Implement origin-based deletion for password manager's auto-signin bit.

chrome/browser/password_manager/native_backend_gnome_x.cc

Index: chrome/browser/password_manager/native_backend_gnome_x.cc
diff --git a/chrome/browser/password_manager/native_backend_gnome_x.cc b/chrome/browser/password_manager/native_backend_gnome_x.cc
index 75dde044a67a6ab0417d498c4412e2a350012ddb..5b3e38ea4e400dd5fd5462ab19683f9738639790 100644
--- a/chrome/browser/password_manager/native_backend_gnome_x.cc
+++ b/chrome/browser/password_manager/native_backend_gnome_x.cc
@@ -699,14 +699,15 @@ bool NativeBackendGnome::RemoveLoginsSyncedBetween(
   return RemoveLoginsBetween(delete_begin, delete_end, SYNC_TIMESTAMP, changes);
 }
 
-bool NativeBackendGnome::DisableAutoSignInForAllLogins(
+bool NativeBackendGnome::DisableAutoSignInForOrigins(
+    const base::Callback<bool(const GURL&)>& origin_filter,
     password_manager::PasswordStoreChangeList* changes) {
   ScopedVector<PasswordForm> forms;
   if (!GetAllLogins(&forms))
     return false;
 
   for (auto& form : forms) {
-    if (!form->skip_zero_click) {
+    if (origin_filter.Run(form->origin) && !form->skip_zero_click) {

[vabr (Chromium), 2016/07/08 19:22:03]

Recalling your earlier planned to accept a "null" callback and treat as always
true here, you don't do it because nobody calls this for all origins, right?

[msramek, 2016/07/11 13:09:17]

Correct - I would leave that for another pass when we start supporting the null
callbacks on BrowsingDataRemover side.

In the meantime, I don't think there's a significant performance hit by not
supporting it, since
- the native backends always did an iteration anyway
- the SQL backend has now been optimized to one call per origin; that is still
slower than it used to be, but the operation is uncommon, and its repeated calls
will be faster (since we don't call the SQL update where unnecessary)

       form->skip_zero_click = true;
       if (!UpdateLogin(*form, changes))
         return false;