Posted by mauromol
on April 22, 2010 at 2:43 AM PDT
first of all I would like to say thanks to Sun developers for their support regarding my previous topic:
The problems mentioned there have all been fixed for 1.6.0_u18. Thank you very much!
However, I'm still here to report some other problems that are affecting Java Web Start (either directly or indirectly) so that our customers are getting more and more the feeling that this technology is actually broken.
They are mainly problems regarding the check of the signatures of JAR files.
Here are the links:
This problem sometimes happens using 1.6.0_u18, while it never happened with 1.6.0_05 or 1.5.0_x. As I wrote in my comment, we are distributing an application via Java Web Start and all our classes are in a JAR correctly signed. However, sometimes Java Web Start says that the signer information of one of the classes in that JAR does not match that of other classes of the same package (in the same JAR)... which is clearly wrong. In fact, trying to restart the application immediately after this error is given and in the exact same manner, makes the application start correctly! And, as I said, this never happened with JWS clients of JREs <= 1.6.0_05.
Review ID: 1750096 (reported by me on Aprile 08, 2010, still not received any feedback)
This is not directly related to Java Web Start, but I didn't see it happening with 1.6.0_05, while it is happening now with 1.6.0_18 and I think it's related to how Java Web Start is now handling the JRE selection to run the application, especially when the JNLP asks for the use of a JRE which is not the newest installed one and when the Java Console is set up to be shown always.
In java.awt.Component.BltBufferStrategy.showSubRegion(int, int, int, int) a NullPointerException can happen if java.awt.Component.BltBufferStrategy.dispose() is called meanwhile. Bug #6281999 fixed a case when this could occur, but it seems that a problem can still happen and I suspect it is due to the fact that there is no null-checking against the Graphics obtained from the backbuffers array.
Another problem that is not affecting us for now (but may do in the near future, since it's quite severe) is the following:
This is related to one of the problems I mentioned in my previous topic and is again about JAR signature verification, which is failing if a JAR is not empty but contains only non-Java classes files (just configuration files, for instance).
I would appreciate very much if I could get some feedback about these topics too.