--- In Boost-Users@y..., "Peter Dimov" wrote:

From: "david_at_2good"

...

I am using the free C++Builder Compiler version 5.5.1 for Win32. (http://www.borland.com/products/downloads/download_cbuilder.html)

I have used Boost 1.27 without problems, but when compiling the same code with Boost 1.28 I get this error message for the compiler:

Error E2113 c:\borland\boost_1_28_0\boost/detail/shared_count.hpp 39: Virtual function 'use_count_is_zero::what() const throw()' conflicts with base class 'exception'

Seems to work for me with the same compiler. Have you installed the service packs?

Good idea! I didn't know there were any for the free compiler but I downloaded and installed them. No difference, though. But trying even harder, I examined the compiler flags used by the framework I use. The framework uses the Stdcall calling sequence for function calls by sending the -ps switch to the compiler. Without that switch, everything works fine. With the flag, the error above. Now I just have to figure out why this is a problem... hints, anyone? \David

From: "david_at_2good"

...

But trying even harder, I examined the compiler flags used by the framework I use. The framework uses the Stdcall calling sequence for function calls by sending the -ps switch to the compiler. Without

--- In Boost-Users@y..., "Peter Dimov" wrote: that

...

switch, everything works fine. With the flag, the error above. Now I just have to figure out why this is a problem... hints, anyone?

The reason is that <exception> (exceptio.h) contains

#pragma option push -b -a8 -pc -Vx- -Ve- -w-inl -w-aus -w-sig

as its first line, making standard-conformant code a bit difficult to compile under -ps.

I can surround the use_count_is_zero definition with #pragma option push -pc, but is this the right thing to do?

I surrounded the place where I include boost/shared_ptr.hpp and boost/shared_array.hpp with the #pragma option push/pop. Not a very elegant solution, but it seems to do the trick. Surrounding the use_count_is_zero definition with the #pragmas would allow Boost to handle this case on its own, which is a Good Thing, but clutters the code with compiler-specific #ifdefs. Thank you very much for your help, David