[llvm-bugs] [Bug 34242] New: inlining a function into itself can result in references to version of variable from wrong frame

via llvm-bugs llvm-bugs at lists.llvm.org
Fri Aug 18 17:45:32 PDT 2017


https://bugs.llvm.org/show_bug.cgi?id=34242

            Bug ID: 34242
           Summary: inlining a function into itself can result in
                    references to version of variable from wrong frame
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Interprocedural Optimizations
          Assignee: unassignedbugs at nondot.org
          Reporter: richard-llvm at metafoo.co.uk
                CC: llvm-bugs at lists.llvm.org,
                    sanjoy at playingwithpointers.com

Reduced miscompiled testcase: (reduced from libaf's af_pan.c)

define i32 @test(i1 %init, i8* %addr) {
bb:  
  %n = alloca i32
  br i1 %init, label %store, label %load

store:
  store i32 0, i32* %n
  %cast = bitcast i32* %n to i8*
  %v = call i32 @test(i1 0, i8* %cast)
  ret i32 %v

load:
  %castback = bitcast i8* %addr to i32*
  %n.load = load i32, i32* %castback
  ret i32 %n.load
}


After inlining, we get this:


define i32 @test(i1 %init, i8* %addr) {
bb:
  %n.i = alloca i32
  %n = alloca i32
  br i1 %init, label %store, label %load

store:                                            ; preds = %bb
  store i32 0, i32* %n
  %cast = bitcast i32* %n to i8*
  %0 = bitcast i32* %n.i to i8*
  call void @llvm.lifetime.start.p0i8(i64 4, i8* %0)
  %n.load.i = load i32, i32* %n.i
  %1 = bitcast i32* %n.i to i8*
  call void @llvm.lifetime.end.p0i8(i64 4, i8* %1)
  ret i32 %n.load.i

load:                                             ; preds = %bb
  %castback = bitcast i8* %addr to i32*
  %n.load = load i32, i32* %castback
  ret i32 %n.load
}

This is wrong: %n.load.i is a load of the *wrong variable*. We should be
loading %n, not %n.i.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20170819/08758b31/attachment.html>


More information about the llvm-bugs mailing list