Added support of X86::CMOVE instruction to the assembly instrumentation.

[ygorshenin]

Summary

Added support of X86::CMOVE instruction to the assembly instrumentation.

[eugenis]

What about other variations of cmov?
Why CMOVE*rr? These are register-to-register moves.
I think you need more complex instrumentation for cmov, as memory access is conditional.

[ygorshenin]

Closing the issue, as cmov instructions are not needed for rflags test.

lib/Target/X86/AsmParser/X86AsmInstrumentation.cpp

void X86AddressSanitizer::InstrumentMOV(
   case X86::MOV8mi:
   case X86::MOV8mr:
   case X86::MOV8rm:
     AccessSize = 1;
     break;
   case X86::MOV16mi:
   case X86::MOV16mr:
   case X86::MOV16rm:
+  case X86::CMOVE16rr:
+  case X86::CMOVE16rm:
     AccessSize = 2;
     break;
   case X86::MOV32mi:
   case X86::MOV32mr:
   case X86::MOV32rm:
+  case X86::CMOVE32rr:
+  case X86::CMOVE32rm:
     AccessSize = 4;
     break;
   case X86::MOV64mi32:
   case X86::MOV64mr:
   case X86::MOV64rm:
+  case X86::CMOVE64rr:
+  case X86::CMOVE64rm:
     AccessSize = 8;
     break;
   case X86::MOVAPDmr:
   case X86::MOVAPSmr:
   case X86::MOVAPDrm:
   case X86::MOVAPSrm:
     AccessSize = 16;
     break;