Re: gtkmm2.4 and MSVC (again)

["Murray Cumming" <murrayc murrayc com>]

> Murray Cumming wrote:
>>>There are some inconsistencies in the way
>>>global variables are dll exported (gendef <->
>>>__declspec(dllimport/dllexport) with GLIBMM_API #dedfines).
>>
>>
>> Now that we are using this gendef thing, shouldn't we just remove
>> GLIBMM_API? It doesn't seem to be used much anyway.
>
> No, we need them for declspec(dllimport). Unlike gcc, global variables
> are not automatically tagged by MSVC with the dllimport flag (there
> might be a compiler flag though...).

Yeah, or classes or functions either. But I thought that was what gendef
fixed.

> Anyway, I will build gtkmm-2.4.6 for mingw this week-end but hopefully
> will commit on monday some changes to g[lib|tk]mmconfig.h.in required
> for MSVC. I now have Debug and Release versions of gtkmm for MSVC with
> the gtkmm-demo fully functional. If somebody could send me a example
> (not k3d !!!) that suffers from the infamous "Access violation - no RTTI
> data!" error, I will happily try to compile it against my build.

Wonderful. Thanks. Feel free to commit anything you need.


Murray Cumming
murrayc murrayc com
www.murrayc.com
www.openismus.com