Re: [Gcl-devel] Porting Femlisp (second trial)

[Camm Maguire]

Greetings!

Earlier versions of GCL pulled all the 'package operations' out of the
initialization vector and ran them first.  This was causing obvious
errors of another kind stemming from the reordering of the top level
forms in the file.  It is on my todo list to make an exhaustive check
of the evaluation of these forms under all eval-when scenarios,  What
we have now might not be quite hardened yet.

Take care,

Nicolas Neuss <address@hidden> writes:

> Camm Maguire <address@hidden> writes:
> 
> > Thank you for this.  This is indeed a reproducible bug.  It will
> > require a radical change which I'm hesitatnt to attempt at this
> > moment.  Basically, GCL writes an initialization lisp code sequence
> > at the end of each compiled file.  It reads the code, loads the
> > object, identifies the objects in the code with their addresses in the
> > loaded binary, then evaluates the forms.  What it needs to do is read
> > one form at a time, identify, and eval, etc.  You make and use the
> > package in the same initialization sequence.  A standard workaround is
> > to put all your defpackage statements in a separate file which is
> > loaded first.  Is this doable for you?
> >
> > We will eventually get to this, but it would appear that other
> > priorities are more pressing at the moment.  Please let me know how
> > this might impact your work -- that is how the pririties are set in
> > the first place :-).
> 
> [additional information, off-list]
> 
> BTW, I am surprised that it is that difficult.  If I remember correctly I
> have tried some earlier GCL version (2.5.x, maybe Debian/stable) and this
> special problem did not happen.
> 
> Yours, Nicolas.
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah