On 10/31/13 9:19 AM, Alexander Scherbatiy wrote:
Could you review the updated fix:
http://cr.openjdk.java.net/~alexsch/8011059/webrev.04/
I'd like to point out a potential issue here with respect to ImageObserver.
Typically, when the ImageObserver is notified that the image loading is
complete then it becomes irrelevant. I'm not sure if we've documented
this anywhere, and I've not done a code survey to see if anyone has ever
created code based on that assumption, but it is true. complete means
the bits are all there and all calls to drawImage() forever-more will
succeed without any problems.
The way that the scaled images are introduced here, there will be
separate image loading for both the regular and @2x versions of the
image. A developer could render the image on one resolution display
which would trigger the loading of that version of the image until it
says it is complete, but the first time they load the image on a new
display of a different resolution, then suddenly a new version of the
image is needed and will not have been loaded - but that has never been
possible before and so the developer could be caught off guard by this.
There would ideally be some way to make the status of the image only
complete when all resolution variants of an image are loaded.
This gets more complicated with the MediaTracker object which is
supposed to be used to load an image completely. We don't explicitly
say that if a MediaTracker says an image is complete that they can then
ignore their observers, and there is already an issue with animated GIFs
where we document that the tracker will claim they are done after their
first frame is loaded, but I don't doubt that many programmers have
written code to funnel all images through MT and then assume they don't
need observers after that. In fact, I believe that Swing has done
things like that internally. The question then is, what happens with MT
and @2x images?
...jim
