That's a useful idea, but doesn't alter the fact that date_time 1.35 has broken previously working code and should be altered to work harmoniously with other boost libraries. The windows include order dependencies are a nightmare and I think it's the responsibility of libs that force their inclusion to avoid making the situation worse than it is. In other words, if there are include order dependencies in boost it's a bug, IMO. Jim

-----Original Message----- From: boost-bounces@lists.boost.org [mailto:boost-bounces@lists.boost.org] On Behalf Of Felipe Magno de Almeida Sent: 31 March 2008 20:22 To: boost@lists.boost.org Subject: Re: [boost] Bug in 1.35: header inclusion order dependencyintroduced between recursive_mutex.hpp and asio.hpp on Windows

On Mon, Mar 31, 2008 at 11:00 AM, James Talbut <James.Talbut@omg3d.com> wrote:

...

recursive_mutex.hpp include windows.h, which includes winsock.h. asio.hpp barfs if winsock.h has been included.

Just define WIN32_LEAN_AND_MEAN globally. Then winsock.h will not be included.

[snip]

...

Jim

-- Felipe Magno de Almeida _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

________________________________________________________________________ This e-mail, and any attachment, is confidential. If you have received it in error, do not use or disclose the information in any way, notify me immediately, and please delete it from your system. ________________________________________________________________________

Anteru wrote:

James Talbut schrieb:

...

That's a useful idea, but doesn't alter the fact that date_time 1.35 has broken previously working code and should be altered to work harmoniously with other boost libraries. The windows include order dependencies are a nightmare and I think it's the responsibility of libs that force their inclusion to avoid making the situation worse than it is. In other words, if there are include order dependencies in boost it's a bug, IMO.

IMO, too, and I've openend a ticket for it: http://svn.boost.org/trac/boost/ticket/1740

Not only because it breaks winsock.h, but I have an enum member ERROR and it's useless after including date_time :/

A global enumerated value called ERROR is not a good idea.

Sohail Somani wrote:

On Tue, 01 Apr 2008 13:48:03 -0400, Edward Diener wrote:

...

A global enumerated value called ERROR is not a good idea.

In C++, a global anything is not a good idea especially if you use other libraries. At the least, you should be doing C-style namespacing, i.e., product_ERROR, if you are into that sort of thing.

You mean like this: #include <boost/date_time.hpp> // or <windows.h> namespace X { class Y { typedef enum {OK=1, ERROR} myEnumT; }; } // namespace X int main(int argc, char* argv[]) { return 0; } The issue is not that the ERROR was globally enumerated. The issue is that ERROR is #defined in one of the headers that is included by windows.h. This is, of course, the fault of the windows headers for not respecting the set of symbols that it is allowed to define, but it is surprising to have otherwise valid code break by the inclusion of a boost header. Here is another example that is even harder to track down: ---8<--- Z.h ---8<--- #ifndef Z_incl #define Z_incl class Z { public: int GetMessage() const { return val; } private: int val; }; #endif ---8<--- main.cpp ---8<--- #include "Z.h" #include <boost/date_time.hpp> // or <windows.h> void foo() { Z z; z.GetMessage(); // error: GetMessageA is not a member of Z } ---8<--- ---8<--- David