🚀 go-pugleaf

In article <436922997wnr@dstrip.demon.co.uk>
           Steve@dstrip.demon.co.uk "Steve Rencontre" writes:

>In article: <4j9lfh$o2l@narses.hrz.tu-chemnitz.de>  
>hfst@hrz.tu-chemnitz.de (Hans Steffani) writes:
>> Steve Rencontre <Steve@dstrip.demon.co.uk> writes:
>> 
>> >Oh yes it can!
>> 
>> >     ...     
>> >     int var1, var2;
>> >     int *var1p = &var1;
>> >     var1p[-1] = 0;
>> >     ...
>> 
>> Post this to comp.lang.c[.moderated] and you will be told that this
>> leads to undefined behavior and something like that. 
>
>Well my newsreader doesn't know how to post to moderated groups, and I'm 
>too lazy to do it by hand :-), but anyway...

Have you tried posting as normal?

>It may be undefined by the *language*, but then so is the size of an 
>int!

The size of an int is implementation defined so it is guaranteed to have a
value. Moreover the language guarantees that you can store any number in
the range -32767 to 32767 inclusive in an int on any implementation. That
is a world apart from the plain undefined behaviour that the code above
results in.

>However, this very behaviour is relied on by stdargs/varargs in 
>order to make variable parameter lists possible. Yes, it's 
>implementation-dependent -- but that doesn't mean it results in 
>unpredictable or unreliable behaviour within that implementation.

True, but it does mean that it is non-portable. Not knowing the intent and
use of the code from what has appeared in comp.lang.c I cannot say
whether that is significant or not. Usually it is.

-- 
-----------------------------------------
Lawrence Kirby | fred@genesis.demon.co.uk
Wilts, England | 70734.126@compuserve.com
-----------------------------------------