<p><br>
On Jan 17, 2011 9:20 AM, "Jon TURNEY" <<a href="mailto:jon.turney@dronecode.org.uk">jon.turney@dronecode.org.uk</a>> wrote:<br>
><br>
> This is more in the nature of a RFC, but this is the least-bad approach I have<br>
> managed to come up with.<br>
><br>
> Why do I want to do this? This set of changes lets me untie a few knots that<br>
> XWin finds itself in at the moment:<br>
><br>
> * libdix comes in 2 different flavours, depending on if it's<br>
> built with ROOTLESS defined or not, so DDXs which don't support ROOTLESS can't<br>
> be built at the same time as those that do.<br>
><br>
> * I'd like for a statically linked DDX to be able install another GLX provider<br>
> such that it gets chosen in preference to swrast, and to solve that problem<br>
> without creating another instance of the problem above.<br>
><br>
> * The first point at which the DDX gets to execute is either<br>
> ddxProcessArgument() (if arguments were supplied), otherwise OsVendorInit().<br>
> This uncertainly makes it unneccessarily painful to implement command line<br>
> options with a default depending on the environment.<br>
> (There's also a specific problem in that I'd like to have XWin default to<br>
> -nolock if the filesystem for /tmp is FAT (as it doesn't support the semantics<br>
> needed to create the lockfile successfully), but OsVendorInit() is too late to<br>
> check this, as by that time, the lockfile has been created.)<br>
><br>
> Jon TURNEY (6):<br>
> Move main() from dix to ddx<br>
> Replace DDXBEFORERESET with a more general way of doing DDX-specific<br>
> hooks<br>
> Add a DDX specific GLX provider push hook<br>
> Add a DDX specific hook into miPaintWindow for DDXen which support<br>
> the rootless extension<br>
> Make building XWin with WINDOWSWM and ROOTLESS mandatory<br>
> Revert libmain.a<br>
></p>
<p>Hi Jon,<br>
I'm the developer working on the X server for Android, Androix (<a href="http://androix.org">androix.org</a>) and your work on XQuartz was indispensible for that, so thanks.</p>
<p>I have a similar problem, though it doesn't effect the present Java JNI (as either main() or dix_main() will work), it will be an issue with native applications in 2.3.</p>
<p>I am all for this, as I don't have arguments or config files to parse but do need to negotiate some things or retreive settings from the Android application having a custom main() is neccessary and being able to call it what I want (for the native app version) is as well.</p>
<p>I can't comment on GLX (or a future GLXES) yet because I haven't begun implementing that in my server.</p>
<p>--<br>
Timothy Meade<br>
tmzt on freenode</p>