Part two of dependency injection with auto-generation support

10 files changed, 398 insertions, 237 deletions

diff --git a/build/algorithm b/build/algorithm
index 59021f5..ad53b95 100644
--- a/build/algorithm
+++ b/build/algorithm
@@ -66,7 +66,7 @@ namespace build
   // in the same or a subdirectory of dir.
   //
   void
-  search_and_match (action, target&, const dir_path& dir);
+  search_and_match (action, target&, const dir_path&);
 
   // Unless already available, match, and, if necessary, execute
   // (not yet implemented) the group in order to obtain its members
@@ -90,6 +90,15 @@ namespace build
   target_state
   execute (action, target&);
 
+  // A special version of the above that should be used for "direct"
+  // and "now" execution, that is, side-stepping the normal target-
+  // prerequisite relationship (so no dependents count is decremented)
+  // and execution order (so this function will never return postponed
+  // target state).
+  //
+  target_state
+  execute_direct (action, target&);
+
   // The default prerequisite execute implementation. It calls execute()
   // on each non-ignored (non-NULL) prerequisite target in a loop. If this
   // target is a member of a group, then it first does this to the group's
diff --git a/build/algorithm.cxx b/build/algorithm.cxx
index f4da3bd..4f30944 100644
--- a/build/algorithm.cxx
+++ b/build/algorithm.cxx
@@ -196,13 +196,13 @@ namespace build
       g.recipe (a, p.first->apply (a, g, p.second));
     }
 
-    // Note the we use execute_impl() rather than execute() here
+    // Note that we use execute_direct() rather than execute() here
     // to sidestep the dependents count logic. In this context,
     // this is by definition the first attempt to execute this
     // rule (otherwise we would have already known the members
     // list) and we really do need to execute it now.
     //
-    execute_impl (a, g);
+    execute_direct (a, g);
 
     r = g.members (a);
     assert (r.members != nullptr); // What "next step" did the group expect?
diff --git a/build/algorithm.ixx b/build/algorithm.ixx
index 3a6f183..7300848 100644
--- a/build/algorithm.ixx
+++ b/build/algorithm.ixx
@@ -83,4 +83,15 @@ namespace build
       }
     }
   }
+
+  inline target_state
+  execute_direct (action a, target& t)
+  {
+    switch (t.state)
+    {
+    case target_state::unchanged:
+    case target_state::changed: return t.state;
+    default: return execute_impl (a, t);
+    }
+  }
 }
diff --git a/build/cli/rule.cxx b/build/cli/rule.cxx
index abcbb70..6f8b648 100644
--- a/build/cli/rule.cxx
+++ b/build/cli/rule.cxx
@@ -8,6 +8,7 @@
 
 #include <build/scope>
 #include <build/target>
+#include <build/context>
 #include <build/algorithm>
 #include <build/diagnostics>
 
@@ -217,68 +218,95 @@ namespace build
       //
       cli* s (execute_prerequisites<cli> (a, t, t.h ()->mtime ()));
 
+      target_state ts;
+
       if (s == nullptr)
-        return target_state::unchanged;
+        ts = target_state::unchanged;
+      else
+      {
+        // Translate source path to relative (to working directory). This
+        // results in easier to read diagnostics.
+        //
+        path relo (relative (t.dir));
+        path rels (relative (s->path ()));
 
-      // Translate source path to relative (to working directory). This
-      // results in easier to read diagnostics.
-      //
-      path relo (relative (t.dir));
-      path rels (relative (s->path ()));
+        scope& rs (t.root_scope ());
+        const string& cli (rs["config.cli"].as<const string&> ());
 
-      scope& rs (t.root_scope ());
-      const string& cli (rs["config.cli"].as<const string&> ());
+        vector<const char*> args {cli.c_str ()};
 
-      vector<const char*> args {cli.c_str ()};
+        // See if we need to pass any --?xx-suffix options.
+        //
+        append_extension (args, *t.h (), "--hxx-suffix", "hxx");
+        append_extension (args, *t.c (), "--cxx-suffix", "cxx");
+        if (t.i () != nullptr)
+          append_extension (args, *t.i (), "--ixx-suffix", "ixx");
 
