Re: [Gcl-devel] conditions/clos/gcl unified build patch and instructions.

[Vadim V. Zhytnikov]

Vadim V. Zhytnikov пишет:
> Camm Maguire пишет:
>
> > Hi Vadim!  Thanks for the patch!!
> >
> > Unfortunately, it breaks the maxima build.  As you have it, the final
> > maxima link complains that the package "COMMON-LISP-USER" does not
> > exist.  When I reverted the missing lines building this package in C,
> > the error is that package "CONDITIONS" does not exist.  Even when one
> > can fire up unixport/saved_gcl and see the conditions package.
> >
> > Hope to look into this soon so we can commit something like the patch
> > below.

I'm not sure why it doesn't work to you but I've just build gcl and
successfully build Maxima with this it.  I can see all new
packages in it (:lisp (list-all-packages)).  Even more it pass
major part of Maxima tests.  The only thing which is broken so far
is errcatch instruction.

errcatch(1/0);

causes Lisp error. The reason of this is more or less clear.
With new packages we redefined several functions (ERROR, BREAK
...) with new ones from CONDITIONS.  I don't know what is the
the essential difference between these functions and shall we
really import them from CONDITIONS or not. More thorough investigation 
is needed.

BTW, have you noticed that GCL's reaction on lisp errors
has changed?  Now it produces some menu.  I'm perfectly sure
that these imported functions are responsible for that.

> It not only breaks Maxima build but in addition new loop.lisp does
> not work as expected. Try
>
> (loop for s being the external-symbols of "COMMON-LISP" collect s)

It is turned out that this loop construction is broken in ecls as well.
It seems that ecls's loop2.lisp which we are trying to implant in GCL
also was borrowed from CMUCL. And the above loop works in CMUCL
abut doesn't work in gcl and ecls. I don't know the fix yet but
it can wait since in general loop works for simpler constructions.

Best wishes,

Vadim