[LLVMbugs] [Bug 12079] New: crash with opaque member

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=12079

             Bug #: 12079
           Summary: crash with opaque member
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Scalar Optimizations
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: nicholas at mxc.ca
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified


Testcase 1:

  %struct.B = type opaque
  %struct.A = type {i32, %struct.B, i8}

  define i32* @test(%struct.A* %x) {
    %A = bitcast %struct.A* %x to i32*
    ret i32* %A      
  }

and testcase 2:

  %struct.B = type opaque
  %struct.A = type {i32, %struct.B, i8}

  @gv = global %struct.A zeroinitializer

  define i32* @test() {
    %A = bitcast %struct.A* @gv to i32*
    ret i32* %A      
  }

both crash trying to build a getelementptr, the first in instcombine and the
second in early-cse (in inst-simplify, in constant folding).

Are all-zero GEPs supposed to be valid on partially-opaque types? Is that type
supposed to be rejected? If you use %struct.B for the test then you'll find
that it stays a bitcast (instruction or constant expression) thus avoiding the
crash. Should that logic be extended?

-- 
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.