-      // See if we need to pass any --?xx-suffix options.
-      //
-      append_extension (args, *t.h (), "--hxx-suffix", "hxx");
-      append_extension (args, *t.c (), "--cxx-suffix", "cxx");
-      if (t.i () != nullptr)
-        append_extension (args, *t.i (), "--ixx-suffix", "ixx");
+        append_options (args, t, "cli.options");
 
-      append_options (args, t, "cli.options");
+        if (!relo.empty ())
+        {
+          args.push_back ("-o");
+          args.push_back (relo.string ().c_str ());
+        }
 
-      if (!relo.empty ())
-      {
-        args.push_back ("-o");
-        args.push_back (relo.string ().c_str ());
-      }
+        args.push_back (rels.string ().c_str ());
+        args.push_back (nullptr);
 
-      args.push_back (rels.string ().c_str ());
-      args.push_back (nullptr);
+        if (verb)
+          print_process (args);
+        else
+          text << "cli " << *s;
 
-      if (verb)
-        print_process (args);
-      else
-        text << "cli " << *s;
+        try
+        {
+          process pr (args.data ());
 
-      try
-      {
-        process pr (args.data ());
+          if (!pr.wait ())
+            throw failed ();
 
-        if (!pr.wait ())
-          throw failed ();
+          timestamp s (system_clock::now ());
 
-        timestamp ts (system_clock::now ());
+          // Update member timestamps.
+          //
+          t.h ()->mtime (s);
+          t.c ()->mtime (s);
+          if (t.i () != nullptr)
+            t.i ()->mtime (s);
 
-        t.h ()->mtime (ts);
-        t.c ()->mtime (ts);
-        if (t.i () != nullptr)
-          t.i ()->mtime (ts);
+          ts = target_state::changed;
+        }
+        catch (const process_error& e)
+        {
+          error << "unable to execute " << args[0] << ": " << e.what ();
 
-        return target_state::changed;
+          if (e.child ())
+            exit (1);
+
+          throw failed ();
+        }
       }
-      catch (const process_error& e)
-      {
-        error << "unable to execute " << args[0] << ": " << e.what ();
 
-        if (e.child ())
-          exit (1);
+      // Update member recipes. Without that the state update below
+      // won't stick.
+      //
+      if (!t.h ()->recipe (a))
+        t.h ()->recipe (a, &delegate);
+
+      if (!t.c ()->recipe (a))
+        t.c ()->recipe (a, &delegate);
 
-        throw failed ();
-      }
+      if (t.i () != nullptr && !t.i ()->recipe (a))
+        t.i ()->recipe (a, &delegate);
+
+      // Update member states.
+      //
+      t.h ()->state = ts;
+      t.c ()->state = ts;
+      if (t.i () != nullptr)
+        t.i ()->state = ts;
+
+      return ts;
     }
 
     target_state compile::
@@ -286,19 +314,30 @@ namespace build
     {
       cli_cxx& t (static_cast<cli_cxx&> (xt));
 
-      target_state ts (target_state::unchanged);
+      // The reverse order of update: first delete the files, then clean
+      // prerequisites. Also update timestamp in case there are operations
+      // after us that could use the information.
+      //
+      //
+      bool r (false);
+
+      if (t.i () != nullptr)
+      {
+        r = rmfile (t.i ()->path (), *t.i ()) || r;
+        t.i ()->mtime (timestamp_nonexistent);
+      }
 
-      if (t.i () != nullptr &&
-          build::perform_clean (a, *t.i ()) == target_state::changed)
-        ts = target_state::changed;
+      r = rmfile (t.c ()->path (), *t.c ()) || r;
+      t.c ()->mtime (timestamp_nonexistent);
 
