[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-2d-dev
Subject:    Re: [OpenJDK 2D-Dev] [9] Review Request: 8039774 [OGL] Image painting is broken if 'sun.java2d.accth
From:       Andrew Brygin <andrew.brygin () oracle ! com>
Date:       2014-04-11 8:28:44
Message-ID: 5347A7BC.4000300 () oracle ! com
[Download RAW message or body]

Hello Sergey,

  the fix looks fine.

Thanks,
Andrew

On 4/10/2014 3:45 PM, Sergey Bylokhov wrote:
> Hello.
> Please review the fix for jdk 9.
> The bug occurs, when we try to call drawImageBg with scale and 
> 'sun.java2d.accthreshold' is set to 0.
>
> The bug exists in the DrawImage.renderImageXform(). This method is 
> called, when all our attempts to scale the image fails, and we create 
> a buffered image, and use it as intermediate buffer.
> Note that the buffered image has predefined image type 
> (TYPE_INT_RGB/TYPE_INT_ARGB), because we know that for these types we 
> have necessary tranform helpers.
> So if "sun.java2d.accthreshold" set to 0 the next code [1] is broken, 
> because dstData.getSourceSurfaceData() returns cached version of the 
> surface, which is unexpected.
>
> [1]:
>             srcData = dstData.getSourceSurfaceData(img,
> SunGraphics2D.TRANSFORM_GENERIC,
>                                                    sg.imageComp,
>                                                    null);
>             srcType = srcData.getSurfaceType();
>             helper = TransformHelper.getFromCache(srcType);
>
> !!!!! helper is null here.
>
> I can reproduce the problem on OGL, and sure that it will be 
> reproduced on d3d as well.
>
> BTW: we have to do something with this code path because DrawImageBg() 
> in this scenario 10000 slower than DrawImage().
>
> Bug: https://bugs.openjdk.java.net/browse/JDK-8039774
> Webrev can be found at: 
> http://cr.openjdk.java.net/~serb/8039774/webrev.00
>

[prev in list] [next in list] [prev in thread] [next in thread]