Make most browser accelerators not repeat.

chrome/browser/ui/views/frame/browser_view.h

Index: chrome/browser/ui/views/frame/browser_view.h
diff --git a/chrome/browser/ui/views/frame/browser_view.h b/chrome/browser/ui/views/frame/browser_view.h
index 137ed8cf335f1b18de6262472bc984ce4a49660a..597fb94cba3a162c046e46a62d2b651c5e4c342f 100644
--- a/chrome/browser/ui/views/frame/browser_view.h
+++ b/chrome/browser/ui/views/frame/browser_view.h
@@ -657,9 +657,14 @@ class BrowserView : public BrowserWindow,
   // The Status information bubble that appears at the bottom of the window.
   std::unique_ptr<StatusBubbleViews> status_bubble_;
 
-  // A mapping between accelerators and commands.
+  // A mapping between accelerators and command IDs.
   std::map<ui::Accelerator, int> accelerator_table_;
 
+  // IDs of commands that should be executed in response to repeated
+  // accelerators (i.e. generated by holding keys down). Repeated accelerators
+  // are ignored otherwise.
+  std::set<int> repeatable_command_ids_;

[sky, 2016/09/01 19:31:06]

How often do we trigger accelerators, especially ones with a repeat? My
suspicion is it isn't worth caching this. WDYT?

[Daniel Erat, 2016/09/01 19:47:31]

the local copy was initially due to mirroring the accelerator_table_ code. it
feels weird to initialize a new set every time we see a repeatable accelerator
(especially since these come in quickly if they're being repeated), but at the
same time, it's small and we're probably doing much bigger allocations for other
reasons.

i'd prefer to leave this, but i'll remove it if you really want me to. :-)

but one other option would be instead have an chrome::IsCommandRepeatable()
function be exposed by accelerator_table.h. it could just iterate through the
few entries in its local array every time it's called.

+
   // True if we have already been initialized.
   bool initialized_;