-      if (build::perform_clean (a, *t.c ()) == target_state::changed)
-        ts = target_state::changed;
+      r = rmfile (t.h ()->path (), *t.h ()) || r;
+      t.h ()->mtime (timestamp_nonexistent);
 
-      if (build::perform_clean (a, *t.h ()) == target_state::changed)
-        ts = target_state::changed;
+      // Clean prerequisites.
+      //
+      target_state ts (reverse_execute_prerequisites (a, t));
 
-      return ts;
+      return r ? target_state::changed : ts;
     }
 
     target_state compile::
diff --git a/build/config/utility b/build/config/utility
index 403a91e..e59f0e5 100644
--- a/build/config/utility
+++ b/build/config/utility
@@ -53,11 +53,13 @@ namespace build
       {
         for (const name& n: val.template as<const list_value&> ())
         {
-          if (!n.simple ())
+          if (n.simple ())
+            args.push_back (n.value.c_str ());
+          else if (n.directory ())
+            args.push_back (n.dir.string ().c_str ());
+          else
             fail << "expected option instead of " << n <<
               info << "in variable " << var;
-
-          args.push_back (n.value.c_str ());
         }
       }
     }
diff --git a/build/cxx/rule b/build/cxx/rule
index 8c575fe..8b42fc2 100644
--- a/build/cxx/rule
+++ b/build/cxx/rule
@@ -30,10 +30,6 @@ namespace build
 
       static target_state
       perform_update (action, target&);
-
-    private:
-      void
-      inject_prerequisites (action, target&, const cxx&, scope&) const;
     };
 
     class link: public rule
diff --git a/build/cxx/rule.cxx b/build/cxx/rule.cxx
index 47f925d..8d91e1c 100644
--- a/build/cxx/rule.cxx
+++ b/build/cxx/rule.cxx
@@ -108,6 +108,9 @@ namespace build
       return nullptr;
     }
 
+    static void
+    inject_prerequisites (action, target&, cxx&, scope&);
+
     recipe compile::
     apply (action a, target& xt, void* v) const
     {
@@ -149,10 +152,11 @@ namespace build
         t.prerequisite_targets.push_back (&pt);
       }
 
-      // Inject additional prerequisites. For now we only do it for
-      // update and default.
+      // Inject additional prerequisites. We only do it for update
+      // since chances are we will have to update some of our
+      // prerequisites in the process (auto-generated source code).
       //
-      if (a.operation () == update_id || a.operation () == default_id)
+      if (a.operation () == update_id)
       {
         // The cached prerequisite target (sp.target) should be the
         // same as what is in t.prerequisite_targets since we used
@@ -161,8 +165,7 @@ namespace build
         prerequisite& sp (*static_cast<prerequisite*> (v));
         cxx& st (dynamic_cast<cxx&> (*sp.target));
 
-        if (st.mtime () != timestamp_nonexistent)
-          inject_prerequisites (a, t, st, sp.scope);
+        inject_prerequisites (a, t, st, sp.scope);
       }
 
       switch (a)
@@ -240,7 +243,7 @@ namespace build
             if (++i == e)
               break; // Let the compiler complain.
 
-            d = dir_path (i->value);
+            d = i->simple () ? dir_path (i->value) : i->dir;
           }
           else if (i->value.compare (0, 2, "-I") == 0)
             d = dir_path (i->value, 2, string::npos);
@@ -370,8 +373,8 @@ namespace build
       return r;
     }
 
-    void compile::
-    inject_prerequisites (action a, target& t, const cxx& s, scope& ds) const
+    static void
+    inject_prerequisites (action a, target& t, cxx& s, scope& ds)
     {
       tracer trace ("cxx::compile::inject_prerequisites");
 
@@ -408,212 +411,298 @@ namespace build
       args.push_back ("-MQ"); // Quoted target name.
       args.push_back ("*");   // Old versions can't handle empty target name.
 
