Index: llvm/trunk/lib/CodeGen/PHIElimination.cpp
===================================================================
--- llvm/trunk/lib/CodeGen/PHIElimination.cpp
+++ llvm/trunk/lib/CodeGen/PHIElimination.cpp
@@ -197,12 +197,11 @@
 /// EliminatePHINodes - Eliminate phi nodes by inserting copy instructions in
 /// predecessor basic blocks.
 bool PHIElimination::EliminatePHINodes(MachineFunction &MF,
-                                             MachineBasicBlock &MBB) {
+                                       MachineBasicBlock &MBB) {
   if (MBB.empty() || !MBB.front().isPHI())
     return false;   // Quick exit for basic blocks without PHIs.
 
-  // Get an iterator to the first instruction after the last PHI node (this may
-  // also be the end of the basic block).
+  // Get an iterator to the last PHI node.
   MachineBasicBlock::iterator LastPHIIt =
     std::prev(MBB.SkipPHIsAndLabels(MBB.begin()));
 
Index: llvm/trunk/test/CodeGen/PowerPC/2013-07-01-PHIElimBug.ll
===================================================================
--- llvm/trunk/test/CodeGen/PowerPC/2013-07-01-PHIElimBug.ll
+++ llvm/trunk/test/CodeGen/PowerPC/2013-07-01-PHIElimBug.ll
@@ -1,28 +0,0 @@
-; RUN: llc < %s -verify-machineinstrs | FileCheck %s
-
-target datalayout = "E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-f128:128:128-v128:128:128-n32:64"
-target triple = "powerpc64-unknown-linux-gnu"
-
-@g_51 = external global [8 x i32], align 4
-
-; CHECK: func_7
-
-; Function Attrs: nounwind
-define fastcc void @func_7() #0 {
-entry:
-  %arrayidx638 = getelementptr inbounds [3 x [1 x i32]], [3 x [1 x i32]]* undef, i64 0, i64 1, i64 0
-  br i1 undef, label %for.cond940, label %if.end1018
-
-for.cond940:                                      ; preds = %for.cond940, %if.else876
-  %l_655.1 = phi i32* [ getelementptr inbounds ([8 x i32], [8 x i32]* @g_51, i64 0, i64 6), %entry ], [ %l_654.0, %for.cond940 ]
-  %l_654.0 = phi i32* [ null, %entry ], [ %arrayidx638, %for.cond940 ]
-  %exitcond = icmp eq i32 undef, 20
-  br i1 %exitcond, label %if.end1018, label %for.cond940
-
-if.end1018:                                       ; preds = %for.end957, %for.end834
-  %l_655.3.ph33 = phi i32* [ %l_655.1, %for.cond940 ], [ getelementptr inbounds ([8 x i32], [8 x i32]* @g_51, i64 0, i64 6), %entry ]
-  store i32 0, i32* %l_655.3.ph33, align 4
-  ret void
-}
-
-attributes #0 = { nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "unsafe-fp-math"="false" "use-soft-float"="false" }
Index: llvm/trunk/test/CodeGen/PowerPC/2013-07-01-PHIElimBug.mir
===================================================================
--- llvm/trunk/test/CodeGen/PowerPC/2013-07-01-PHIElimBug.mir
+++ llvm/trunk/test/CodeGen/PowerPC/2013-07-01-PHIElimBug.mir
@@ -0,0 +1,92 @@
+# RUN: llc -mtriple powerpc64-unknown-linux-gnu -run-pass livevars -run-pass phi-node-elimination -verify-machineinstrs -o - %s | FileCheck %s
+
+# This test case was originally known as
+#   test/CodeGen/PowerPC/2013-07-01-PHIElimBug.ll
+#
+# It was discovered that the original test case no longer reproduced the bug
+# from PR16508 (solved in rL185363). That could have been resolved by adding
+# -O1 (or possibly -O0) to the run line, but instead the test case was
+# converted into this .mir test case. Having it as a .mir test should make it
+# less sensitive to changes in earlier passes.
