[GRASS-dev] debugging nviz with TclTk 8.5 aqua

[Glynn Clements]

William Kyngesburye wrote:

> >>>>> Looks like we need the tcltk header paths.
> >>>>
> >>>> In that case, I think that we're better off ditching the OSX- 
> >>>> specific
> >>>> stuff in its entirety and going back to -I, -L and -l switches.  
> >>>> No -F,
> >>>> no -framework.
> >>>>
> >>> We needed the -framework flags because -l would get the system tcltk
> >>> symlinks.
> >>
> >> Er, not if you also use -L, surely?
> >>
> > Yes, because the aqua framework does not install lib symlinks -  
> > there is nothing to link to using -ltcl -ltk.  So we must use the - 
> > framework flags.
> 
> 
> So, reinstate the txltk aqua option?  And then let the user specify  
> tcltk-includes.  I'd say default the headers to /Library/Frameworks  
> tcl & tk, and recommend that the user installs tcltk aqua 8.5, but  
> requiring the user to supply the paths is also OK.

IMHO, that has to be absolutely the last resort. I'd rather just say
"you need a *working* version of Tcl/Tk".

Mac users can try harassing ActiveState into providing something that
actually works, or they can provide a "community" package which
actually works, or they can even decide that they're fed up with the
Mac trying to be Unix and NotUnix at the same time and switch to
something sane.

It's not as if the same issue won't apply to other projects (many of
which use automake, which certainly isn't going to handle that kind of
weirdness).

Ultimately, the effort would probably be better spent on hastening the
migration away from Tcl/Tk to Python.

-- 
Glynn Clements <glynn at gclements.plus.com>