Ah, yes, now I see it. I was combing through the various classes in
java/awt/image that *are* public to do some research while I was reading
the code review and somehow thought that the classes being modified were
in the same directory/package.
I now see that all of the modified constructors are in sun/awt which is
fine.
Apologies for the diversion...
...jim
On 3/29/16 1:24 PM, Sergey Bylokhov wrote:
On 29.03.16 23:19, Jim Graham wrote:
This sounded like such a great idea that I don't think we realized the
one obvious reason we can't do this...
We should take into account that this classes in sun/awt/image and is
not accessible on old jdk when the securitymanager is enabled, and will
be completely encapsulated in jdk9.
You are talking about changing the method signatures on existing public
constructors. While it might be nice to have had them typed from the
get-go, I don't think we can change them now. We might be able to add
new overloaded typed constructors (or typed factories?), but we can't
just delete or modify the existing constructors without creating a
compatibility issue...
...jim
On 3/10/16 4:48 AM, Sergey Bylokhov wrote:
Hi, Ajit.
What about other cases in Raster.java, where the DataBuffer is passed as
a parameter? Probably we can simplify the code and find all such issues
if we changes the ByteInterleavedRaster/etc. For example:
ByteInterleavedRaster(SampleModel sampleModel,DataBuffer dataBuffer,...)
to
ByteInterleavedRaster(SampleModel sampleModel,DataBufferByte
dataBuffer,...)
Because we throw an exception in such cases anyway:
if (!(dataBuffer instanceof DataBufferByte)) {
throw new RasterFormatException("ByteInterleavedRasters must have "
+ "byte DataBuffers");
}
And the compiler will help us, what everybody think about it?
On 09.03.16 17:38, Ajit Ghaisas wrote:
Hi,
Modified the test and added check for MultiPixelPackedSampleModel
condition.
Please review updated webrev :
http://cr.openjdk.java.net/~arapte/ajit/6353518/webrev.01/
Regards,
Ajit
-----Original Message-----
From: Sergey Bylokhov
Sent: Wednesday, March 09, 2016 4:06 PM
To: Ajit Ghaisas; awt-...@openjdk.java.net; Semyon Sadetsky; Ambarish
Rapte; 2d-dev
Subject: Re: [9] Review-request for 6353518: Creation of a
WritableRaster with a custom DataBuffer causes erroneous Exception
Changes for 2d area.(cc)
On 07.03.16 11:20, Ajit Ghaisas wrote:
Hi,
Bug : https://bugs.openjdk.java.net/browse/JDK-6353518
Issue : (Text from bug description)
An attempt to create a WritableRaster via
Raster.createWritableRaster(SampleModel sm, DataBuffer db, Point
location) using a custom DataBuffer causes an erroneous
RasterFormatException.
Apparently the reason for this bug is that IntegerComponentRaster
insists on being passed an instance of DataBufferInt rather than just
a DataBuffer with a DataType of TYPE_INT.
This is quite annoying since DataBufferInt is declared final and thus
cannot be extended.
Fix :
The last line of Raster.createWritableRaster() method already has a
way to handle generic case if the input does not match any of the
cases in switch statements in that method.
The fact that " *InterleavedRaster() constructors throw exception if
DataBuffer passed to them does not match the expected type" was
ignored earlier.
This fix adds a check of "DataBuffer type" before creating respective
*InterleavedRaster() objects.
It constructs the SunWritableRaster in case DataBuffer type does not
match any data specific DataBuffer classes (DataBufferByte,
DataBufferUShort, DataBufferInt)
Request to review webrev containing this fix :
http://cr.openjdk.java.net/~arapte/ajit/6353518/webrev.00/
Regards,
Ajit
--
Best regards, Sergey.
