[LLVMbugs] [Bug 4739] New: GlobalAlias of ConstantExpr emits incorrect code

[bugzilla-daemon at cs.uiuc.edu]

http://llvm.org/bugs/show_bug.cgi?id=4739

           Summary: GlobalAlias of ConstantExpr emits incorrect code
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: csdavec at swan.ac.uk
                CC: llvmbugs at cs.uiuc.edu


A global alias may be defined as a GEP pointing to an element in the middle of
a structure, in which case the generated assembly is incorrect.  Consider this
simple example:

%0 = type { i32, i32 }
@structure = internal global %0 { i32 0, i32 1 }
@element1 = alias getelementptr( %0* @structure, i32 0, i32 1)
@.str = private constant [12 x i8] c"Value:  %d\0A\00"

declare i32 @printf(i8*, ...)

define i32 @main() {
entry:
  %tmp1 = load i32* getelementptr( %0* @structure, i32 0, i32 1)
  call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([12 x i8]* @.str,
i32 0, i32 0), i32 %tmp1); <i32> [#uses=0]
  %tmp2 = load i32* @element1
  call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([12 x i8]* @.str,
i32 0, i32 0), i32 %tmp2); <i32> [#uses=0]
  ret i32 0
}

Running this prints:
Value:  1
Value:  0

This is because the @element1 alias is being resolved at to the global
referenced in the GEP, not to the result of the GEP.  Looking at the generated
assembly, this is obvious:

    .globl  element1
    .set     element1, structure

This ought to be:
    .globl  element1
    .set     element1, structure+4

Making this change in the asm gives:
Value:  1
Value:  1

This is the correct output.

If there is some other way of emitting a symbol in the middle of an LLVM
structure so I can work around this bug, please let me know.


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.