Re: [PHP4BETA] cvs: php4 /ext/standard/ string.c

On Thu, 1 Jun 2000 [email protected] wrote:

> 
> 
> On Thu, 1 Jun 2000, Andrei Zmievski wrote:
> 
> > On Thu, 01 Jun 2000, Jouni Ahto wrote:
> > > jah		Thu Jun  1 06:38:23 2000 EDT
> > > 
> > >   Modified files:
> > >     /php4/ext/standard	string.c 
> > >   Log:
> > >   
> > >   (ucwords) Fixed fix for fix for fix for #4748.
> > >   # Today's most fixed bug.
> > >  
> > >  	r=return_value->value.str.val;
> > >  	*r=toupper((unsigned char)*r);
> > > -	for(r_end = r + return_value->value.str.len -1 ; r < r_end ; r++ ) {
> > > +	for(r_end = r + return_value->value.str.len -1 ; r < r_end ; ) {
> > >  		if(isspace(*r)) {
> > > -			*++r=toupper((unsigned char)*r);
> > > +			*r=toupper((unsigned char)*++r);
> > 
> > Are you sure about this? It seems that you're replacing space character
> > with the uppercased next one.
> 
> Didn't happen for me. At what point of the line does r actually get
> incremented? I read the first version as 'uppercase the char r points to
> and assign it to incremented r', the second 'increment r, uppercase the
> char it points to, and assign it to r (ie. itself)'. But in fact, both
> versions did work for me. I was in fact fixing something else, skipping
> over when there are an even number of space chars.

What you're doing is not defined in C/C++.  The order of evaluation of
lvalue vs. rvalue is undefined.

So,

*r = foo(++*r);

Could be equivalent to both:

*r = foo(*r);
++*r;

and

++*r;
*r = foo(*r);

It's compiler dependant.  You need to fix it...

Zeev

-- 
Zeev Suraski <[email protected]>
http://www.zend.com/