Experimental Physics and
Industrial Control System

Benjamin Franksen via Core-talk <[email protected]> · Wed, 18 Sep 2019 15:47:23 +0200

Am 05.09.19 um 16:45 schrieb Michael Davidsaver via Core-talk:
> On 9/5/19 2:26 AM, Johnson, Andrew N. wrote:
>> The callback.h header contains this:
>>
>>> /*
>>>  * WINDOWS also has a "CALLBACK" type def
>>>  */
>>> #ifdef _WIN32
>>> #       ifdef CALLBACK
>>> #               undef CALLBACK
>>> #       endif /*CALLBACK*/
>>> #endif /*_WIN32*/
>>
>> I think that was set up so that as long as you include the clashing windows header first (and don’t need to use their CALLBACK facilities) you should be able to include callback.h afterwards and it will hide the Windows definition.
>>
> 
> Sadly, not.  I know this is an ordering issue.  What I
> don't know __where__ the issue comes from as the MSVC
> error output doesn't mention any line numbers in pdbsingle.cpp.
> 
> I'm also unable to find where (all) the windows headers
> define CALLBACK.  A (case insensitive) google search
> finds a lot of unrelated chatter about callback functions.
> The mingw headers define CALLBACK in a half dozen different
> headers.

Another trick to avoid name clashes is to rename the symbol:

#define CALLBACK EPICS_CALLBACK
typedef struct whatever CALLBACK;

The idea is that anything that uses CALLBACK would have to include
callback.h and thus in the module that includes it we redefine it to
EPICS_CALLBACK. This requires that system headers are included last.

One can also rename the windows CALLBACK as in

#define CALLBACK WINDOWS_CALLBACK
#include <windows.h>
#undef CALLBACK

Cheers
Ben

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System