Re: [PATCH v2 07/13] topology: Add private data parser

[Liam Girdwood <liam.r.girdwood@linux.intel.com>]

On Tue, 2015-07-07 at 18:19 +0200, Takashi Iwai wrote:
> On Tue, 07 Jul 2015 17:54:02 +0200,
> Liam Girdwood wrote:
> > 
> > On Wed, 2015-07-01 at 18:20 +0200, Takashi Iwai wrote:
> > 
> > > 
> > > > +static int get_hex_num(const char *str)
> > > > +{
> > > > +	char *tmp, *s = NULL;
> > > > +	int i = 0;
> > > > +
> > > > +	tmp = strdup(str);
> > > > +	if (tmp == NULL)
> > > > +		return -ENOMEM;
> > > > +
> > > > +	s = strtok(tmp, ",");
> > > > +	while (s != NULL) {
> > > > +		s = strtok(NULL, ",");
> > > > +		i++;
> > > > +	}
> > > > +
> > > > +	free(tmp);
> > > > +	return i;
> > > 
> > > Hmm, this just counts the number of comma + 1, so you don't need to
> > > duplicate the string?
> > > 
> > 
> > The string here is duplicated since strtok is destructive (it overwrites
> > the delimiters with NULL) and the string is being used later on by the
> > calling function.
> 
> Yes, but what I meant is something like below:
> 
> static int get_hex_num(const char *str)
> {
> 	int i = 0;
> 
> 	if (!*str)
> 		return 0;
> 	for (;;) {
> 		i++;
> 		str = strchr(str, ',');
> 		if (!str)
> 			return i;
> 		str++;
> 	}
> }
> 
> ... so that it works without strdup().
> 
> But it seems that strtok() skips the starting delimiter or handles
> multiple delimiters as a single, so the result will become
> inconsistent.  That is, all the following strings will give "a" and
> "b" via strtok():
> 	a,b
> 	a,,b
> 	,a,b
> 	a,b,
> 
> I guess you don't want to have an empty list element, right?
> 

Lets ask the author :) but IMO an empty list should be skipped here.

Yao, what's your rational behind this code ?

Liam