-      // We are using absolute source file path in order to get
-      // absolute paths in the result. Any relative paths in the
-      // result are non-existent generated headers.
+      // We are using absolute source file path in order to get absolute
+      // paths in the result. Any relative paths in the result are non-
+      // existent, potentially auto-generated headers.
       //
       // @@ We will also have to use absolute -I paths to guarantee
-      // that.
+      // that. Or just detect relative paths and error out?
       //
       args.push_back (s.path ().string ().c_str ());
-
       args.push_back (nullptr);
 
-      if (verb >= 2)
-        print_process (args);
-
       level5 ([&]{trace << "target: " << t;});
 
-      try
-      {
-        process pr (args.data (), false, false, true);
-        ifdstream is (pr.in_ofd);
-        prefix_map pm; // Build it lazily.
+      // Build the prefix map lazily only if we have non-existent files.
+      // Also reuse it over restarts since it doesn't change.
+      //
+      prefix_map pm;
+
+      // If any prerequisites that we have extracted changed, then we
+      // have to redo the whole thing. The reason for this is auto-
+      // generated headers: the updated header may now include a yet-
+      // non-existent header. Unless we discover this and generate it
+      // (which, BTW, will trigger another restart since that header,
+      // in turn, can also include auto-generated headers), we will
+      // end up with an error during compilation proper.
+      //
+      // One complication with this restart logic is that we will see
+      // a "prefix" of prerequsites that we have already processed
+      // (i.e., they are already in our prerequsite_targets list) and
+      // we don't want to keep redoing this over and over again. One
+      // thing to note, however, is that the prefix that we have seen
+      // on the previous run must appear exactly the same in the
+      // subsequent run. The reason for this is that none of the files
+      // that it can possibly be based on have changed and thus it
+      // should be exactly the same. To put it another way, the
+      // presence or absence of a file in the dependency output can
+      // only depend on the previous files (assuming the compiler
+      // outputs them as it encounters them and it is hard to think
+      // of a reason why would someone do otherwise). And we have
+      // already made sure that all those files are up to date. And
+      // here is the way we are going to exploit this: we are going
+      // to keep track of how many prerequsites we have processed so
+      // far and on restart skip right to the next one.
+      //
+      // Also, before we do all that, make sure the source file itself
+      // if up to date.
+      //
+      execute_direct (a, s);
 
-        for (bool first (true), second (true); !is.eof (); )
-        {
-          string l;
-          getline (is, l);
+      size_t skip_count (0);
+      for (bool restart (true); restart; )
+      {
+        restart = false;
 
-          if (is.fail () && !is.eof ())
-            fail << "io error while parsing g++ -M output";
+        if (verb >= 2)
+          print_process (args);
 
-          size_t pos (0);
+        try
+        {
+          process pr (args.data (), false, false, true);
+          ifdstream is (pr.in_ofd);
 
-          if (first)
+          size_t skip (skip_count);
+          for (bool first (true), second (true); !(restart || is.eof ()); )
           {
-            // Empty output should mean the wait() call below will return
-            // false.
-            //
-            if (l.empty ())
-              break;
+            string l;
+            getline (is, l);
 
-            assert (l[0] == '*' && l[1] == ':' && l[2] == ' ');
+            if (is.fail () && !is.eof ())
+              fail << "io error while parsing g++ -M output";
 
-            first = false;
+            size_t pos (0);
 
-            // While normally we would have the source file on the
-            // first line, if too long, it will be move to the next
-            // line and all we will have on this line is "*: \".
-            //
-            if (l.size () == 4 && l[3] == '\\')
-              continue;
-            else
-              pos = 3; // Skip "*: ".
+            if (first)
+            {
+              // Empty output should mean the wait() call below will return
+              // false.
+              //
+              if (l.empty ())
+                break;
 
-            // Fall through to the 'second' block.
-          }
+              assert (l[0] == '*' && l[1] == ':' && l[2] == ' ');
 
-          if (second)
-          {
-            second = false;
-            next (l, pos); // Skip the source file.
-          }
+              first = false;
 
-          auto g (
-            make_exception_guard (
-              [](const target& s)
-              {
-                info << "while extracting dependencies from " << s;
-              },
-              s));
+              // While normally we would have the source file on the
+              // first line, if too long, it will be moved to the next
+              // line and all we will have on this line is "*: \".
+              //
+              if (l.size () == 4 && l[3] == '\\')
+                continue;
+              else
+                pos = 3; // Skip "*: ".
 
