Hello Alexander,

 the test looks fine to me.

 Regarding explicit activation of xrender pipeline, there are three tests
 which does the same thing:

test/sun/java2d/XRenderBlitsTest.java
test/sun/java2d/pipe/InterpolationQualityTest.java
test/sun/java2d/AcceleratedXORModeTest.java

 Probably, they have to be modified in order to avoid false alarms.

Thanks,
Andrew

On 10/29/2013 6:37 PM, alexander stepanov wrote:
Please see the updated webrev:
http://cr.openjdk.java.net/~bae/8024767/webrev.01/

Regards,
Alexander

On 29.10.2013 16:48, alexander stepanov wrote:
Hello Phil,

OK, thanks, I'll remove the redundant options.

Regards,
Alexander

On 29.10.2013 2:03, Phil Race wrote:
Sorry for the delay in replying to this but I don't think you should specify these options
* @run main/othervm -Dsun.java2d.xrender=True TextRenderingTest

1 - othervm is frowned on these days as it slows down testing
2. If the reason you are using it is to pass in the VM option, then its not right here. You are forcing the (platform specific) pipeline to be on against the better judgement of the pipeline. The test is only valid if the pipeline
decides its valid.

-phil.

On 9/13/2013 5:58 AM, Andrew Brygin wrote:
Hi Alexander,

 the test looks fine to me.

Thanks,
Andrew

On 9/13/2013 1:20 PM, alexander stepanov wrote:
Hello,

could you please review the fix for the following [TEST] bug:
https://bugs.openjdk.java.net/browse/JDK-8024767
(sorry, the bug info isn't available at bugs.sun.com yet).

The webrev corresponding:
http://cr.openjdk.java.net/~bae/8024767/webrev.00/

Just a new test added; no other code affected.
The patch contains a simple automated test (check if source offset for text rendering is handled correctly).

Thanks.

Regards,
Alexander





Reply via email to