Experimental Physics and
Industrial Control System

Michael Davidsaver <[email protected]> · Tue, 21 Mar 2017 11:09:29 -0400

On 03/21/2017 09:23 AM, Mark Rivers wrote:
>> extern "C" is always needed for export from C++ due to name mangling.  I
>> don't think there is anything windows/msvc specific here.
> 
> It links OK without extern "C" on Linux, but not on Windows.

You're right, I see the same at least for gcc 4.9.2.  However, I think
this is implementation defined behavior on the part of gcc, which is
choosing in this instance not to mangle variable names.

I would maintain that extern "C" is correct as this will match what is
found in the generated _registerRecordDeviceDriver.cpp

> 
> Mark
> 
> ________________________________________
> From: [email protected] [[email protected]] on behalf of Michael Davidsaver [[email protected]]
> Sent: Tuesday, March 21, 2017 8:15 AM
> To: Ralph Lange; EPICS Core Talk
> Subject: Re: epicsExport.h on Windows
> 
> extern "C" is always needed for export from C++ due to name mangling.  I
> don't think there is anything windows/msvc specific here.
> 
> The first definition is never used and imo. should be removed.
> 
>>> extern statements in epicsExport.h are ignored by VS compiler
> 
> They are not ignored.  They just don't have the same effect. 'extern'
> and 'extern "C"' behave differently despite using the same keyword.
> 
> 
> On 03/21/2017 08:16 AM, Ralph Lange wrote:
>> Dear Core-Talkers,
>>
>> For a device support partly written in C++, we have been getting a pull
>> request for Windows builds that adds extern "C" declarations around
>> EPICS exports, e.g.
>>
>>     -epicsExportAddress(double, connectInterval);
>>     +extern "C" {
>>     +    epicsExportAddress(double, connectInterval);
>>     +}
>>
>>
>> with the comment
>>
>>     I am afraid that the extern statements in epicsExport.h are ignored
>>     by VS compiler when expanding the epicsRegisterFunction macro.
>>     Therefore they need to be repeated here otherwise we get a linking
>>     failure when building test IOC.
>>
>>
>> As my Windows experience is very limited - is there a better way to
>> handle that?
>>
>> Thanks,
>> ~Ralph
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System