Deprecate support for xhr.withCredentials for synchronous requests

Source/core/xml/XMLHttpRequest.cpp

Index: Source/core/xml/XMLHttpRequest.cpp
diff --git a/Source/core/xml/XMLHttpRequest.cpp b/Source/core/xml/XMLHttpRequest.cpp
index b4505ee04b4ef1190410adbf3f560f01c3b16cba..097ae1e8e1459d66c852bf904bfaf0091c2765e7 100644
--- a/Source/core/xml/XMLHttpRequest.cpp
+++ b/Source/core/xml/XMLHttpRequest.cpp
@@ -35,6 +35,8 @@
 #include "core/fileapi/Blob.h"
 #include "core/fileapi/File.h"
 #include "core/fileapi/Stream.h"
+#include "core/frame/Settings.h"
+#include "core/frame/UseCounter.h"
 #include "core/frame/csp/ContentSecurityPolicy.h"
 #include "core/html/DOMFormData.h"
 #include "core/html/HTMLDocument.h"
@@ -42,7 +44,6 @@
 #include "core/inspector/InspectorInstrumentation.h"
 #include "core/inspector/InspectorTraceEvents.h"
 #include "core/loader/ThreadableLoader.h"
-#include "core/frame/Settings.h"
 #include "core/xml/XMLHttpRequestProgressEvent.h"
 #include "core/xml/XMLHttpRequestUpload.h"
 #include "platform/Logging.h"
@@ -486,6 +487,11 @@ void XMLHttpRequest::setWithCredentials(bool value, ExceptionState& exceptionSta
         return;
     }
 
+    // FIXME: According to XMLHttpRequest Level 2 we should throw InvalidAccessError exception here.
+    // However for time being only print warning message to warn web developers.
+    if (!m_async && executionContext()->isDocument())

[sof, 2014/06/18 08:16:06]

No need to check isDocument(), just pass in executionContext() to count() (or
countDeprecation().)

That has the added benefit of also making the non-Document (i.e., Workers) case
eventually work.

[maheshkk, 2014/06/18 15:55:08]

You are right. Once http://code.google.com/p/chromium/issues/detail?id=376039 is
done it will make much more sense. 

+        UseCounter::countDeprecation(document(), UseCounter::SyncXHRWithCredentials);

[sof, 2014/06/18 08:16:06]

With https://codereview.chromium.org/337283003 adding a global syncXHR use
counter, should we turn this into a count() for SyncXHRWithCredentials initially
& return once data for that has settled?

[maheshkk, 2014/06/18 15:55:08]

IMO, printing deprecated message will serve as early warning without waiting for
useCount. Because blink doesn't match other browsers and spec. And remove the
support once useCount is less after significant time provided to developers with
deprecated warning message. Thoughts? 

[sof, 2014/06/18 16:13:36]

I understand what you hope is the side-effect of also emitting the deprecation
message. I could be interpreting this wrongly, but I don't think we're quite
there yet here wrt deprecation and removal -- what if the usage data (and other
considerations) show that we cannot go ahead with this, wouldn't it be
counterproductive to have emitted a deprecation message for a while?

[maheshkk, 2014/06/18 23:53:07]

Agreed. We could go by only adding UseCounter now and iterate again after
obtaining the use count to deprecate. 
Mozilla, however throws exception. So web developers who might be using this
already have split in browsers behavior. I understand that it would be much
safer to deprecate after getting use count.

+
     m_includeCredentials = value;
 }