boost_stacktrace_impl_return_nullptr as an exported symbol

[lufriem]

#159 added the 'stacktrace from current exception for MSVC' feature but one (I guess) sideeffect is that I now get an exported symbol 'boost_stacktrace_impl_return_nullptr', which makes MSVS always generate an input library (.lib) for my (non-library, ie EXE) project (its linker just triggers off the __declspec(dllexport) tokens, with no way of suppressing this, see https://developercommunity.visualstudio.com/t/why-is-vs2017-building-an-import-library-for-my-ex/233569#T-N237446-N243028). Due to some (admittedly dodgy) project setup I ran into troubles with this file present (when it did not exist before), which I fixed on my side, but I'm wondering about this export..

I guess there's a reason for having that symbol be exported? I see that using the BOOST_SYMBOL_EXPORT macro I guess I could suppress its creation, but I wouldn't really want to deviate from your default project/configuration defaults..

Sorry for the rambling, I guess my question is, is this expected and something that users just have to deal with?

Thank you

[apolukhin]

Make a PR with a fix and let the CI run. We'll see, if removing that BOOST_SYMBOL_EXPORT breaks something

[huangqinjin]

You may use /NOIMPLIB linker option to suppress the generation of .lib file. Without BOOST_SYMBOL_EXPORT, no symbol boost_stacktrace_impl_return_nullptr exist and /ALTERNATENAME would not work.

[lufriem]

Thank you both for your replies! I also searched around a bit but I'm afraid I'm out of my depth in regards to the boost build system.

One thing I saw that I neglected to mention (sorry!): I'm using a static build of the boost libraries. I found an old thread discussing the 'issue' of boost exporting symbols (also in the context of exception reporting, funnily enough), I suppose under these circumstances (static lib) it would also be possible to globally suppress their creation (as they wouldn't be needed/wanted anyway) using the BOOST_DISABLE_EXPLICIT_SYMBOL_VISIBILITY switch. Under any other circumstances (ie shared library) I don't think that would be a good idea.

[huangqinjin]

I mean to apply /NOIMPLIB to your EXE linker option, not the boost library itself.

[lufriem]

Yes thank you I understood that, I was just wondering whether there was still something that could/should be done to the library itself

[huangqinjin]

I have no idea if there is a better method to mimic the __attribute__((weak)) behavior. Maybe just provide a macro to #ifdef-out the symbol if boost_stacktrace_from_exception is linked.

[apolukhin]

Maybe just provide a macro to #ifdef-out the symbol if boost_stacktrace_from_exception is linked.

Hm, the BOOST_STACKTRACE_DYN_LINK macro is defined if dynamic library is being built. I'll try the ifdefing tomorrow

[lufriem]

Thank you both for your efforts!