I'm personally alarmed whenever I hear the word 'Apple'. I wouldn't touch half a bit of what that company supports or does. It's by wide and far employed policies that are more harmful to users than Microsoft's.
Then again, LLVM and Clang both seem to be licensed bsd-like, so if it's similar to the Google situation with Chromium, it's acceptable.
Still, I wouldn't want to use anything written by anyone that gets their paycheck from Apple.
They have an svn repo where the source code resides, go check what those scary Apple engineers are up to.
http://clang.llvm.org/get_started.html
Though seems by that page that there are a few catches for early adaptors with even compiling C (C++ is way too immature to even mention):
"The semantic analyzer does not produce all of the warnings it should."
"We don't consider the API to be stable yet, and reserve the right to change fundamental things."
"The driver is currently implemented in python and is 'really slow'."
So this might at worst increase the required amount of maintenance significantly if put to use too early.
Clang-on-llvm will ultimately be a big improvement over GCC in a lot of respects (speed, ease-of-use), but it seems like it's currently very immature. Perhaps the BSD developers are doing this in order to get the development resources put towards clang in hopes that it will get the development resources necessary to become a real replacement for GCC. Perhaps it will also help them make their code more portable by having to remove GCC extensions from it.
No, it's really the license that interests them. They have some hate of sorts against GPL stuff. I highly doubt this LLVM will offer any real advantages over GCC. Ever.
I'd actually be more inclined to believe that reason than any other. The BSDs have historically tried to replace GPL'd software in their codebase, OpenBSD/NetBSD and the PCC is no exception.
de Raadt actually commented on it in more constructive vain:
Originally Posted by Theo de Raadt