Bug 456399

Created attachment 312461 [details]
Log output from xscreensaver

CCing to jwz.

Need a stack trace.

Yeah, will try to capture a trace in a moment ...

Pasting in here 'cause it's fairly short

#0  0x0011e402 in __kernel_vsyscall ()
#1  0x00149690 in raise () from /lib/libc.so.6
#2  0x0014af91 in abort () from /lib/libc.so.6
#3  0x08051b65 in saver_exit (si=0xbf945244, status=-1, 
    dump_core_reason=0x806aff8 "because of synchronous X Error") at
../../driver/windows.c:904
#4  0x0804d5c8 in saver_ehandler (dpy=0x99a6430, error=0xbf944de8)
    at ../../driver/xscreensaver.c:336
#5  0x0030e71e in _XError (dpy=0x99a6430, rep=0x99ae160) at XlibInt.c:2905
#6  0x00316263 in _XReply (dpy=0x99a6430, rep=0xbf944f10, extra=0, discard=0) at
xcb_io.c:417
#7  0x0066f77b in XRRGetScreenResources (dpy=0x99a6430, window=0) at XrrScreen.c:81
#8  0x080538b3 in update_screen_layout (si=0xbf945244) at ../../driver/screens.c:386
#9  0x0804d893 in main (argc=Cannot access memory at address 0x6
) at ../../driver/xscreensaver.c:784
#10 0x00136390 in __libc_start_main () from /lib/libc.so.6
#11 0x0804bbc1 in _start ()

As the error appears within XRRGetScreenResources I'm assuming this is probably
a bug in libXrandr or in the X server rather than xscreensaver, feel free to
reassign it if you like.

I don't know how RootWindow(dpy,i) is returning 0, unless you have both A) multiple real X screens 
(ScreenCount > 0) and B) RANDR stuff happening at the same time.  I wasn't even sure that was possible.  
But I can't tell whether that's the case because it died before it was able to print out a debug log of your 
screens and geometry.

Also, update_screen_layout does not call XRRGetScreenResources, so that stack trace is garbled. I suspect 
you're not compiled with -g.

I do indeed have multiple real X screens, and was running from a terminal on
screen 1, although I wouldn't have expected that to matter ...

Anyway, it appears that randr_scan_monitors() has been inlined, which gcc likes
to do these days; repeated backtrace is the same. I double checked that I have
the correct version of the debuginfo package installed.

#0  0x0011e402 in __kernel_vsyscall ()
#1  0x00149690 in raise () from /lib/libc.so.6
#2  0x0014af91 in abort () from /lib/libc.so.6
#3  0x08051b65 in saver_exit (si=0xbf86f264, status=-1, 
    dump_core_reason=0x806aff8 "because of synchronous X Error") at
../../driver/windows.c:904
#4  0x0804d5c8 in saver_ehandler (dpy=0x973e450, error=0xbf86ee08)
    at ../../driver/xscreensaver.c:336
#5  0x0030e71e in _XError (dpy=0x973e450, rep=0x9746188) at XlibInt.c:2905
#6  0x00316263 in _XReply (dpy=0x973e450, rep=0xbf86ef30, extra=0, discard=0) at
xcb_io.c:417
#7  0x0066f77b in XRRGetScreenResources (dpy=0x973e450, window=0) at XrrScreen.c:81
#8  0x080538b3 in update_screen_layout (si=0xbf86f264) at ../../driver/screens.c:386
#9  0x0804d893 in main (argc=Cannot access memory at address 0x6
) at ../../driver/xscreensaver.c:784
#10 0x00136390 in __libc_start_main () from /lib/libc.so.6
#11 0x0804bbc1 in _start ()

Lemme know if there's anything else I can do (have the exploded source here, can
add printf()s or whatever tomorrow).

Please build and run "test-randr" from the xscreensaver-5.06/driver/ src.

Also, xdpyinfo.

Also, by "real X screens" I mean you address them with $DISPLAY :0.0 and :0.1.  Just because you have 
mutliple monitors doesn't mean you have real X screens.  That's kind of the whole point of 
RANDR/Xinerama.

(In reply to comment #10)
> Please build and run "test-randr" from the xscreensaver-5.06/driver/ src. 
> Also, xdpyinfo.
> 

Hi, Bill:
For now I created a subpackage "xscreensaver-tests" which contains some binaries
for testing like "/usr/libexec/xscreensaver/test-randr" or so. Would you try this
from the following?
http://kojipkgs.fedoraproject.org/packages/xscreensaver/5.06/1.fc8.1/

xdpyinfo is in xorg-x11-utils package.

For Jamie:
Do you want me to ship test-* binaries under driver/ directory for Fedora
xscreensaver rpms
from next time?

No, they're separate protocol screens (.0 and .1) - I have two cards, and this
way I get DRI on both heads, hence nice fast 3d screensavers :o)

Created attachment 312539 [details]
Output from test-randr

The "event" I sent it was just changing the refresh rate on one screen, hence
the lack of apparent change.

Created attachment 312575 [details]
patch

Ok, I think I just fixed this -- but the other folks weren't getting crashes. 
(Congratulations, I think you're the first person ever to try and use real
multi-screen on a system with randr 1.2 available...)

Jamie: you wouldn't believe how many people tell me "don't do that" ;o)

Thanks for patch, building now ...

... fixed, thanks.

Good news, thanks!

> For Jamie:
> Do you want me to ship test-* binaries under driver/ directory for Fedora
> xscreensaver rpms
> from next time?


Nah, they're almost never needed...

xscreensaver-5.06-2.fc9 has been submitted as an update for Fedora 9

xscreensaver-5.06-2.fc8 has been submitted as an update for Fedora 8

It seems the update packages are already pushed.

If the bug still exists, please feel free to reopen this bug.
Thank you, Bill and Jamie. Closing.