should consider faults when determining register liveness

_From [derek.br...@gmail.com](https://code.google.com/u/117968039472581148324/) on December 10, 2010 17:57:59_

PR 463053

My register stealing is a little over-agressive: I'm technically
being a little unsafe. Even if a register is dead on the normal
path, if a fault can occur before the register is written to, we
should restore its value in the fault handler. Pretty
pathological so for now we ignore it, given that we have larger
problems and we need the speed.

**\* Bug 485216 has been marked as a duplicate of this bug. ***

_Original issue: http://code.google.com/p/drmemory/issues/detail?id=142_