On 7/2/14, 2:48 AM, Alan Bateman wrote:
On 02/07/2014 08:20, Alan Bateman wrote:
:
I think the only case that need detailed examination now is whether it
is possible for SharedSecrets.getJavaAWTFontAccess to return null. I'm
thinking about the case where 2+ threads use Bidi for the first time,
or say something else initializes TextAttribute at the same time that
another thread uses Bidi for the first time.
On second look then I think this is okay because the static initializer
must happen-before the use of the class.
Exactly.
So I think the only question
now is the test case and understanding why that needs to be updated.
The test case ensures that even BidiBase class was loaded before
TextAttribte/NumericShaper classes, it should work correctly. Version 1
of this fix (webrev.1) actually fails with this test case.
Naoto
