Re: [Gnustep-cvs] r30361 - in /libs/gui/trunk: ChangeLog Source/GSNibLoading.m

[Wolfgang Lux]

I wrote:

> Fred Kiefer wrote:
>
> > Am 12.05.2010 09:59, schrieb Wolfgang Lux:
> > > Author: wlux
> > > Date: Wed May 12 09:59:57 2010
> > > New Revision: 30361
> > >
> > > URL: http://svn.gna.org/viewcvs/gnustep?rev=30361&view=rev
> > > Log:
> > > Fix bug where some objects loaded from a .nib file were not sent an
> > > -awakeFromNib message.
> > >
> > > Modified:
> > >     libs/gui/trunk/ChangeLog
> > >     libs/gui/trunk/Source/GSNibLoading.m
> >
> > Hi Wolfgang,
> >
> > are you sure this new code is correct now? I changed that same code
> > backwards and forwards a few time in March and your change could  
> > break
> > things for Doug again. (See our mails from March, 25th on the GNUstep
> > developer mailing list) If I remember correctly the problem Doug  
> > had at
> > that time was that my change resulted in the NIB owner not getting a
> > awakeFromNib message call.
>
> I'm not absolutely sure that the new code now is correct, but the old
> code for sure was incorrect as some nib objects (in my particular case
> some submenus of the main menu) definitely were not sent an  
> awakeFromNib
> message. If the nib owner does not receive an awakeFromNib, I guess  
> that
> we have to fix the nib loader to send this message explicitly to the
> owner. If so, should this be done before or after awakening the other
> objects?

Update: Thanks for making me aware of the issue with the owner, which  
indeed was no longer sent an -awakeFromNib message. Investigating  
things a bit further revealed that your change in r30018 to iterate  
over the _objects map instead of the _names map essentially was  
correct and I now have restored it together with an explicit - 
awakeFromNib sent to the owner after all other objects have been  
woken (which seems to happen on Cocoa -- at least according to my  
limited tests).

> > Somehow the way Apple uses the names and objects maps in the NIB  
> > format
> > is hard to understand. If we ever get it right here, we then  
> > should have
> > a look at the XIB loading, this surely is broken as well.
>
>
> Indeed, it just took me a while longer to get this (hopefully) right
> because I first had to find out a way to enumerate all objects  
> contained
> in the xib file (commit r30365).

Turns out that there are more issues with xib loading. I guess it  
will take me a bit longer to fix things here (without any claim of  
getting things right ;-).

Wolfgang