Use HTTP response headers for PNaCl caching instead of bitcode hash.

ppapi/ppapi_tests.gypi

Index: ppapi/ppapi_tests.gypi
diff --git a/ppapi/ppapi_tests.gypi b/ppapi/ppapi_tests.gypi
index 28961575b7f55d312d670f21982d876d7365b49d..3980445f72f8d0a3b730a456a1e88b0671d00416 100644
--- a/ppapi/ppapi_tests.gypi
+++ b/ppapi/ppapi_tests.gypi
@@ -152,10 +152,18 @@
         '../testing/gtest.gyp:gtest',
         '../ui/surface/surface.gyp:surface',
       ],
+      # For the nacl_http_response_headers_unittest below.
+      'include_dirs': [
+        '../ppapi',
+      ],
       'sources': [
         'proxy/run_all_unittests.cc',
 
         'host/resource_message_filter_unittest.cc',
+        # Piggy back on ppapi_unittests for a simple NaCl unittest,
+        # which must not have dependencies on anything other than stdlibs.
+        'native_client/src/trusted/plugin/nacl_http_response_headers.cc',

[dmichael (off chromium), 2013/06/14 16:21:43]

I'm not a big fan of putting the source directly in the test target. Could you
explain a bit more in the comment why this is necessary and have a TODO to fix
it when the code moves to the browser?

[jvoung (off chromium), 2013/06/14 20:38:03]

Done.  I wasn't sure how to trigger the build of the code without adding a
dependency on the ppapi/native_client/src/trusted/plugin/plugin.gyp here, so the
lightest weight thing seemed to be just to put the source here.  That's also the
reason for the comment about not having many dependencies for the implementation
of that header parser.

[dmichael (off chromium), 2013/06/14 21:09:03]

Yeah, I understand why you did it. But it's weird enough to deserve a comment.
Thanks.

+        'native_client/src/trusted/plugin/nacl_http_response_headers_unittest.cc',
         'proxy/device_enumeration_resource_helper_unittest.cc',
         'proxy/file_chooser_resource_unittest.cc',
         'proxy/flash_resource_unittest.cc',