Do not use bundles for the CDM on Mac.

webkit/media/webkit_media.gypi

Index: webkit/media/webkit_media.gypi
diff --git a/webkit/media/webkit_media.gypi b/webkit/media/webkit_media.gypi
index faf92579d4d178b29ca714d005049b6c62eede10..ce3ab1670c189c85cd57f6edb09b9bdc0cd277b6 100644
--- a/webkit/media/webkit_media.gypi
+++ b/webkit/media/webkit_media.gypi
@@ -233,7 +233,6 @@
         }],
         ['OS == "mac"', {
           'type': 'loadable_module',
-          'mac_bundle': 1,
           'product_extension': 'plugin',
           'xcode_settings': {
             'OTHER_LDFLAGS': [
@@ -241,16 +240,9 @@
               '-Wl,-exported_symbol,_PPP_GetInterface',
               '-Wl,-exported_symbol,_PPP_InitializeModule',
               '-Wl,-exported_symbol,_PPP_ShutdownModule'
-            ]},
-          'copies': [
-            {
-              'destination': '<(PRODUCT_DIR)/clearkeycdmadapter.plugin/Contents/MacOS/',
-              'files': [
-                '<(PRODUCT_DIR)/libclearkeycdm.dylib',
-                '<(PRODUCT_DIR)/ffmpegsumo.so'
-              ]
-            }
-          ]
+            ],
+            'DYLIB_INSTALL_NAME_BASE': '@loader_path',

[xhwang, 2013/05/02 19:39:17]

does our content_browsertests still work with this change?

[ddorwin, 2013/05/02 20:35:57]

Works on my box. Trybots are running.
I think it's fine - since the two dylibs (no bundles) are side-by-side,
everything works fine. If this was still a bundle, we'd have to fix the CDM's
loader_path to account for the real binary in the bundle being 3 levels deep.

+          },
         }],
       ],
     }