Re: [Axiom-developer] Re: 2.7 build

[Stephen Wilson]

Camm Maguire <address@hidden> writes:

> Greetings!  Did you stop here?
> 
>       TopLevelDrawFunctions 
>    Loading /fix/t1/camm/axiom/int/algebra/DRAWCFUN.nrlib/code for 
>       package TopLevelDrawFunctionsForCompiledFunctions 
>  
>    >> System error:
>    
> 
> (1) -> cp: cannot stat `/fix/t1/camm/axiom/int/algebra/*.daase': No such file 
> or directory
> make[3]: *** [/fix/t1/camm/axiom/mnt/linux/algebra/*.daase] Error 1

Yes, this is where I am at.  But my goodness, you must have a fast
machine.  I started my latest build perhaps 6 hours ago and its still
on layer 14 of the algebra build.  I envy you :)

> At some point we should discuss the replacement for emit-fn in 2.7.0.
> 
> 1) If one has #+native-reloc, (si::do-recompile) in final image and
>    save.  Optionally supply and argument and link with the recompiled
>    source.
> 
> 2) Otherwise, or if one prefers, wrap build stages in
>    'with-compilation-unit, which will automatically recompile the
>    original loaded compiled source files for use in a subsequent
>    reload and save, or link.

Personally, I favor 2), insofar as there are efforts to use ASDF as a
make replacement, which automatically wraps its components in a
with-compilation-unit.  Given this is the typical procedure for Lisp
programs, perhaps the general audience for GCL would appreciate that
support.

On the other hand, I think that (si::do-recompile) is the preferred
route for Axiom when based on iterative make style builds, as is
currently the case.

So unfortunately I cant make a conclusive decision, as both seem to me
equally important.

> The interface is still flexible -- I'd appreciate feedback in making
> it optimally useful.

I am still on the extreme uphill side of understanding both Axiom and
GCL.  If I gain any potentially useful insights, I will certainly let
you know.
 
> I take it at present, none of the patches you've made against silver
> were required to workaround gcl bugs?  If you find same, please let me
> know.  Thus far, I know of only Waldek's issue outstanding.

This is correct.  Every issue I found with GCL has been fixed.  Thank
you!!!

> emit-fn is still there and has not been changed, but has bugs.

OK.  One of the challenges at this point is to convert this `base
line' patchset into one which cooperates well with both 2.7.0 and the
2.6.8pre snapshot what has served Axiom so well.  It will likely take
some time before Axiom can drop support for CLTL1 builds altogether.

Thanks again Camm for all your hard work!  It is completely pointless
to try and measure the importance of your contributions!!

Sincerely,
Steve


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