Re: CVS commit: src

[David Laight <david%l8s.co.uk@localhost>]

On Sat, May 28, 2011 at 11:42:37PM +0100, Julio Merino wrote:
> On 5/28/11 10:46 PM, Christos Zoulas wrote:
> >In article<20110528161256.AB898175D0%cvs.netbsd.org@localhost>,
> >Matthias Scheler<source-changes-d%NetBSD.org@localhost>  wrote:
> >[...]
> >>+   assert(close(fd) == 0);
> >
> >Please don't create assertions that contain code, because compiled with
> >-DNDEBUG they vanish.
> 
> Are these test code?  If so, just replace assert by one of:
> 
> - ATF_REQUIRE(boolean_expression)
> - ATF_REQUIRE_EQ(expected_value, actual_value)
> 
> like:
> 
> ATF_REQUIRE(pipe(fds) != 0);
> ATF_REQUIRE_EQ(1, write(fds[1], "", 1));

Actually, the assert() define (and similar) are often less than useful
since they don't give any indication of why things are wrong.
The ATF_REQUIRE_EQ() might be more useful (I've not looked at its definition),
but it ought to be possible to use something like:

#define CHECK(type, fmt, v1, op, v2) \
    do { \
        type _v1 = (v1), _v2 = (v2); \
        if (!((_v1) op (_v2))) \
            check_error(#v1, #op, #v2, __LINE__, fmt, _v1, _v2); \
    } while (0);

#define CHECK_I(v1, op, v1) CHECK(int, "%d", v1, op, v2)

So check_error() can print the both values of the erronous test.

        David

-- 
David Laight: david%l8s.co.uk@localhost