+
+--- |
+  target datalayout = "E-m:e-i64:64-n32:64"
+  target triple = "powerpc64-unknown-linux-gnu"
+
+  @g_51 = external global [8 x i32], align 4
+  define void @func_7() {
+    bb.0:
+      ret void
+  }
+...
+---
+name:            func_7
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $x2
+
+    %0:g8rc_and_g8rc_nox0 = ADDIStocHA $x2, @g_51
+    %1:g8rc_and_g8rc_nox0 = LDtocL @g_51, killed %0, implicit $x2 :: (load 8)
+    %2:gprc = LI 0
+    %3:crrc = CMPLWI killed %2, 0
+    BCC 76, killed %3, %bb.2
+
+  bb.1:
+    %4:g8rc = ADDI8 killed %1, 24
+    B %bb.4
+
+  bb.2:
+    %5:g8rc = ADDI8 killed %1, 24
+    %6:g8rc = LI8 0
+    %7:crrc = CMPLWI undef %8:gprc, 20
+
+  bb.3:
+    %9:g8rc = PHI %5, %bb.2, %10, %bb.3
+    %10:g8rc = PHI %6, %bb.2, undef %11:g8rc, %bb.3
+    BCC 68, %7, %bb.3
+    B %bb.4
+
+  bb.4:
+    %12:g8rc_and_g8rc_nox0 = PHI %4, %bb.1, %9, %bb.3
+    %13:g8rc = LI8 0
+    STW8 killed %13, 0, killed %12 :: (store 4)
+    BLR8 implicit $lr8, implicit $rm
+
+...
+
+# Original TR (and 2013-07-01-PHIElimBug.ll) was about verifier errors for bb.3.
+#
+# I got a feeling that we also need to have some checks to see that # the code
+# is correct in some way. Hopefully this test case is stable enough to verify
+# the full MIR like this.
+#
+# CHECK:       bb.0:
+# CHECK:         %0:g8rc_and_g8rc_nox0 = ADDIStocHA $x2, @g_51
+# CHECK-NEXT:    %1:g8rc_and_g8rc_nox0 = LDtocL @g_51, killed %0, implicit killed $x2 :: (load 8)
+# CHECK-NEXT:    %2:gprc = LI 0
+# CHECK-NEXT:    %3:crrc = CMPLWI killed %2, 0
+# CHECK-NEXT:    BCC 76, killed %3, %bb.2
+# CHECK:       bb.1:
+# CHECK:         %4:g8rc = ADDI8 killed %1, 24
+# CHECK-NEXT:    %16:g8rc_and_g8rc_nox0 = COPY killed %4
+# CHECK-NEXT:    B %bb.4
+# CHECK:       bb.2:
+# CHECK:         %5:g8rc = ADDI8 killed %1, 24
+# CHECK-NEXT:    %6:g8rc = LI8 0
+# CHECK-NEXT:    %7:crrc = CMPLWI undef %8:gprc, 20
+# CHECK-NEXT:    %14:g8rc = COPY killed %5
+# CHECK-NEXT:    %15:g8rc = COPY killed %6
+# CHECK:       bb.3:
+# CHECK:         %10:g8rc = COPY killed %15
+# CHECK-NEXT:    %9:g8rc = COPY killed %14
+# CHECK-NEXT:    %14:g8rc = COPY killed %10
+# CHECK-NEXT:    %15:g8rc = IMPLICIT_DEF
+# CHECK-NEXT:    %16:g8rc_and_g8rc_nox0 = COPY killed %9
+# CHECK-NEXT:    BCC 68, %7, %bb.3
+# CHECK-NEXT:    B %bb.4
+# CHECK:       bb.4:
+# CHECK:         %12:g8rc_and_g8rc_nox0 = COPY killed %16
+# CHECK-NEXT:    %13:g8rc = LI8 0
+# CHECK-NEXT:    STW8 killed %13, 0, killed %12 :: (store 4)
+# CHECK-NEXT:    BLR8 implicit $lr8, implicit $rm