From: Lyle Johnson <lyle@...>
Date: 2003-04-03T03:32:19+09:00
Subject: Re: FXRuby 1.0.21 FXOptionMenu problem

James Adam wrote:

> Basically, whenever I use an FXOptionMenu, or specifically an FXOption
> to which i then ".connect(SEL_COMMAND)" some code to, changing to that
> option when the app is running results in the following error being
> printed out on the commandline:
> 
> FXRuby.cpp(667): FXASSERT(FALSE) failed.

<snip>

This is a bug. If you want to locally patch your code, add the following 
snippet in FXRbConvertMessageData():

   else if(sender->isMemberOf(FXMETACLASS(FXOption))){
     if(type==SEL_COMMAND)
       return to_ruby(reinterpret_cast<FXEvent*>(ptr));
     }

Based on the fact that you've already been looking at that function, I 
think you'll see where to add it. This will be fixed for FXRuby-1.0.22.

> [nb. as an aside, it seems to trigger segfaults within random OpenGL
> calls, usually resulting in [BUG] rb_gc_mark() errors. this might just
> be my code, but then i 'neuter' the FXOptionMenu the problem
> disappears]

Yes. If we don't recognize the sender & message type for the message 
(e.g. a SEL_COMMAND message from an FXOption instance) we just pass 
through the message data as-is, assuming it's a valid Ruby object. This 
was done in the event that you create your own widgets with special 
message data. Since in this case the message data is not a Ruby object 
(but rather a pointer to a C++ FXEvent object) you'll eventually get a 
runtime error of some sort.