Re: [Axiom-developer] build-improvements and more-rules.mk

[Waldek Hebisch]

Martin Rubey wrote:
> Dear Waldek, Tim, Cliff,
> 
> Waldek Hebisch <address@hidden> writes:
> 
> > > Waldek Hebisch <address@hidden> writes:
> > > > 
> > > > I agree that algebra bootstrap should be revisited.  But IMHO the
> > > > main work to do is outside Makefiles: I think that algebra should
> > > > build just from spad files using Spad compiler and almost empty 
> > > > databases.
> > > 
> > > Agree, but that's more than just the main work - it's a massive
> > > undertaking.
> > >
> > 
> > I think it is easier then you think.  
> 
> I agree. The only thing one has to do is to rebuild the databases after each
> layer. In fact, I did this manually to build an Axiom with my Guessing
> Package...
> 
> I said 
> 
> make DAASE=../../int
> 
> after each layer. (It's a pain, but it works.)
> 

I do not think it is _that_ easy.  As I wrote, Axiom remembers (large part
of) old database in the image, so at first you may not notice missing
database entries.  But then you are likely to hit something not cached
into the image and you will get runtime error.  Have you tried to
dump database after LAYER3?  For me Axiom crashed.

I have fixed Axiom so that before reading new data Axiom erases old data.
After that if you load empty database you wil notice compilation errors
since Axiom lost (most of) its knowledge about types.  So, to bootstarap
from scratch you need some fake categories and carefully chosen compilation
order (quite different than the current order).


-- 
                              Waldek Hebisch
address@hidden