-          while (pos != l.size ())
-          {
-            path f (next (l, pos));
-            f.normalize ();
+              // Fall through to the 'second' block.
+            }
 
-            if (!f.absolute ())
+            if (second)
             {
-              // This is probably as often an error as an auto-generated
-              // file, so trace at level 3.
-              //
-              level3 ([&]{trace << "non-existent header '" << f << "'";});
+              second = false;
+              next (l, pos); // Skip the source file.
+            }
 
-              // If we already did it and build_prefix_map() returned empty,
-              // then we would have failed below.
-              //
-              if (pm.empty ())
-                pm = build_prefix_map (t);
+            // If things go wrong (and they often do in this area), give
+            // the user a bit extra context.
+            //
+            auto g (
+              make_exception_guard (
+                [](target& s)
+                {
+                  info << "while extracting dependencies from " << s;
+                },
+                s));
 
-              // First try the whole file. Then just the directory.
-              //
-              // @@ Has to be a separate map since the prefix can be
-              //    the same as the file name.
-              //
-              // auto i (pm.find (f));
+            while (pos != l.size ())
+            {
+              string fs (next (l, pos));
 
-              // Find the most qualified prefix of which we are a
-              // sub-path.
+              // Skip until where we left off.
               //
-              auto i (pm.end ());
+              if (skip != 0)
+              {
+                skip--;
+                continue;
+              }
 
-              if (!pm.empty ())
+              path f (move (fs));
+              f.normalize ();
+
+              if (!f.absolute ())
               {
-                const dir_path& d (f.directory ());
-                i = pm.upper_bound (d);
-                --i; // Greatest less than.
+                // This is probably as often an error as an auto-generated
+                // file, so trace at level 3.
+                //
+                level3 ([&]{trace << "non-existent header '" << f << "'";});
+
+                // If we already did it and build_prefix_map() returned empty,
+                // then we would have failed below.
+                //
+                if (pm.empty ())
+                  pm = build_prefix_map (t);
+
+                // First try the whole file. Then just the directory.
+                //
+                // @@ Has to be a separate map since the prefix can be
+                //    the same as the file name.
+                //
+                // auto i (pm.find (f));
+
+                // Find the most qualified prefix of which we are a
+                // sub-path.
+                //
+                auto i (pm.end ());
+
+                if (!pm.empty ())
+                {
+                  const dir_path& d (f.directory ());
+                  i = pm.upper_bound (d);
+                  --i; // Greatest less than.
 
-                if (!d.sub (i->first)) // We might still not be a sub.
-                  i = pm.end ();
-              }
+                  if (!d.sub (i->first)) // We might still not be a sub.
+                    i = pm.end ();
+                }
 
-              if (i == pm.end ())
-                fail << "unable to map presumably auto-generated header '"
-                     << f << "' to a project";
+                if (i == pm.end ())
+                  fail << "unable to map presumably auto-generated header '"
+                       << f << "' to a project";
 
-              f = i->second / f;
-            }
+                f = i->second / f;
+              }
 
-            level5 ([&]{trace << "injecting " << f;});
+              level5 ([&]{trace << "injecting " << f;});
 
-            // Split the name into its directory part, the name part, and
-            // extension. Here we can assume the name part is a valid
-            // filesystem name.
-            //
-            // Note that if the file has no extension, we record an empty
-            // extension rather than NULL (which would signify that the
-            // default extension should be added).
-            //
-            dir_path d (f.directory ());
-            string n (f.leaf ().base ().string ());
-            const char* es (f.extension ());
-            const string* e (&extension_pool.find (es != nullptr ? es : ""));
+              // Split the name into its directory part, the name part, and
+              // extension. Here we can assume the name part is a valid
+              // filesystem name.
+              //
+              // Note that if the file has no extension, we record an empty
+              // extension rather than NULL (which would signify that the
+              // default extension should be added).
+              //
+              dir_path d (f.directory ());
+              string n (f.leaf ().base ().string ());
+              const char* es (f.extension ());
+              const string* e (&extension_pool.find (es != nullptr ? es : ""));
 
