Adding --disable-plugins-discovery command line switch to not load third-party plugins from common …

chrome/browser/browser_process_impl.h

Index: chrome/browser/browser_process_impl.h
===================================================================
--- chrome/browser/browser_process_impl.h	(revision 204062)
+++ chrome/browser/browser_process_impl.h	(working copy)
@@ -248,6 +248,9 @@
   // Sequenced task runner for local state related I/O tasks.
   const scoped_refptr<base::SequencedTaskRunner> local_state_task_runner_;
 
+  // Command line of the current process.
+  const CommandLine& command_line_;

[Bernhard Bauer, 2013/06/06 08:28:50]

Why do you need this as a member variable?

[seva, 2013/06/06 22:11:25]

This saves two CommandLine::ForCurrentProcess() calls and keeps things a bit
cleaner, more testable (BrowserProcessImpl constructor parameter "command_line"
is now used) and easier extensible (command_line is available in case it's
needed for other checks in the future, which seems likely), to my taste.

One CommandLine::ForCurrentProcess() call I removed from browser_process_impl.cc
line 902 and the other I would have to add there for my
HasSwitch(switches::kDisablePluginsDiscovery) check.

Though if you think this has issues (or too unrelated to the intention of the
patch) I'm totally open to reverting this part and just making another
CommandLine::ForCurrentProcess() call in browser_process_impl.cc.

[Bernhard Bauer, 2013/06/07 10:33:43]

No, I think it's fine. I was going to say that this isn't tested very
extensively anyway (and if it was, it's not much more difficult to override the
current command line temporarily), but I didn't see that it already gets a
command line as a construction parameter. In that case it definitely makes sense
to keep it around, to make sure we always use the same instance.

+
   // Ensures that the observers of plugin/print disable/enable state
   // notifications are properly added and removed.
   PrefChangeRegistrar pref_change_registrar_;