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, 8 insertions, 0 deletions

diff --git a/src/manifest_parser.cc b/src/manifest_parser.cc
index 044b259..f0457dd 100644
--- a/src/manifest_parser.cc
+++ b/src/manifest_parser.cc
@@ -340,6 +340,14 @@ bool ManifestParser::ParseEdge(string* err) {
   edge->implicit_deps_ = implicit;
   edge->order_only_deps_ = order_only;
 
+  if (edge->outputs_.empty()) {
+    // All outputs of the edge are already created by other edges. Don't add
+    // this edge.
+    state_->edges_.pop_back();
+    delete edge;
+    return true;
+  }
+
   // Multiple outputs aren't (yet?) supported with depslog.
   string deps_type = edge->GetBinding("deps");
   if (!deps_type.empty() && edge->outputs_.size() > 1) {