From: Chris Lattner <sabre@nondot.org>
Date: Fri, 4 Oct 2002 14:50:20 +0000 (+0000)
Subject: Handle post dominance correctly in the case where blocks do not have a path to
X-Git-Url: http://plrg.eecs.uci.edu/git/?a=commitdiff_plain;h=7d821db9586dfcd8301957c768c68c58b1ead34a;p=oota-llvm.git

Handle post dominance correctly in the case where blocks do not have a path to
the exit node.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@4038 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Analysis/PostDominators.cpp b/lib/Analysis/PostDominators.cpp
index 8b66ff6cad1..f027949793f 100644
--- a/lib/Analysis/PostDominators.cpp
+++ b/lib/Analysis/PostDominators.cpp
@@ -59,6 +59,18 @@ bool PostDominatorSet::runOnFunction(Function &F) {
 	  if (PredSet.size())
 	    set_intersect(WorkingSet, PredSet);
 	}
+      } else if (BB != Root) {
+        // If this isn't the root basic block and it has no successors, it must
+        // be an non-returning block.  Fib a bit by saying that the root node
+        // postdominates this unreachable node.  This isn't exactly true,
+        // because there is no path from this node to the root node, but it is
+        // sorta true because any paths to the exit node would have to go
+        // through this node.
+        //
+        // This allows for postdominator properties to be built for code that
+        // doesn't return in a reasonable manner.
+        //
+        WorkingSet = Doms[Root];
       }
 	
       WorkingSet.insert(BB);           // A block always dominates itself