[SVG2] Add support for the 'turn' unit in <angle>.

Source/core/svg/SVGAngleTearOff.cpp

Index: Source/core/svg/SVGAngleTearOff.cpp
diff --git a/Source/core/svg/SVGAngleTearOff.cpp b/Source/core/svg/SVGAngleTearOff.cpp
index 8a5f832ace8e6ee456c1aa82d5aebe6f3d21c665..053cf5094e946760c7b745faeecae3929500e7c6 100644
--- a/Source/core/svg/SVGAngleTearOff.cpp
+++ b/Source/core/svg/SVGAngleTearOff.cpp
@@ -109,6 +109,12 @@ void SVGAngleTearOff::setValueAsString(const String& value, ExceptionState& exce
     }
 
     target()->setValueAsString(value, exceptionState);
+
+    if (!exceptionState.hadException() && !isValidAngle()) {
+        exceptionState.throwDOMException(SyntaxError, "The value provided ('" + value + "') is invalid.");

[fs, 2014/06/10 11:13:30]

Hmm, the "damage" has been done here already (state modified). Save and
rollback?

[krit, 2014/06/10 11:45:48]

Does unknown unit type need to imply that we do not support the unit as string
in general? Internally we still do after all.

It is logical to not accept it and would be more consistent either I guess.

[Erik Dahlström (inactive), 2014/06/10 16:57:42]

If we want to behave as if the "turn" unit wasn't supported to a user of the
SVGAngle DOM interface then this is necessary. My preference is still that we
should follow the existing architecture of SVGAngle (extending it with the new
IDL constant), and then deprecating the whole interface as soon as we have a
proper replacement API (if any).
I don't fully understand what you mean.

[Erik Dahlström (inactive), 2014/06/10 16:57:42]

Done.

+        return;
+    }
+
     commitChange();
 }