Since rust contains an entire copy of llvm and builds it while completely ignoring the system copy, it would be nice if there was at least a way to tell it not to waste half the build compiling dead code like the amdgpu or nvptx targets.
(In reply to Anthony Parsons from comment #0) > Since rust contains an entire copy of llvm and builds it while completely > ignoring the system copy, it would be nice if there was at least a way to > tell it not to waste half the build compiling dead code like the amdgpu or > nvptx targets. Agreed. Also, is it really necessary to use the bundled copy of llvm rather than system llvm? And, if it is, are there quantifiable benefits of using llvm rather than the system gcc, i.e. what is the real point of the clang use flag (whether correctly used or not)?
We used to support the system LLVM but due to the way Gentoo modified LLVM it made it not work. I cannot recall the underlying bugs but I'm sure you can find them in Bugzilla. I can take a second look to see if things have improved since.
*** This bug has been marked as a duplicate of bug 627288 ***