The basic problem with the above combination is that JButtons actionPerformed is executed even if some InputVerifier to some component objects. That may (and in fact does) to storing of invalid data in some cases. As a comment to the above bug states, the main reason for this behaviour is that in BasicButtonListener mousePressed the state of the buttonModel is set to armed and pressed _before_ requesting the focus to the button. Some experiments showed that the only reliable way to prevent it and leave the button in a consistent state is to reverse the order of calls: first request the focus and then if the button actually received it go ahead and change the state. (Every of the other suggested workarounds at least lead to an unconsistent state of "half-pressed": button looks pressed whenever the mouse enters). Now the question: Does any of you see any deeper reason to have the sequence of method calls as is? Would there be any hidden inconsistencies? To solve the problem would require to reverse the order in the keybindings to mnemonic, default button handling as well... :-( Any comments/ideas? Jeanette _______________________________________________ Advanced-swing mailing list [EMAIL PROTECTED] http://eos.dk/mailman/listinfo/advanced-swing
