Eric Niebler <eric <at> boostpro.com> writes:
On 1/6/2010 2:38 PM, Gennadiy Rozental wrote:
Any reason to prefer this implementation to the one currently in trunk?
You mean, aside from the fact that it is standard C++03, conforms to Boost's coding guidelines, requires no compiler-specific workarounds, and is unlikely to be broken on platforms that are not being actively tested; whereas the one currently on trunk is none of those things? Not really, no. :-P
Current one does not require compiler specific workarounds, is unlikely to be broken on platforms that are not being actively tested, is cleaner, uses correct tools for the job and conforms to the commonly supported (included next C++) C standard (not sure about what specific coding standard you refer to). Boost.Test does use several interfaces in various places which are not 100% portable according to C++03. Not sure if this innocent macro worth the trouble of maintaining the version which needs to be eventually changed (to use va_copy once it is commonly available). That said, if you feel strongly about this - go ahead and check this in. Gennadiy