I'm not completely clear on why the variable is becoming nil, but I
don't think it should.  Coding around it bothers me -- if the object
engine is acting wierd, this is serious!  FWIW, it doesn't happen on my
machine, so I'm all the more clueless.  I tried it with a perfectly
vanilla 2.9a-2915 image.

Did you test with a freshly downloaded image, or with a development
image?  If freshly downloaded, I'm wondering about your VM.  What VM are
you using?  If development, could you try it with a development image? 
(And, could you try a simple "revert" to the older method?  Perhaps a
recompile will fix things).

(And if anyone else wants to try this, it would also make for nice data.
 Start up an image, open a morphic project, open an IRC chat window,
choose "connect", and then hit the "connect" button.  Do you get a
walkback?)


By the way, even if we end up having to code around this thing, the "m
accept's" really are important.  If they are simply removed, then users
will have to hit alt-s on all the windows on the connection dialoge
before they can connect.  I'm sure this would be confusing.  Overall,
it's kinda tricky right now to have a big "accept" button that will
accept for several windows at once.  But that's what is wanted here,
tricky or not.


Your fix for colors is good and welcome!



-Lex



Geoffrey Corey <gcorey at redsky.com> wrote:
> Replacement method for IRCConnection fixes two problems:
>   1) Changes the background of the connection dialog so you can actually
> read the text
>   2) Fixes a problem under NT where a walkback window appears when the
> connect button is pressed
>