Re: This NSConstantString thing

[Markus Hitter]

Am Dienstag den, 12. März 2002, um 22:50, schrieb Richard 
Frith-Macdonald:

> PS. If the Apple compiler really *does* insist on specific ivar 
> names in some way, we need to
> do something like -

It doesn't insist on names but checks the declarations. A plain 
char* is recognized but not the same inside a union. To check 
all possible valid declarations, the check could become quite 
complex.

> @interface NXConstantString : NSString
> {
> #ifdef  PRIVATE_CONST_STRING
>   union {
>     unichar     *u;
>     unsigned char   *c;
>   } _contents;
>   unsigned int  _count;
> #else
>   char *cString;
>   unsigned int len;
> #endif
> }
>
> Then define PRIVATE_CONST_STRING before including NSString.h in 
> the file where NXConstantString is actually implemented.

This solution would make the Foundation compile, of course. But 
it would make the unicode issue even worse as far as I can see 
for now.

> However, I suspect it does not insist on that at all ...

It does. The error message is:

GSMime.m: In function `decodeWord':
GSMime.m:200: interface `NSConstantString' lacks required ivars

(as GSMime.m is the first file using a @"..." thing)

To my knowledge, this change in behaviour is a requirement to 
make ObjC++ work. Details are in the depths of gcc sources ...

> and there is rather some problem with include files or some 
> other problem is confusing you.

For a simpler example, you can have a look at the code configure 
uses to check for -fconstant-string-class.


For the unicode stuff, I'll have a look tomorrow ;-)


Markus

- - - - - - - - - - - - - - - - - - -
Dipl. Ing. Markus Hitter
http://www.jump-ing.de/