Re: [Adonthell-devel] Freetype and SWIG configure checks

[Alexandre Courbot]

> > > Actually, we could check if a 'CVS' directory exists and let configure
> > > fail if SWIG is not found or not recent enough. Otherwise we might
> > > display a message telling people to update SWIG if they plan to make
> > > changes to the code.
> >
> > Checking for the CVS directory wouldn't cover all cases IMO. Someone
> > could want to develop with the released versions - actually I find it
> > better to display a message each time py_adonthell_wrap.cc need to be
> > rebuild and SWIG isn't good. This will always abord compilation when
> > needed - even when people want to mess with releases.
> 
> Uh, seems my writing was a bit misleading: 
> I meant configure should fail if SWIG is not found or not recent _and_ a CVS 
> directory exists. That's a sure thing, so nothing would speak against it.
> 
> If there is no CVS directory, SWIG is not required, so configure mustn't 
> fail. Instead it displays a simple warning message and continues. That's what 
> I have done an hour ago or so.
> 
> Regarding your point: it doesn't hurt to have the check in the Makefile as 
> well. That way people who did not notice the warning will know what is wrong. 
> However, the current check only test whether swig is installed or not. We 
> should change that to look for the proper version too. Otherwise people might 
> be able to compile Adonthell, but will not be able to run it afterwards, 
> which is much worse.

Unless I missed something, it seems to me that testing SWIG in the
Makefile whenever py_adonthell_wrap.cc needs to be rebuild is sufficiant
and covers all cases, no? So why adding an additional check in the
configure? The user would be warned that he needs SWIG sooner, true. But
he might also just want some Makefiles ready to be able to build the
tools which doesn't depend on the Python wrappers - and this would be
impossible in this case. I'd say, making the Makefile test more accurate
so it correctly detects SWIG's version and displays a message would do
it better. The user would only be warned a few minutes later that he has
to get SWIG - and he wouldn't even need to run the configure script
again to continue his build, so that catches with the warning delay ;)

Alex.
-- 
http://www.gnurou.org