Pete Bartlett wrote:
Hi Robert,
The reason you haven't had any user complaints is that there hasn't been a release with the new #error-ing version, has there? It's a racing certainty that if you did do such a release then you would get requests to make the old behaviour work when a #define was set (BOOST_PERMIT_DEPRECATED_STATIC_WARNING or something). These files have had a strange status though so perhaps being more cavalier than usual with deprecation is possible.
My first inclination was to make a "warning pragma". but it turns out to be compiler dependent and I didn't want to mess with it unless it was really necessary. Right now, I only want to merge in the serialization library and leave this as a separate question to be left up to concensus. I have no special interest in the manner which it is handled.
On an implementation note - why do you have an unconditional #error for all compilers and then follow that up with a couple of conditional #{pragma message,warning}s for various compilers... the latter seem redundant (see e.g. http://svn.boost.org/trac/boost/browser/trunk/boost/static_warning.hpp)
It is redundent. I put the #error in there to be sure that when I ran my tests I would detect any cases where I failed to change the library/tests to use the new locations of these items. I wasn't sure whether I should leave in the #error, leave in the #warning, or just remove the files. Robert Ramey