[OpenJDK 2D-Dev] RFR: 8166038 BufferedImage methods getTileGridXOffset() and getTileGridYOffset() return a non 0 value for sub images

[Martin Desruisseaux]

Hello Philip

Le 17/04/2020 à 21:08, Philip Race a écrit :

> So if this is "always zero", for BufferedImage, why aren't we just 
> directly returning zero ?
>
For consistency with existing BufferedImage.getMinX() and getMinY() 
method implementations. Those methods shall always return zero (this is 
also stated in the specification), but BufferedImage implementation 
nevertheless delegates to raster.getMinX() and getMinY(). The 
BufferedImage constructor verifies that (raster.minX == 0 && raster.minY 
== 0), but despite this restriction some BufferedImage methods are 
implemented for arbitrary raster.minX and raster.minY values (see for 
example getData() — note however that setData(Raster) is not robust to 
arbitrary minX and minY).

I don't know why some (but not all) BufferedImage methods have been made 
robust to arbitrary minX and minY. So for deciding what to put in 
getTileGridXOffset(), I took the formula and substituted each term by 
the corresponding BufferedImage implementation. This process was 
documented in a previous version of this change request:

http://cr.openjdk.java.net/~serb/8166038/webrev.00/src/java.desktop/share/classes/java/awt/image/BufferedImage.java.udiff.html


> I'm a bit on the fence about this wondering whether to submit a CSR 
> here, since a behavioural change is being introduced in order to 
> conform to the spec, and there is some compatibility issue possible.
>
Note that previous behavior was mathematically inconsistent with 
RenderedImage specification and with Java Advanced Imaging (JAI) 
implementation (e.g. PlanarImage.XtoTileX(int)). It went unnoticed maybe 
because Java2D has optimization for the BufferedImage case which bypass 
the getTileGridXOffset() and getTileGridYOffset() values.

     Regards,

         Martin