Detach the dependency from host_content_settings_map to extension.

chrome/browser/content_settings/host_content_settings_map.h

 // Copyright (c) 2012 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.
 
 // Maps hostnames to custom content settings.  Written on the UI thread and read
 // on any thread.  One instance per profile.
 
 #ifndef CHROME_BROWSER_CONTENT_SETTINGS_HOST_CONTENT_SETTINGS_MAP_H_
 #define CHROME_BROWSER_CONTENT_SETTINGS_HOST_CONTENT_SETTINGS_MAP_H_
 
@@ skipping 15 matching lines @@
 class ExtensionService;
 class GURL;
 class PrefService;
 
 namespace base {
 class Clock;
 class Value;
 }
 
 namespace content_settings {
+class ObservableProvider;
 class ProviderInterface;
 class PrefProvider;
 }
 
 namespace user_prefs {
 class PrefRegistrySyncable;
 }
 
 class HostContentSettingsMap
     : public content_settings::Observer,
       public base::RefCountedThreadSafe<HostContentSettingsMap> {
  public:
   enum ProviderType {
     INTERNAL_EXTENSION_PROVIDER = 0,

[Bernhard Bauer, 2014/09/11 09:39:29]

Hm... here we still have references to extensions.

I suppose if we were to do this really cleanly, we would have a skeleton HCSM
that allows registering providers but doesn't know anything about them, and one
embedder (maybe a subclass, maybe owning it) which would add thoses providers.
Not sure if that's worth it though or whether we can live with this layering
violation.

[Jun Mukai, 2014/09/11 22:58:37]

What I care most in this CL is the coding level dependency, i.e. #include of
"chrome/browser/extensions".  I have little concerns on 'EXTENSION' names.

However, I agree taht ideally this types are not great. Therefore, I uploaded a
new patchset, which removes ProviderTypes. Can you take another look?

[Bernhard Bauer, 2014/09/12 09:07:25]

Yeah, it was more of a layering violation in spirit, not in code (and like I
said, maybe we can just live with it for now).
This enum also defined the precedence of the different providers, making sure
that e.g. policy-defined content settings take precedence over user-defined,
etc. If you just keep the providers in an array, are you still going to have the
right precedences?

[Jun Mukai, 2014/09/12 20:21:05]

Hm. Considering the situation, I think it's better to keep this enum and the
structure of the code. EXTENSION_PROVIDER things are okay as the names.
We may want to remove this enum eventually, but not now.

Revert the patchset back to the original one. Do you think this is okay?

     POLICY_PROVIDER,
     CUSTOM_EXTENSION_PROVIDER,
     PREF_PROVIDER,
     DEFAULT_PROVIDER,
     NUM_PROVIDER_TYPES,
   };
 
   HostContentSettingsMap(PrefService* prefs, bool incognito);
 
-#if defined(ENABLE_EXTENSIONS)
-  // In some cases, the ExtensionService is not available at the time the
-  // HostContentSettingsMap is constructed. In these cases, we register the
-  // service once it's available.
-  void RegisterExtensionService(ExtensionService* extension_service);
-#endif
+  static void RegisterProfilePrefs(user_prefs::PrefRegistrySyncable* registry);
 
-  static void RegisterProfilePrefs(user_prefs::PrefRegistrySyncable* registry);
+  // Adds a new provider for |type|.
+  void RegisterProvider(
+      ProviderType type,
+      scoped_ptr<content_settings::ObservableProvider> provider);
 
   // Returns the default setting for a particular content type. If |provider_id|
   // is not NULL, the id of the provider which provided the default setting is
   // assigned to it.
   //
   // This may be called on any thread.
   ContentSetting GetDefaultContentSetting(ContentSettingsType content_type,
                                           std::string* provider_id) const;
 
   // Returns a single |ContentSetting| which applies to the given URLs.  Note
@@ skipping 237 matching lines @@
   // time and by RegisterExtensionService, both of which should happen
   // before any other uses of it.
   ProviderMap content_settings_providers_;
 
   ObserverList<content_settings::Observer> observers_;
 
   DISALLOW_COPY_AND_ASSIGN(HostContentSettingsMap);
 };
 
 #endif  // CHROME_BROWSER_CONTENT_SETTINGS_HOST_CONTENT_SETTINGS_MAP_H_