[Axiom-developer] Re: Database bootstrap!

[Waldek Hebisch]

> Waldek,
> 
> That's excellent! Hopefully you've kept good notes about which
> things are needed at each stage of the process so I can try to
> reproduce it. Great work, painfully achieved, I'm sure.
> 

I have rather sparse notes.  My experiece is that even with best
notes it may be quite hard to reproduce build process -- this is
not restricted to Axiom.  So I normaly use scripts, they are
much better at recording needed steps and much easier to test
for correctnes.  And when I read code 6 month later I can
immediatly see what is done, no need to guess what given
piece of prose is intended to say.

I you (or somebody else) want to see details I have put the
diff (against wh-sandbox revision 593) at:

http://www.math.uni.wroc.pl/~hebisch/prog/db-boo4.diff

The diffs hooks into algebra Makefile, but the essential part
is really a build script: stamp-bootstrap rule has as a
dependency stamp-db rule.  stamp-db rule has as dependency
all spad file, so all spad files are unpacked from 
pamphlets before bootstrap.  The action on stamp-db rule runs
main database build script boot-sum.sh.  Once the stamp-db
is made we have databases which are good enough for the next
step and also a set of object files corresponding to categories
and core domains.  

As I wrote, one need a bunch of tweaks to the build process, and
action on stamp-bootstrap is really a build script implementing
those tweaks.

I think that the only part needed for bootstrap which is not in
the patch is $bootstrapDomains flag (implementation of this flag
consists of two lines).

Those scripts needs cleanup.  ATM the process is very sequential
and hard to restart in the middle if something goes wrong.  Many
files are recompiled needlessly.  Also, I use lists of files
which hardcode order -- ideally we should automatically generate
such lists.

-- 
                              Waldek Hebisch
address@hidden