Re: [boost] BOOST_WORKAROUND and g++ -Wundef

From: David Abrahams (dave_at_[hidden])
Date: 2005-10-15 13:38:44

Angus Leeming <angus.leeming_at_[hidden]> writes:

> David Abrahams wrote:
>
>> I object to adding all these specific silencing tests.  IIRC we were
>> going to handle this by creating boost macros for each compiler,
>> similar to BOOST_MSVC, that were always #defined (possibly to 0), and
>> using those in the BOOST_WORKAROUND tests.  Otherwise, we will end up
>> uglifying a lot of code, where BOOST_WORKAROUND is supposed to make
>> the code much simpler and more readable.
>
> Right. Agree 100%.
>
> The problem is that, at the moment, no such macro is always
> defined. 

That should be done in boost/config/suffix.hpp

> How about macros like this in workaround.hpp:
>
> ======================================================
> #if defined(__BORLANDC__)
> # define BOOST_WORKAROUND_BORLANDC(X) BOOST_WORKAROUND(__BORLANDC__, X)
> #else
> # define BOOST_WORKAROUND_BORLANDC(X) 0
> #endif

No thanks.  We need BOOST_BORLANDC, etc.  There are times when those
values will be useful without the BOOST_WORKAROUND macro.

> Sorting through the sources, it appears that there are BOOST_WORKAROUNDS
> used for the following macros (extraction script also below). Would you
> like me to proceed on a bitwise basis, defining macros as above and then
> applying it to the code base, or can you describe a better way (for me) to
> proceed?

It isn't just the uses of BOOST_WORKAROUND that will benefit.  I'm
sure there are plenty of legacy tests that should be converted.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com