Experimental Physics and
Industrial Control System

Michael Davidsaver via Core-talk <core-talk at aps.anl.gov> · Thu, 8 Oct 2020 10:18:59 -0700

On 10/8/20 9:56 AM, Michael Davidsaver wrote:
> On 10/8/20 9:30 AM, Konrad, Martin via Core-talk wrote:
>> Hi Ralph,
>>> ITER needed to switch EPICS Base to use "-std=c++11" because without
>>> that setting, linking EPICS libraries would not work for other code that
>>> was compiled using C++11.
>> Weird. I tried to reproduce this by building pvxs (which builds using
>> -std=c++11) on CentOS 7 against EPICS 7 (built without -std=c++11).
>> Works fine for me.
>>
>> Are you linking against any libraries provided by a vendor as binaries?
>> Is there a way we can reproduce the problem?
> 
> @Ralph  Ah, I had forgotten about this...
> 
> @Martin  This isn't a question of libstdc++, but rather of how/where
> pvDataCPP gets 'shared_ptr<T>'.
> 
> https://github.com/epics-base/pvDataCPP/blob/320cc3c60b575fa49e49a21619472f01b8890049/src/misc/pv/sharedPtr.h#L62-L86

This means that all modules depending on pvDataCPP (which does not include PVXS)
must be built with the same selection of in sharedPtr.h.

Also, I realize that I have been getting this GCC situation confused with
a related linking problem I ran into OSX/clang when the compiler and linker
were using inconsistent implementations of the c++ standard library (from clang
vs. gcc).

https://github.com/mdavidsaver/epicscorelibs/blob/5d798854b398366cdaa4cf634651a14aad8f15d7/src/python/epicscorelibs/config.py#L114-L115

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System