-            // Find or insert target.
-            //
-            // If there is no extension (e.g., standard C++ headers),
-            // then assume it is a header. Otherwise, let the standard
-            // mechanism derive the type from the extension. @@ TODO.
-            //
+              // Determine the target type.
+              //
+              const target_type* tt (nullptr);
+
+              // See if this directory is part of any project out_root
+              // hierarchy. Note that this will miss all the headers
+              // that come from src_root (so they will be treated as
+              // generic C headers below). Generally, we don't have
+              // the ability to determine that some file belongs to
+              // src_root of some project. But that's not a problem
+              // for our purposes: it is only important for us to
+              // accurately determine target types for headers that
+              // could be auto-generated.
+              //
+              if (scope* r = scopes.find (d).root_scope ())
+              {
+                // Get cached (or build) a map of the extensions for the
+                // C/C++ files this project is using.
+                //
+                const ext_map& m (build_ext_map (*r));
+
+                auto i (m.find (e));
+                if (i != m.end ())
+                  tt = i->second;
+              }
 
-            // Determine the target type.
-            //
-            const target_type* tt (nullptr);
-
-            // See if this directory is part of any project out_root
-            // hierarchy. Note that this will miss all the headers
-            // that come from src_root (so they will be treated as
-            // generic C headers below). Generally, we don't have
-            // the ability to determine that some file belongs to
-            // src_root of some project. But that's not a problem
-            // for our purposes: it is only important for us to
-            // accurately determine target types for headers that
-            // could be auto-generated.
-            //
-            if (scope* r = scopes.find (d).root_scope ())
-            {
-              // Get cahed (or build) a map of the extensions for the
-              // C/C++ files this project is using.
+              // If it is outside any project, or the project doesn't have
+              // such an extension, assume it is a plain old C header.
               //
-              const ext_map& m (build_ext_map (*r));
+              if (tt == nullptr)
+                tt = &h::static_type;
 
-              auto i (m.find (e));
-              if (i != m.end ())
-                tt = i->second;
-            }
+              // Find or insert target.
+              //
+              path_target& pt (
+                static_cast<path_target&> (search (*tt, d, n, e, &ds)));
 
-            // If it is outside any project, or the project doesn't have
-            // such an extension, assume it is a plain old C header.
-            //
-            if (tt == nullptr)
-              tt = &h::static_type;
+              // Assign path.
+              //
+              if (pt.path ().empty ())
+                pt.path (move (f));
 
-            path_target& pt (
-              static_cast<path_target&> (search (*tt, d, n, e, &ds)));
+              // Match to a rule.
+              //
+              build::match (a, pt);
 
-            // Assign path.
-            //
-            if (pt.path ().empty ())
-              pt.path (move (f));
+              // Update it.
+              //
+              // There would normally be a lot of headers for every source
+              // file (think all the system headers) and this can get
+              // expensive. At the same time, most of these headers are
+              // existing files that we will never be updating (again,
+              // system headers, for example) and the rule that will match
+              // them is fallback path_rule. So we are going to do a little
+              // fast-path optimization by detecting this common case.
+              //
+              recipe_function* const* recipe (
+                pt.recipe (a).target<recipe_function*> ());
 
-            // Match to a rule.
-            //
-            build::match (a, pt);
+              if (recipe == nullptr || *recipe != &path_rule::perform_update)
+              {
+                // We only want to restart if our call to execute() actually
+                // caused an update. In particular, the target could already
+                // have been in target_state::changed because of a dependency
+                // extraction run for some other source file.
+                //
+                target_state os (pt.state);
+                execute_direct (a, pt);
+
+                if (pt.state != os && pt.state != target_state::unchanged)
+                {
+                  level5 ([&]{trace << "updated " << pt << ", restarting";});
+                  restart = true;
+                }
+              }
 
