Build self-consistent graphs for dupe edges with multiple outputs.

Fixes #867, both the crashes and "[stuck]" issues.

The problem was that a duplicate edge would modify the in_edge of the
outputs of the new build rule, but the edge corresponding to the old
build rule would still think that the in_edge points to itself.
`old_edge->outputs_[0]->in_edge()` would not return `old_edge`, which
confused the scan logic.

As fix, let `State::AddOut()` reject changing in_edge if it's already
set.  This changes behavior in a minor way: Previously, if there were
multiple edges for a single output, the last edge would be kept.  Now,
the first edge is kept.  This only had mostly-well-defined semantics if
all duplicate edges are the same (which is the only case I've seen in
practice), and for that case the behavior doesn't change.

For testing, add a VerifyGraph() function and call that every time any
test graph is parsed.  That's a bit more code than just copying the test
cases from the bug into build_test.cc, but it also yields better test
coverage overall.

1 files changed, 2 insertions, 1 deletions

diff --git a/src/state.cc b/src/state.cc
index 89f90c0..18c0c8c 100644
--- a/src/state.cc
+++ b/src/state.cc
@@ -141,13 +141,14 @@ void State::AddIn(Edge* edge, StringPiece path, unsigned int slash_bits) {
 
 void State::AddOut(Edge* edge, StringPiece path, unsigned int slash_bits) {
   Node* node = GetNode(path, slash_bits);
-  edge->outputs_.push_back(node);
   if (node->in_edge()) {
     Warning("multiple rules generate %s. "
             "builds involving this target will not be correct; "
             "continuing anyway",
             path.AsString().c_str());
+    return;
   }
+  edge->outputs_.push_back(node);
   node->set_in_edge(edge);
 }