Re: Handling testcases that are known to fail?

From: rogeeff (rogeeff_at_[hidden])
Date: 2002-01-23 22:50:44

I was not clear enough in previos message.  You can specify a number 
of assertions that could fail inside any test case. Your point is 
still valid but it should be rather rare case. Any other more 
sophisticated method would affect usability. For example,

void foo()
{
    BOOST_CHECK(...);
    BOOST_CHECK_EQUAL( ... );
    BOOST_CHECK_CLOSE( ..., 1e-5 );
}

Sunce floating-point arithmentic working differently on different 
platforms, you may expect that last assertion may fail on some of 
them. So when adding new test case you would write:

...
ts->add( BOOST_TEST_CASE( &foo, 
#if defined(__SUNPRO_CC)
1
#else
0
#endif
) );

Gennadiy.

--- In boost_at_y..., Jeremy Siek <jsiek_at_c...> wrote:
> 
> Actually, just having the expected number is not very accurate... 
what if
> by some freak accident an expected failure turns into a success 
while at
> the same time a usually successful test starts to fail. The number 
of
> failures will stay the same, telling us nothing about the problems 
that
> are occuring. It would be much better to have a specific list of 
the tests
> that are expected to fail.
> 
> On Thu, 24 Jan 2002, rogeeff wrote:
> rogeef> New Test Library present mechanism for defining expected 
number of
> rogeef> failures. You still need to set appropriate value for each 
compiler.
> rogeef>
> rogeef> Gennadiy.
> 
> --------------------------------------------------------------------
--
>  Jeremy Siek                          http://php.indiana.edu/~jsiek/
>  Ph.D. Student, Indiana Univ. B'ton   email: jsiek_at_o...
>  C++ Booster (http://www.boost.org)   office phone: [phone]
> --------------------------------------------------------------------
--