Re: [PATCH v4 2/2] lib/test-string_helpers.c: add string_get_size() tests

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Wed, 2015-09-16 at 13:21 +0200, Rasmus Villemoes wrote:
> On Tue, Sep 15 2015, Andy Shevchenko <
> andriy.shevchenko@linux.intel.com> wrote:
> 
> > On Tue, 2015-09-15 at 15:55 +0200, Vitaly Kuznetsov wrote:
> > > +static __init void test_string_get_size_one(u64 size, u64 
> > > blk_size,
> > > +					    const enum 
> > > string_size_units units,
> > > +					    const char 
> > > *exp_result)
> > > +{
> > > +	char buf[16];
> > > +
> > > +	string_get_size(size, blk_size, units, buf, 
> > > sizeof(buf));
> > > +	if (!memcmp(buf, exp_result, strnlen(exp_result, 
> > > sizeof(buf) 
> > > - 1) + 1))
> > 
> > Actually you don't need to do this +- 1. Either you will have '\0' 
> > or
> > not, it will be checked by memcmp() anyway.
> > 
> > Thus,
> > memcmp(buf, exp_result, strnlen(exp_result, sizeof(buf))).
> 
> Huh? How does that ensure that string_get_size put a '\0' at the 
> right
> spot? We do need the comparison to also cover the terminating '\0' in
> exp_result.

Ah, you are right.

But seems we have length of the exp_result always smaller than buffer
size, so, would we change this to 
 memcmp(…, strlen(exp_result) + 1);
?

> [It would be nice if we could assert at compile-time that
> strlen(exp_result) < sizeof(buf).]

Interesting if BUILD_BUG_ON can help here. Can we use
sizeof(exp_result) since all of them are literal constants?

> 
> > Perhaps one line comment here
> > /* Make sure that buf will be always NULL-terminated */
> > 
> > > +	buf[sizeof(buf) - 1] = '\0';
> 
> <bikeshed>Could we pretty-please use different names for 0 the 
> pointer
> and 0 the character, say in this case nul or NUL or '\0' or simply
> 0. Also, I don't see the value of the comment; that line is a totally
> standard idiom.</bikeshed>.

Got your point.

-- 
Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Intel Finland Oy