[OpenJDK 2D-Dev] Review Request: (JDK-8048782) OpenJDK: PiscesCache : xmax/ymax rounding up can cause RasterFormatException.

[Jim Graham]

Hi Prasanta,

This fixes the symptom without fixing the underlying issue which is that 
there is poor sharing of information between the objects in the Pisces 
package.  The cache adds 1 to bboxX1 and bboxY1 for its own internal 
purposes - and it seems extremely questionable that it should do so as 
reading through the code it isn't clear that it should - and then 
another object grabs its bbox data, making assumptions about how those 
fields are managed, and gets it wrong.  Filtering the answer for 
"rightness" does not fix the underlying problem of how PTG gets its 
data.  I'd recommend a solution that brings us closer to having the code 
make sense, though we should look for a minimal solution since Marlin 
may patch much of this code when that project gets integrated.  Some 
suggestions:

- Add PiscesCache.getBBox() which does the proper "- 1"s

- Adjust PiscesCache to not need to add one.  This is probably the best 
overall solution since it gets rid of "voodoo" code in PiscesCaceh, but 
it would require a lot more testing.  This could get risky, so the first 
solution above, though it might add more code, is probably better for 
now.  Note that in PiscesCache, allocating the rowAARLE array already 
adds one to the number of rows allocated which seems odd because now we 
allocate 2 additional rows due to the bbox+1 adjustment.  Also, bboxX1 
is never even used so adding 1 to it just seems pointless other than to 
confuse code that reads the values.

			...jim

On 2/19/2015 8:39 PM, prasanta sadhukhan wrote:
> Hi,
>
> I would like a review for a solution of this bug.
> In OpenJDK9, "java.awt.image.RasterFormatException: (x + width) is
> outside raster" is thrown when the bounding box width outgrows original
> raster. This is because PiscesRenderingEngine does not take care of
> clipping the bounding box correctly.
>
> Bug:https://bugs.openjdk.java.net/browse/JDK-8048782
> webrev: http://cr.openjdk.java.net/~serb/prasanta/8048782/webrev.00
>
> Regards
> Prasanta