Re: NSPopUpButton

[Fred Kiefer]

Just when I thought we have resolved the popup problems things are
turning rather complicate again. But Nicola is right:

> This is quite strange as I would have thought as obvious that the sender
> would have always been the NSPopUpButton ... and Georg agrees with me
> otherwise he wouldn't have been surprised by finding that the sender was a
> matrix in this case ... I would consider that an implementation bug in the
> apple code ... surprising APIs are bad APIs ... it's not really meaningful
> that the sender changes depending on which action is used ... the general
> rule is that the sender is the control, in this case the NSPopUpButton ...  
> there is even less sense in exposing the NSMatrix as the sender, because
> the NSMatrix in this case is an internal private object used by the code
> to implement the behaviour ... the public interface goes through
> NSPopUpButton, so that is the sender which should be exposed to
> applications.

We should take care to produce the least surprising results. And this
would be to send the action always from the same object. For GNUstep
this cannot be the NSMatrix, as we currently don't use one in our
implementation of NSMenu. (Should we? Nicola, you are an expert on
NSMatrix, would there be any benefit in switching from our current
NSMenuView implementation into one that uses a NSMatrix?)
But before we decide on any final solution, I still would like to here
the MacOSX behaviour. Could anyone with access to MacOSX ask Georg for
his test program and report the results on this mailing list?

Fred