[OpenJDK 2D-Dev] Review Request: JDK-8144735 [hidpi] javax/swing/JWindow/ShapedAndTranslucentWindows/TranslucentPerPixelTranslucentGradient.java fails

Tue Aug 30 06:17:29 UTC 2016

Reminder

From: Prasanta Sadhukhan 
Sent: Friday, August 26, 2016 12:08 PM
To: Prem Balakrishnan; 2d-dev at openjdk.java.net
Subject: Re: Review Request: JDK-8144735 [hidpi] javax/swing/JWindow/ShapedAndTranslucentWindows/TranslucentPerPixelTranslucentGradient.java fails

Ok. so long you are addressing the opengl issue albeit via different bugid, it's ok with me. +1.

Regards
Prasanta

On 8/26/2016 11:56 AM, Prem Balakrishnan wrote:

Hi Prasanta,

Thankyou for the review.

The code compiles , Actual scenario is as below:

public class Temp {

    AccelSurface as = new AccelSurface() {};

    SurfaceData sd = (SurfaceData)as;

}

class SurfaceData {}

interface Surface {}

interface AccelSurface extends Surface{}

class D3DSurfaceData extends SurfaceData implements AccelSurface{}

class OGLSurfaceData extends SurfaceData implements AccelSurface{}

-------------------

In suggested fix, the SurfaceData is always of type D3DSurfaceData.

And hence getDefaultScaleX/Y holds good. OpenGL rendering is handled in a different flow.

HYPERLINK "http://cr.openjdk.java.net/%7Epkbalakr/8144735/webrev.02/"http://cr.openjdk.java.net/~pkbalakr/8144735/webrev.02/

Suggested fix also resolves the following failures on hidpi windows 8

javax/swing/JWindow/ShapedAndTranslucentWindows/PerPixelTranslucent.java 
javax/swing/JWindow/ShapedAndTranslucentWindows/PerPixelTranslucentGradient.java 
javax/swing/JWindow/ShapedAndTranslucentWindows/SetShapeAndClickSwing.java 
javax/swing/JWindow/ShapedAndTranslucentWindows/ShapedPerPixelTranslucentGradient.java 
javax/swing/JWindow/ShapedAndTranslucentWindows/ShapedTranslucentPerPixelTranslucentGradient.java 
javax/swing/JWindow/ShapedAndTranslucentWindows/TranslucentPerPixelTranslucentGradient.java

----------------------
OpenGL Rendering issue will be addressed in HYPERLINK "https://bugs.openjdk.java.net/browse/JDK-8164811"JDK-8164811 

Regards,

Prem

From: Prasanta Sadhukhan 
Sent: Monday, August 22, 2016 12:32 PM
To: Prem Balakrishnan; Rajeev Chamyal; HYPERLINK "mailto:awt-dev at openjdk.java.net"awt-dev at openjdk.java.net; HYPERLINK "mailto:2d-dev at openjdk.java.net"2d-dev at openjdk.java.net
Subject: Re: Review Request: JDK-8144735 [hidpi] javax/swing/JWindow/ShapedAndTranslucentWindows/TranslucentPerPixelTranslucentGradient.java fails

I wonder how it compiles?
I tried a small program
------------------
public class Test {
        AccelSurface b = new AccelSurface();
        SurfaceData c = (SurfaceData)b;
}

class SurfaceData extends Surface {}
class Surface {}
class AccelSurface extends Surface {}
---------
and it fails to compile 
Test.java:3: error: incompatible types: AccelSurface cannot be converted to SurfaceData
    SurfaceData c = (SurfaceData)b;

ANyways, did you check with opengl pipeline? It seems getDefaultScaleX/Y is not present in OGLSurfaceData which will result in falling back to SurfaceData in which case the scale will be 1.

Regards
Prasanta

On 8/18/2016 3:12 PM, Prem Balakrishnan wrote:

Added "2d-dev" team for review

Regards,

Prem

From: Alexandr Scherbatiy 
Sent: Thursday, August 18, 2016 2:57 PM
To: Prem Balakrishnan; Rajeev Chamyal; HYPERLINK "mailto:awt-dev at openjdk.java.net"awt-dev at openjdk.java.net; Sergey Bylokhov
Subject: Re: Review Request: JDK-8144735 [hidpi] javax/swing/JWindow/ShapedAndTranslucentWindows/TranslucentPerPixelTranslucentGradient.java fails

Could you also send the review to the 2d-dev alias?

Thanks,
Alexandr.

On 8/18/2016 9:59 AM, Prem Balakrishnan wrote:

Hi Alexandr,

AccelSurface is implemented by *ONLY* D3DSurfaceData and OGLSurfaceData classes,

Both of these classes extend SurfaceData as well.

