Re: Strange initialization problem with Gtk::Main

[John Emmas <johne53 tiscali co uk>]

On 29/10/2013 17:12, Murray Cumming wrote:
> Are you trying to instantiate a gtkmm type before gtkmm has been
> initialized? For instance are you using a gtkmm type in a global static
> object?

Not that exactly but essentially I think that's the problem.  The 
program that works is a very simple exe which doesn't need to link to 
external DLLs.  The one which fails is much more complex and it needs to 
link to several other DLLs - but some of those DLLs also link to glibmm 
and giomm.  Remember that (up until today) I'd build glibmm and giomm as 
static libs (mainly for simplicity).  So I think what's happening is 
that giomm gets initialised (probably in the main exe) then later, one 
of the DLLs makes a giomm call - but whatever giomm code is in the DLL 
hasn't been initialised yet so the call fails.  I've now managed to 
build glibmm as a DLL and I'll rebuild giomm tomorrow, along with 
pangomm, cairomm et al.  My only concern is that the 'gendef' approach 
seems to be exporting a lot of stuff which seems unnecessary (such as 
std::vector / std::bad_alloc etc) so I'm fearful that by the time 
they're all rebuilt as DLLS there might be a lot of duplicated symbols 
when I come to re-link the main program :-(

Time will tell and hopefully namespacing might come to my rescue!  I 
should have a better handle on this tomorrow.

John