[Evolution-hackers] Re: dlopen / API hooks for evolution ...

[michael meeks <michael meeks novell com>]

Hi Frank,

On Mon, 2005-01-31 at 15:45 +0200, Frank Schönheit wrote:
> According to Heiner Rechtien: No way. His points are
> - we're developing C++, not C, so what do you need from this lib what
>   isn't (semantically equivalent) available elsewhere?

	The glib api is part of the e-d-s API - however, it is the stable part.

> - glib has just recently been removed - the only thing still depending
>   on it is the Gnome integrator. RE would be quite unhappy with
>   re-introducing it.

	Well - they need it for the gtk+/vcl backend on Linux & Solaris -
so ... it has to be there for the same platforms we want to build this
code on.

> - This would create yet another "version hell", if we'd do this.

	Not really; glib (in stark contrast to e-d-s) is extremely stable at
both the API and ABI level - and has been since March 2002 when it was
released - approaching 3 years ago.

> So I fear we need to find another solution :(
> What kind of code do you need from glib?

	Well; we call ~another g_object type 10 methods. Of course we can grab
these in the same way via dlopen - but since that is not even slightly
pleasant or useful for us, it'd be best if Sun engineering do that I
think - I'll have a poke at MHU first though.

	Anyhow - I'm not eager to do that personally; however - the EApi code
is in-place to let that be done fairly easily.

	I suggest we make it easy to disable the evoution2.0 integration, and
have it default disabled in the build until such time as you want it
built & can be bothered to do this extra, unpleasant work.

	Regards,

		Michael.

-- 
 michael meeks novell com  <><, Pseudo Engineer, itinerant idiot