fix D78849 for g++ < 7.1

[nickdesaulniers]

Summary

Looks like g++ < 7.1 has a bug resolving calls to member functions without
this-> in lamdas with auto types. It looks like multiple build bots are
using g++-5.

https://stackoverflow.com/questions/32097759/calling-this-member-function-from-generic-lambda-clang-vs-gcc
https://godbolt.org/z/MiaRt-

[MaskRay]

llvm/lib/CodeGen/MachineVerifier.cpp

172

I think [&](unsigned Reg) { return addRequired(Reg); } works as well...

There might be a this capturing bug in older GCC.

I seem to recall older MSVC has similar bugs.

[nickdesaulniers]

committed as 59acdf0aca10b78c5d3885f9721779585593074a

llvm/lib/CodeGen/MachineVerifier.cpp

172

Specifically, it's the case below that's problematic. The use of auto is problematic. If I use & capture and spell out the type, it works for g++-5.1, but I think spelling out the type hurts readability, compared to a call through this.

llvm/lib/CodeGen/MachineVerifier.cpp

struct BBInfo {
           return false;
         if (regsLiveOut.count(Reg))
           return false;
         return vregsRequired.insert(Reg).second;
       }
 
       // Same for a full set.
       bool addRequired(const RegSet &RS) {
-        return llvm::any_of(RS,
-                            [this](unsigned Reg) { return addRequired(Reg); });
+        return llvm::any_of(
+            RS, [this](unsigned Reg) { return this->addRequired(Reg); });
       }
 
       // Same for a full map.
       bool addRequired(const RegMap &RM) {
         return llvm::any_of(
-            RM, [this](const auto &P) { return addRequired(P.first); });
+            RM, [this](const auto &P) { return this->addRequired(P.first); });
       }
 
       // Live-out registers are either in regsLiveOut or vregsPassed.
       bool isLiveOut(unsigned Reg) const {
         return regsLiveOut.count(Reg) || vregsPassed.count(Reg);
       }
     };