Re: 23.0.60; Seg fault in xfaces.c at line 6703 (Emacs.app on GNUstep)

[Chris Hall]

> > With the patch I supplied, at least the user knows there is an issue 
> > with 
> > realizing the default face, rather than SIGSEGV (11).

> Since this is a programming error in an internal structure, in a 
> development 
> version of the code, it is letting the developer know there is an 
> error. 
> Developers have debuggers, they don't need code to catch bugs and 
> exit 
> gracefully. To write code to catch every potential NULL pointer 
> exception in 
> the internal structures would make Emacs bloated and slow.

Thanks, but I think I already pretty much understood that.  If I may 
quote myself:

> > > I am aware that sometimes some classes of errors are perhaps best 
> > > allowed to 
> > > happen and to result in catastrophic failures like segmentation 
> > > faults,

From an uninitiated point of view, I saw an attempt to reference a 
member of an uninitialized structure, while in the process of 
initializing, as you point out, a development version of the program.

Based on the program's subsequent behavior, this structure is clearly 
*required* to be properly initialized in order to successfully realize 
a face, during the first call to 
'Fdisplay_supports_face_attributes_p'.

Further, the struct in question is called 'face_cache', which suggests 
to me that once one has realized a face, one can cache a reference to 
it in this struct.  And further suggests to me that before realizing a 
face, one would have no need of a cache, so why go to the trouble of 
initializing it, if one wasn't yet sure whether a face even *could* be 
realized?  (Never mind *requiring* it to be initialized.)

Yet this struct clearly needs to be properly initialized as a 
necessary prerequisite of realizing a face.

Maybe this is part of some sort of optimization, i.e., by delaying 
attempting to realize a face until the first time someone asks if the 
display supports face attributes?  Or am I perhaps misinterpreting the 
name/purpose of the struct?


TIA,

Chris Hall