Okay, this is a pretty intrusive commit.

Basically, I've reworked the system a bit to try and use the cmake variables a little more properly.

First, I've moved the qtdbus headers into header_list and removed the corresponding stuff from the .pl files. We can determine if the QtDbus stuff is installed via cmake now.

Second, I've commented out the call to grab_qglobal_symbols in qtguess.pl.cmake.

I can't see what it's actually helping to do anymore since we can use the compilation test to determine where there are QT_NO or QT_ defines. Moreso, the latest versions of Qt define license information in QT_EDITION variables in qglobal.h, and this subroutine cannot read them properly and parse them. This causes build failures due to errors about "QtLicenseForModule". Not calling this method anymore fixes these errors and doesn't seem to change the overall way things are built as best as I can tell.

Third, all of the headers in header_list have had the prefix directory stripped and I've reverted to the simple .h name. This is because we can grab each of the individual package directories from cmake, and for frameworked directories this makes things nicer.

Fourth, I've recommented out a lot of the excludes in generate.pl.cmake. I've actually been successful in getting smoke to build with some of these headers, so I thought perhaps starting over from scratch would be good here. We can re-uncomment them out as needed.

Last, I've reworked some of the build system items in CMakeLists.txt. This is again to be somewhat smart of how cmake is able to figure out things for us.

I've tested this on OSX (Framework Qt 4.2.2) and Linux (normal Qt-4.2.2) and it builds, compiles, and links on both. That's not to say I didn't make a mistake, or overlook something egregious. Please let me know if I did, hopefully in a nice way.