Hence, casting of 'as' variable which is of type AccelSurface object to SurfaceData is always VALID.

Regards,
Prem

From: Alexandr Scherbatiy 
Sent: Wednesday, August 17, 2016 4:42 PM
To: Prem Balakrishnan; Rajeev Chamyal; HYPERLINK "mailto:awt-dev at openjdk.java.net"awt-dev at openjdk.java.net; Sergey Bylokhov
Subject: Re: Review Request: JDK-8144735 [hidpi] javax/swing/JWindow/ShapedAndTranslucentWindows/TranslucentPerPixelTranslucentGradient.java fails

On 8/17/2016 11:30 AM, Prem Balakrishnan wrote:

Hi Alexandr,

Thankyou for the review.

YES scaled SurfaceData returns proper scale values from getDefaultScaleX()/getDefaultScaleY(), which I have used in the current patch.

Please review the updated patch

HYPERLINK "http://cr.openjdk.java.net/%7Epkbalakr/8144735/webrev.02/"http://cr.openjdk.java.net/~pkbalakr/8144735/webrev.02/

  Is it always true that the 'as' variable which has type AccelSurface in the fix is always instance of SurfaceData?

  Thanks,
  Alexandr.

Regards,
Prem

From: Alexandr Scherbatiy 
Sent: Tuesday, August 16, 2016 10:06 PM
To: Prem Balakrishnan; Rajeev Chamyal; HYPERLINK "mailto:awt-dev at openjdk.java.net"awt-dev at openjdk.java.net; Sergey Bylokhov
Subject: Re: Review Request: JDK-8144735 [hidpi] javax/swing/JWindow/ShapedAndTranslucentWindows/TranslucentPerPixelTranslucentGradient.java fails

On 8/16/2016 7:35 AM, Prem Balakrishnan wrote:

Reminder

From: Prem Balakrishnan 
Sent: Friday, August 12, 2016 6:36 PM
To: Alexander Scherbatiy; Rajeev Chamyal; HYPERLINK "mailto:awt-dev at openjdk.java.net"awt-dev at openjdk.java.net; Sergey Bylokhov
Subject: RE: Review Request: JDK-8144735 [hidpi] javax/swing/JWindow/ShapedAndTranslucentWindows/TranslucentPerPixelTranslucentGradient.java fails

Hi Alexandr and Sergey,

Please review the updated patch.

HYPERLINK "http://cr.openjdk.java.net/%7Epkbalakr/8144735/webrev.01/"http://cr.openjdk.java.net/~pkbalakr/8144735/webrev.01/

>>"It is a little bit strange bug, because VolatileImage should handle this scale internally, and create double sized surface when necessary".-Sergey

Yes as you mentioned Volatile Image is getting scaled internally. Thanks for the feedback.

Cause: In  VIOptWindowPainter::updateWindowAccel(psdops, w, h) call, width and height were passed without scaling,
which was creating a bitmap of specified width and height, hence the output was clipped.

  I just have two general questions.
  - The scaled SurfaceData should return proper scales from getDefaultScaleX()/getDefaultScaleY() methods. Do these methods return right values after setting the scaled image sizes in the fix?
  - Region.clipScale() which is used in many places rounds values. The usual rule is to use Math.floor() for image coordinates rounding and Math.ceil() for sizes.
    Should the same rule be applicable here?

  Thanks,
  Alexandr.

Regards,
Prem

From: Alexandr Scherbatiy 
Sent: Thursday, August 04, 2016 6:23 PM
To: Prem Balakrishnan; Rajeev Chamyal; HYPERLINK "mailto:awt-dev at openjdk.java.net"awt-dev at openjdk.java.net; Sergey Bylokhov
Subject: Re: Review Request: JDK-8144735 [hidpi] javax/swing/JWindow/ShapedAndTranslucentWindows/TranslucentPerPixelTranslucentGradient.java fails

On 8/3/2016 10:04 AM, Prem Balakrishnan wrote:

Hi,

Please review fix for JDK9,

Bug: https://bugs.openjdk.java.net/browse/JDK-8144735  

Webrev: http://cr.openjdk.java.net/~pkbalakr/8144735/webrev.00/ 

Issue:

javax/swing/JWindow/ShapedAndTranslucentWindows/TranslucentPerPixelTranslucentGradient.java fails 

Cause:

While creating Transparent VolatileImage, width and height was NOT hidpi scaled.

Fix: VolatileImage width and height are scaled.

   I believe this is an issue in AWT and needs to be discussed on awt-dev alias.

   Should the backbuffer width and height be also scaled for the BIWindowPainter?

  Thanks,
  Alexandr.

Thanks,

Prem

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/2d-dev/attachments/20160829/af48fe7b/attachment.html>