[OpenJDK 2D-Dev] [9] request for review: 8068749: Restrict javax.imageio.spi.ServiceRegistry to ImageIO types

[Phil Race]

There is no need to deprecate them. They are not harmful.

-phil.

On 08/04/2015 03:25 AM, Sergey Bylokhov wrote:
> Probably we can mark these classes as deprecated, in favor to 
> java.util.ServiceLoader?
>
> On 03.08.15 23:49, Phil Race wrote:
>> CCC is approved. So I will give this a "+1" .. you just need one 
>> other reviewer (if
>> you did not get that already) and you can push.
>>
>> -phil.
>>
>> On 07/27/2015 07:10 PM, Stuart Marks wrote:
>>> Hi all,
>>>
>>> Please review this following code and API change:
>>>
>>> Bug:
>>>     https://bugs.openjdk.java.net/browse/JDK-8068749
>>>
>>> Webrev:
>>> http://cr.openjdk.java.net/~smarks/reviews/8068749/webrev.0/
>>>
>>> The change is to be pushed into the jdk9/client forest.
>>>
>>> The background is that this is a "preparation for Jigsaw" bug. The 
>>> javax.imageio.spi.ServiceRegistry class was introduced long ago in 
>>> order to load Image I/O service providers; in fact it was also fully 
>>> general service loading mechanism, capable of loading providers for 
>>> arbitrary service types. This latter function has been superseded by 
>>> java.util.ServiceLoader (introduced in JDK 6), although iio.SR can 
>>> still be used for general purpose service provider loading.
>>>
>>> In JDK 9, Jigsaw modularization will require additional work, and 
>>> use of new, modular APIs, to implement general purpose service 
>>> loading. Loading of specific APIs -- which is the primary use case 
>>> for iio.SR -- is considerably simpler. Given that general purpose 
>>> service loading is now handled by j.u.SL, we wish to remove general 
>>> purpose support from iio.SR and restrict its use to loading only 
>>> Image I/O related service providers.
>>>
>>> This change adds explicit checks in iio.SR to restrict the service 
>>> types to the known set of Image I/O interfaces, and it will throw 
>>> IAE if asked to load a provider for any other service type. Use of 
>>> iio.SR to load Image I/O providers will continue to work unchanged.
>>>
>>> This is clearly an incompatible change. I've done some internet 
>>> searching, and use of iio.SR to load Image I/O providers is quite 
>>> common. I was able to find one use of iio.SR to load a non Image I/O 
>>> provider, but the project where this occurred migrated to j.u.SL 
>>> several years ago. Based on this, we believe the impact of this 
>>> change to be small.
>>>
>>> s'marks
>>
>
>