[OpenJDK 2D-Dev] [8] request for review: 7124347: [macosx] "java.lang.InternalError: not implemented yet" on call Graphics2D.drawRend

[Andrew Brygin]

Hello Jim,

  I do not think that suggested change allows a null composite per se:
  the check for null was introduced just to make sure that we can safely
  call isDerivedFrom method. However, it does not change the method
  behavior for the case of null composite.

  If it is impossible to get null composite here for some reason 
(unfortunately
  I can not say this for sure), then this check is not required. 
However, I do not
  think that the extra check for null may cause any harm.

Thanks,
Andrew


On 29.11.2012 4:07, Jim Graham wrote:
> Is there a reason for allowing a null comp in the isSupported method?
>
>             ...jim
>
> On 11/26/2012 4:38 AM, Andrew Brygin wrote:
>> Hello,
>>
>>   could you please review a fix for 7124347?
>>
>>   This fix does not implement getRater() in ogl surfaces.
>>   Instead, it provides a blit for custom composite, which
>>   prepares a snapshot of the destination surface (which
>>   is one of ogl surfaces) and delegates the work to the
>>   general blit, which now extracts raster from the snapshot.
>>   A result of general blit's work is transferred to the
>>   original destination.
>>
>>   Changes in OGLSurfaceDataProcy are required in order to prevent
>>   getting an accelerated copy of original source image as an
>>   operand of the blit. Now we first check for composite type,
>>   and only then (if composite is not specified, or is a kind
>>   of alpha composite) we take into account other conditions.
>>
>>   So this fix does not change anything for the case of alpha composites
>>   but affects only XOR and custom composites case.
>>
>>   Note that the fix solves the problem only for the case of blits
>>   (i.e. drawImage()). Other operations may result into the same
>>   problem if custom composite is used. However, in case of XOR,
>>   this change solves the problem completely, because other rendering
>>   operations in the ogl pipeline are ready for XOR composite.
>>
>>   Supplied regression test verifies that the fix makes the problem gone.
>>   I have verified both windows and macosx platforms.
>>
>>   Please take a look and share your comments.
>>
>> Bug: http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=7124347
>> Webrev: http://cr.openjdk.java.net/~bae/7124347/8/webrev.00/
>>
>> Thanks,
>> Andrew