Why not renaming the d-bus service files dir to dbus-installed (like you did in gabble)?

Some tests depend on the examples, not on the source code. I think those tests should be installed too.

(In reply to comment #1)
> Why not renaming the d-bus service files dir to dbus-installed (like you did in
> gabble)?

done

> Some tests depend on the examples, not on the source code. I think those tests
> should be installed too.

done.

Looks fine to me.

Actually:


+  gchar *command_installed[] = {
+      g_strdup_printf ("%s/%s",
+          libexec,

libexec is potentially NULL. I think printf-style functions on some platforms are not NULL-safe.

-  if (!g_spawn_async (NULL, command, NULL, 0, NULL, NULL, NULL, &error))
+  if (abs_top_builddir != NULL)
+    {
+      if (!g_spawn_async (NULL, command_uninstalled, NULL, 0, NULL, NULL, NULL,
+          &error))
+        {
+          g_error ("g_spawn_async: %s", error->message);
+        }
+    }
+  else
     {
-      g_error ("g_spawn_async: %s", error->message);
+      if (!g_spawn_async (NULL, command_installed, NULL, 0, NULL, NULL, NULL,
+          &error))
+        {
+          g_error ("g_spawn_async: %s", error->message);
+        }
     }

This code duplication is a bit ugly. You could rework it to make it more readable, for instance by having command[0] assigned based on abs_top_builddir.

(In reply to comment #4)
> Actually:
> 
> 
> +  gchar *command_installed[] = {
> +      g_strdup_printf ("%s/%s",
> +          libexec,
> 
> libexec is potentially NULL. I think printf-style functions on some platforms
> are not NULL-safe.
(...)
> 
> This code duplication is a bit ugly. You could rework it to make it more
> readable, for instance by having command[0] assigned based on abs_top_builddir.


fixed.

Looks good. If make check keeps working in every case, please merge with master.

and merged to git master: 45b8e51..d2e6b77