-            // Add to our prerequisite target list.
-            //
-            t.prerequisite_targets.push_back (&pt);
+              // Add to our prerequisite target list.
+              //
+              t.prerequisite_targets.push_back (&pt);
+              skip_count++;
+            }
           }
-        }
 
-        // We assume the child process issued some diagnostics.
-        //
-        if (!pr.wait ())
-          throw failed ();
-      }
-      catch (const process_error& e)
-      {
-        error << "unable to execute " << args[0] << ": " << e.what ();
+          // We may not have read all the output (e.g., due to a restart),
+          // so close the file descriptor before waiting to avoid blocking
+          // the other end.
+          //
+          is.close ();
 
-        // In a multi-threaded program that fork()'ed but did not exec(),
-        // it is unwise to try to do any kind of cleanup (like unwinding
-        // the stack and running destructors).
-        //
-        if (e.child ())
-          exit (1);
+          // We assume the child process issued some diagnostics.
+          //
+          if (!pr.wait ())
+            throw failed ();
+        }
+        catch (const process_error& e)
+        {
+          error << "unable to execute " << args[0] << ": " << e.what ();
 
-        throw failed ();
+          // In a multi-threaded program that fork()'ed but did not exec(),
+          // it is unwise to try to do any kind of cleanup (like unwinding
+          // the stack and running destructors).
+          //
+          if (e.child ())
+            exit (1);
+
+          throw failed ();
+        }
       }
     }
 
diff --git a/build/rule.cxx b/build/rule.cxx
index baed3ba..34aae9d 100644
--- a/build/rule.cxx
+++ b/build/rule.cxx
@@ -39,7 +39,7 @@ namespace build
     // are not doing anything for this action so not checking if the file
     // exists seems harmless. What about, say, configure_update? Again,
     // whether we match or not, there is nothing to be done for this
-    // action. And who knows, maybe the file doesn't exist during
+    // action. And, who knows, maybe the file doesn't exist during
     // configure_update but will magically appear during perform_update.
     // So the overall guideline seems to be this: if we don't do anything
     // for the action (other than performing it on the prerequisites),
@@ -160,10 +160,6 @@ namespace build
   recipe fsdir_rule::
   apply (action a, target& t, void*) const
   {
-    // Inject dependency on the parent directory.
-    //
-    inject_parent_fsdir (a, t);
-
     switch (a.operation ())
     {
     // For default, we don't do anything other than letting our
@@ -171,6 +167,11 @@ namespace build
     //
     case default_id:
     case update_id:
+      // Inject dependency on the parent directory. Note that we
+      // don't do it for clean since we shouldn't be removing it.
+      //
+      inject_parent_fsdir (a, t);
+
       search_and_match (a, t);
       break;
     // For clean, ignore prerequisites that are not in the same or a
diff --git a/build/target b/build/target
index 1de521e..8161438 100644
--- a/build/target
+++ b/build/target
@@ -35,6 +35,9 @@ namespace build
   //
   enum class target_state {unknown, postponed, unchanged, changed, failed};
 
+  std::ostream&
+  operator<< (std::ostream&, target_state);
+
   // Recipe.
   //
   // The returned target state should be changed, unchanged, or
diff --git a/build/target.cxx b/build/target.cxx
index 3b030ee..6cd1fc9 100644
--- a/build/target.cxx
+++ b/build/target.cxx
@@ -17,6 +17,17 @@ using namespace std;
 
 namespace build
 {
+  // target_state
+  //
+  static const char* target_state_[] = {
+    "unknown", "postponed", "unchanged", "changed", "failed"};
+
+  ostream&
+  operator<< (ostream& os, target_state ts)
+  {
+    return os << target_state_[static_cast<size_t> (ts)];
+  }
+
   // recipe
   //
   const recipe empty_recipe;