[cells-gtk-devel] Re: Cells-gtk]

Peter Denno peter.denno at nist.gov
Fri May 27 03:06:56 UTC 2005


On Thursday 26 May 2005 20:13, Kenny Tilton wrote:
> >Hi Ralf,
> >
> >Thanks. I think you mentioned this before, but now it is starting to sink
> > in. It works (just loading "libglib-2.0.so" etc). So for linux/cmucl at
> > least, we don't need all this searching for paths to the libraries.
> >
> >I'll need to check what lispworks and allegro are doing.
> >
> >For each of macosx and windows we look in only one place for the
> > libraries.
>
> Oy, this discussion reminds me that I have been meaning to extend hello-c
> to Just Work(tm) on loading libraries. Back when Cello got ported to Linux,
> Frank Goenninger added a bunch of fancy code to Cello to get libraries
> loaded. I had the same thought then: let's do this in Hello-C, not Cello or
> Cells-Gtk.
>
> Gee, for once the Open Source fairy is whining at other people to do work
> for him. :):) But seriously, please, whoever tackles this, put any fixes in
> Hello-C and send me what you do and I will commit it to Hello-C.

But I am not adding fancy code, I am removing fancy code. It appears that our 
design was built on the faulty premise that we had to specify the full path 
to the library. At least on linux, (I've tested cmucl and lispworks so far) 
we don't have to. (Unless, as Ralf points out, it isn't one of the places the 
system looks for libraries).

On macosx and win32 we always had a hard-coded path. We still do, but I moved 
that to the configuration file. It doesn't matter with win32 because the user 
presumably followed our directions to place the GTK libs where we want them.

There isn't any new code in hello-c/uffi to make this happen.

-- 
- Best regards,
   Peter



More information about the cells-gtk-devel mailing list