[llvm] r289266 - Fix LLVM's use of DW_OP_bit_piece in DWARF expressions.

[Joerg Sonnenberger via llvm-commits]

On Fri, Dec 09, 2016 at 08:43:40PM -0000, Adrian Prantl via llvm-commits wrote:
> Author: adrian
> Date: Fri Dec  9 14:43:40 2016
> New Revision: 289266
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=289266&view=rev
> Log:
> Fix LLVM's use of DW_OP_bit_piece in DWARF expressions.

Hello Adrian,
this breaks debugging on Sparc64. Try to build:

  long double a; fn1() { long double _Complex b = a = b * 1.0L; }

with -O2 -g -target sparc64 and it will assert. It seems like the new
logic in addMachineRegister gets confused by the multiple layers of
aliasing for Q0.

Joerg