[PATCH] Consider the frame base isn't moving in calculateDbgValueHistory.

[David Blaikie]

>>! In D5366#5, @friss wrote:
> For the sake of completeness, it’s not the whole logic that is stopped by the call instructions, but only the liveness range of stack based variables that are unable to be propagated. This is especially an issue for ASAN instrumented code at O0 (and it’s really a pain not to be able to debug code compiled at O0). I think the code in there is still improves the situation with register variables, but I haven’t extracted any statistics to prove me right.

Right, but so far as I understand it this liveness range propagation you're describing was, I thought, the precise thing Alexey was trying to fix with a bunch of work/refactoring he did a few weeks ago. So I'm a bit surprised to learn that it didn't actually do what I thought it did - or that it had a substantial limitation. 

Perhaps I misunderstood the prior work.

http://reviews.llvm.org/D5366