Make setting the user agent work with the zygote on linux.

webkit/glue/webkit_glue.h

 // Copyright (c) 2011 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #ifndef WEBKIT_GLUE_WEBKIT_GLUE_H_
 #define WEBKIT_GLUE_WEBKIT_GLUE_H_
 
 #include "base/basictypes.h"
 
 #if defined(OS_WIN)
@@ skipping 73 matching lines @@
 string16 DumpFrameScrollPosition(WebKit::WebFrame* web_frame, bool recursive);
 
 // Returns a dump of the given history state suitable for implementing the
 // dumpBackForwardList command of the layoutTestController.
 string16 DumpHistoryState(const std::string& history_state, int indent,
                           bool is_current);
 
 // Sets the user agent.  Pass true for overriding if this is a custom
 // user agent instead of the default one (in order to turn off any browser
 // sniffing workarounds). This must be called before GetUserAgent() can
-// be called.
+// be called, and must not be called after GetUserAgent() with a different
+// value (use ResetUserAgent() for that, if necessary).
 void SetUserAgent(const std::string& user_agent, bool overriding);
 
+// Like SetUserAgent(), but can be used to forcibly change the user agent
+// (disabling the checks that ensure that GetUserAgent() hasn't already
+// been called).

[jam, 2011/10/15 01:32:04]

it seems a little redundant to provide two methods that do the same thing with
the only difference being that one has an extra check. why not just take out the
check altogether? if another caller runs into it, they'll just use the second
version

[Dirk Pranke, 2011/10/15 02:14:07]

Well, Reset() should really only be called in extenuating circumstances (like
the zygote case). We could decide that we don't want to be this paranoid, but I
worry that we might introduce weird bugs by doing so. 

On the other hand, it would probably be better to get rid of this reset call and
the setuseragent call in content client, and just have the one set() call done
at the appropriate times and in the appropriate places.

WDYT?

As an aside, I would like to add some tests to make sure this doesn't break in
the future, but I'm not sure what the best way to test this would be. Would a
browser_test be the right way to test it (since we need to launch a renderer and
ensure it uses the right string)? Any idea how I could figure out what user
agent was actually used?

[jam, 2011/10/17 03:43:23]

sure, if you can find one place that calls this and which works in all processes
that would be better
browser_tests run in the context of the browser process, so I'm not sure how you
will verify what's happening in the renderer?

+void ResetUserAgent(const std::string& user_agent, bool overriding);
+
 // Returns the user agent to use for the given URL. SetUserAgent() must
 // be called prior to calling this function.
 const std::string& GetUserAgent(const GURL& url);
 
 // Creates serialized state for the specified URL. This is a variant of
 // HistoryItemToString (in glue_serialize) that is used during session restore
 // if the saved state is empty.
 std::string CreateHistoryStateForURL(const GURL& url);
 
 // Removes any form data state from the history state string |content_state|.
@@ skipping 127 matching lines @@
 bool GetFontTable(int fd, uint32_t table, uint8_t* output,
                   size_t* output_length);
 #endif
 
 // ---- END FUNCTIONS IMPLEMENTED BY EMBEDDER ---------------------------------
 
 
 } // namespace webkit_glue
 
 #endif  // WEBKIT_GLUE_WEBKIT_GLUE_H_