[Xquartz-dev] Server binary 1.7.1.901

[Eeri Kask]

Am 11/07/2009 01:49 AM, Jeremy Huddleston <jeremyhu at apple.com> schrieb:
> So along with the 2.4.1_beta1 release, some of you braver souls might like to replace the 1.6 server binary with one built from the 1.7 branch.  This binary contains everything from the 1.7.1.901 release that was made earlier today as well as a few patches that didn't make the deadline but will probably wind up in 1.7.2 when it finally ships.
> 
> http://static.macosforge.org/xquartz/downloads/SL/X11.bin-1.7.1.901.bz2
> 
> Just download, decompress, and replace /A/U/XQuartz.app/Contents/MacOS/X11.bin with it.
> 
> If all goes well, I'll switch over to the 1.7 server for the next release, but I want some more feedback first.  I've been using it and master (1.8) extensively, and this version is what is currently in MacPorts, so I do believe it is release-quality.  Please let me know if you run into any trouble.


May I ask is there any version for intel platform non-snow leopard
to download?  ... mostly in order to check how the "border issue"
(resp. background issue) is evolving.


Otherwise 1.5.3-apple14 has proven very stable too since its
release; there have been only rare crashes while doing some
AppleWM-extension stressing, traces usually going to the pixman
library, like

Thread 2 Crashed:
0   libpixman-1.0.dylib           0x003f2b1e pixman_fill + 170
1   X11.bin                       0x000323c9 fbFill + 294
2   X11.bin                       0x000329fd fbPolyFillRect + 299
3   X11.bin                       0x000170e7 DRIPolyFillRect + 175
4   X11.bin                       0x00076804 damagePolyFillRect + 559
5   X11.bin                       0x000e4b99 miPaintWindow + 524
6   X11.bin                       0x000f488e
miHandleValidateExposures + 77
7   X11.bin                       0x000f5a44 miChangeBorderWidth + 336
8   X11.bin                       0x00022603
RootlessChangeBorderWidth + 442
9   X11.bin                       0x000d927b ConfigureWindow + 2516
10  X11.bin                       0x000b7795 ProcConfigureWindow + 116
11  X11.bin                       0x000bcd26 Dispatch + 739
12  X11.bin                       0x000cb4a9 dix_main + 1312
13  X11.bin                       0x000147b4 server_thread + 53
14  libSystem.B.dylib             0x9217a155 _pthread_start + 321
15  libSystem.B.dylib             0x9217a012 thread_start + 34



Then, probably it is off topic as not being particularly Apple or
XQuartz related and goes back even to XFree86 X11 Tiger releases,
but it seems sometimes the server misses, i.e. probably does not
deliver some events, especially "EnterNotify" events in certain
circumstances (i.e. if some new client window opens exactly under
the mouse => the ancient "XLess" "search" transient window being the
most prominent and reproducible test case).  It appears under
XQuartz this phenomenon is the easiest reproducible, even often with
other clients like xterm.

This matters insomuch as it intervenes with the focus management in
most window managers which rely on that particular EnterNotify
event; so one has to move the mouse out-of- and again into that
window.   :-)



Greetings,

    Eeri Kask

(P.S. To prove/observe this issue one only has to let some simple
event scanner run (at Xlib level at the client side) on a blank
X11-server which prints out all Enter- and related events the server
delivers to any window.  According to the description given in
"Scheifler/Gettys Xlib book" it seems if some client window is
opened exactly under the mouse the CreateNotify, MapNotify,
EnterNotify have to arrive at the client.)