----- Original Message ----- From: "Andrey Semashev" <andrey.semashev@gmail.com> To: <boost@lists.boost.org> Sent: Wednesday, March 17, 2010 10:50 PM Subject: Re: [boost] [log] Boost.Log Formal Review
On 03/18/2010 12:37 AM, vicente.botet wrote:
maybe
#if defined (LOG_NTRACE) #define LOG_TRACE(logger, stream) if (true); else logger #else #define LOG_TRACE(logger, stream) BOOST_LOG_SEV(logger, trace)<< stream; #endif
I would like all the BOOST_LOG logger macros works this way.
That still leaves the streaming statements compiled. Although they will likely be optimized away, this doesn't eliminate the compilation cost and does not guarantee that no other traces of logging are left (such as string constants or functions).
I think that we need to solve two issues when log is not needed. 1st : no time at all spent at runtime if log is not needed. 2nd : reduce the compilation time to a minimum log is not needed.
The first is mandatory, the second is nice to have.
One of the first reasons for this request was to remove any trace of logging from the resulting binary (e.g., for security reasons). This includes the things I mentioned.
You lost me. Could you clarify which security reasons are you talking about? I expect the compiler after compiling the expression will not generate any code associated to if (true); else ..... Vicente