2274
1   Xquartz    0x000dbdf8 fbPushFill + 450
2   Xquartz    0x000dbf9b fbPushImage + 343
3   Xquartz    0x000d3c7b fbPolyGlyphBlt + 729
4   Xquartz    0x00046067 miPolyText8 + 184
5   Xquartz    0x00148a42 damagePolyText8 + 247
6   Xquartz    0x00022fae RootlessPolyText8 + 225
7   Xquartz    0x0001420e doPolyText + 1695
8   Xquartz    0x00014431 PolyText + 140
9   Xquartz    0x00082cb6 ProcPolyText + 483
10 Xquartz    0x000fcbdf XaceCatchDispatchProc + 146
11 Xquartz    0x0007d531 Dispatch + 640
12 Xquartz    0x00098b7d main + 1921
13 Xquartz    0x0000c834 server_thread + 59
14 libSystem.B.dylib    0x91c17075 _pthread_start + 321
15 libSystem.B.dylib    0x91c16f32 thread_start + 34

Call stacks are read bottom-to-top. The function main() called the function Dispatch(), which called XaceCatchDispatchProc [...] which called RootlessPolyText8 [...] which called fbBltOne. The function that actually crashed was fbBltOne, and it crashed because it tried to access memory which did not belong to it (EXC_BAD_ACCESS).

"fb" stands for Frame Buffer, and it's the portion of the X server which does the actual drawing of pixels on the screen. You'll see "Blt" a lot -- that's short for "bitblitting" a bit more on Wikipedia here: http://en.wikipedia.org/wiki/Bitblt

Reading back up the chain, you see that this was called by the Rootless layer, while trying to draw some text on a window (RootlessPolyText). I already have a bug open here for this: <rdar://problem/5602290> rootless crash: nedit (fbBltOne / fbPushFill)

Here's the problem:

* The Rootless code is specific to OS X, and is much less tested than the rest of the cross-platform X code

* all of the 'fb' code is almost impossible to read and debug, because it's all highly optimized for speed.

Ben Byer

CoreOS / BSD Technology Group, XDarwin maintainer

 _______________________________________________
Do not post admin requests to the list. They will be ignored.
X11-users mailing list      (email@hidden)

This email sent to email@hidden

References:
	>Xquartz Crashes (From: Chip Griffin <email@hidden>)

Prev by Date: Re: Xterm not reading dotfiles
Next by Date: Re: Where art thou, xinit?
Previous by thread: Re: Xquartz Crashes
Next by thread: New Posting Limits on X11-Users
Index(es):
- Date
- Thread