Re: Possible bug with adding subviews?

[Richard Frith-Macdonald]

On Tuesday, February 12, 2002, at 07:07 PM, Stephen Brandon wrote:

> Hi,
>
> I converted a MacOSX nib using nib2gmodel, and am debugging the 
> results :=(
>
> The issue here is that GNUstep complains thus:
>
> Uncaught exception NSInvalidArgumentException, reason: addSubview: 
> creates a
> loop in the views tree!
>
> The problem occurs under the following conditions:
>
> - unarchiving from gmodel file
> - NSBox object ("box") has contentview "a"
> - nsview "a" also appears in box's subviews array
> - nsview "a" has superview "box"

All the above look correct to me.

> - during the unarchiving of "a" (superview "box" is decoded, then 
> [superview
> addSubview:self] is called)
>
> The exception is thrown in -[NSView addSubview:], where the box's
> _content_view object is asked to add *itself*, since _content_view == 
> "a":
>
> if ([self isDescendantOf: aView]) {...   (throws exception)
>
>
> I'm not sure where the best place to fix this is. Should the unarchiver 
> do a
> check to see if the superview's contentview equals itself? (but then the
> superview would never add the view to its subview list). Or should 
> NSView
> just do a check to see if a view is adding itself as a subview, and 
> silently
> ignore it?
>
> It's also possible that the situation as described above should never
> actually happen, and that either my original nib is in a mess, or 
> nib2gmodel
> made an error in converting it.

It sounds to me like there is an error in the decoder ... the NSView 
decoder
is setting the subviews, then the NSBox specific code is setting the 
content
view ... which adds it as a subview ... but it's already there.

There are three places it could be fixed ...
1. in the gmodel NSBox decoder
2. in the NSBox setContentView: method
3. in the NSView addSubview: method

I think 1 is almost certainly the correct place to change.