[LLVMbugs] [Bug 1148] NEW: Code quality issue with multiple inheritance

[bugzilla-daemon at cs.uiuc.edu]

http://llvm.org/bugs/show_bug.cgi?id=1148

           Summary: Code quality issue with multiple inheritance
           Product: tools
           Version: 1.0
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: llvm-g++
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: sabre at nondot.org


Consider:

struct base { virtual ~base(); int X; };
struct A { 
  virtual ~A(); 
int B;
};
struct Derived : public A, public base {
  int field;
};
int foo(base *B) {
  return static_cast<Derived*>(B)->field;
}

The static_cast in foo is actually compiled to the equivalent of "B ? B-8 : 0", producing this llvm code:

define i32 @_Z3fooP4base(%struct.A* %B) {
entry:
        %tmp = icmp eq %struct.A* %B, null              ; <i1> [#uses=1]
        br i1 %tmp, label %cond_next, label %cond_true

cond_true:              ; preds = %entry
        %B = bitcast %struct.A* %B to i8*               ; <i8*> [#uses=1]
        %ctg2 = getelementptr i8* %B, i32 -8            ; <i8*> [#uses=1]
        %tmp = bitcast i8* %ctg2 to %struct.Derived*            ; <%struct.Derived*> [#uses=1]
        %tmp36 = getelementptr %struct.Derived* %tmp, i32 0, i32 2              ; <i32*> [#uses=1]
        %tmp47 = load i32* %tmp36               ; <i32> [#uses=1]
        ret i32 %tmp47

cond_next:              ; preds = %entry
        %tmp4 = load i32* inttoptr (i32 16 to i32*)             ; <i32> [#uses=1]
        ret i32 %tmp4
}


However, because the pointer is loaded, we know that B could not have been a null pointer.  It's unclear 
how to handle this in LLVM, loading from a constant address is legal.  This has to be handled on the 
tree level somehow.

-Chris



------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.