Issue #41 - Bring method of determining IE version up to date

src/plugin/PluginWbPassThrough.cpp

Index: src/plugin/PluginWbPassThrough.cpp
===================================================================
--- a/src/plugin/PluginWbPassThrough.cpp
+++ b/src/plugin/PluginWbPassThrough.cpp
@@ -9,7 +9,7 @@
 #include "PluginSettings.h"
 #include "PluginClass.h"
 #include "PluginSystem.h"
-
+#include "../shared/IE_version.h"
 #include "wtypes.h"
 
 EXTERN_C IMAGE_DOS_HEADER __ImageBase;
@@ -96,7 +96,7 @@
 {
   // No referer or mime type
   // BINDSTRING_XDR_ORIGIN works only for IE v8+
-  if (mimeType.IsEmpty() && domain.IsEmpty() && CPluginClient::GetInstance()->GetIEVersion() >= 8)
+  if (mimeType.IsEmpty() && domain.IsEmpty() && AdblockPlus::IE::installed_major_version() >= 8)

[sergei, 2014/07/28 11:46:27]

Just wonder, what happens if we throw an exception while the execution in some
callback of `WBPassthruSink`?

[Eric, 2014/07/29 12:42:25]

As of today, the exception will propagate up through an entry point, with
undefined behavior. It's the main reason we need to get try/catch(...) in place
throughout the code. We might already have had failures in the field from
something throwing bad_alloc.

Instead of altering all the PassthroughAPP code right now, the version functions
don't throw at all, but rather convert exceptions into default values.

[sergei, 2014/07/29 14:45:56]

I cannot say, that I'm happy having such call here
- it's not clear here that this function does not throw any exception. At the
same time similar functions does throw and the hell of calls (they are even from
different threads) in WbPassThrough looks very fragile, so to have at least
something potentially harmful here looks very dangerous.
- try{}catch(...) looses the information about the problem.
- I expect this method to be called several times, and to read the version each
time from the windows registry looks very inefficient.

We might should create a ticket for that to inject the version in the
constructor and use a constant access time value without any possibility for the
exception, at least here.

[Eric, 2014/07/29 15:17:22]

How was it clear that the previous code didn't throw an exception?

If you want to insist on solving every problem that any code modification
raises, development will grind to a halt. Incremental improvement is the name of
the game.
We can memoize the retrieval functions later. Right now, I'm trying to get
something like a small change in place.
That's one way to solve the problem, and it's not the one I'd use. If you want a
ticket here, make the task to improve the efficiency, not to pick a particular
solution.

   {
     return CFilter::contentTypeXmlHttpRequest;
   }
@@ -235,7 +235,7 @@
 #endif
 
   //Fixes the iframe back button issue
-  if (client->GetIEVersion() > 6)
+  if (AdblockPlus::IE::installed_major_version() > 6)
   {
     if ((contentType == CFilter::contentTypeImage) && (isBlocked))
     {