Re: [boost] [test] Failing to build on Sun

From: John Maddock (john_at_[hidden])
Date: 2007-11-01 13:33:53

Gennadiy Rozental wrote:
> John Maddock <john <at> johnmaddock.co.uk> writes:
>
>>
>> Now that the Sun regression tests have started up again (with thanks
>> to Noel Belcourt and Sandia), Boost.Test is failing to build on that
>> platform: see http://tinyurl.com/ywm486 for an example but there are
>> probably other errors as well.
>>
>> I wonder if this is as simple as missing the correct includes?
>> Certainly to use ::perror you would need <stdio.h> rather than
>> <cstdio> (as the latter only defines std::perror).  I wonder if this
>> is the cause of not finding snprintf as well?
>
> Can you add
>
> #if defined(__SUNPRO_CC)
> #  include <stdio.h>
> #endif
>
> in BOOST_SUN_BASED_DEBUG section of debug.ipp configuration. I can do
> it later today.

I committed a slightly different patch earlier today: 
http://svn.boost.org/trac/boost/changeset/40639/trunk/boost/test/impl/debug.ipp

Note that regardless of the platform, <stdio.h> is required in order to use 
::snprintf and all of <sys/stat.h>, <sys/wait.h> and <sys/time.h> is 
required to use ::select, so while it *may* be compiling on other platforms, 
it's kind of by accident, and these headers should really always be included 
when using those API's.

HTH, John.