axiom-developer
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Axiom-developer] Re: algebra Makefiles with explicit dependencies, boot


From: root
Subject: [Axiom-developer] Re: algebra Makefiles with explicit dependencies, bootstrap, fixed-points etc.
Date: Mon, 3 Jan 2005 18:29:13 -0500

Bill,

Ah, you've hit the nail in my memory. The cyclic dependencies will
cause the system to build in an arbitrary way. Adding the dependencies
causes the algebra to violate the build order because make thinks it
knows better.

Once the algebra exists you can rebuild in any order so the bug only
shows up in the first build. Thus the algebra does not "really" depend
on the dependencies you are trying to hard code. You can remake any
algebra file without making the files it depends on if you have a 
running axiom with already built algebra. Indeed, if you modify one
of the .spad.pamphlet files then typing "make" will do the minimum
amount of work correctly.

re: BOOTSTRAP everything

The problem with this idea is that it is not needed. The bootstrap
files are very fragile but only for the first build. Once the algebra
files have been built it is possible to change the .spad files any
way you'd like. The 'hole' that a user can fall thru happens when they
change an algebra file that has BOOTSTRAP code (which they did NOT
update) and then try to build the system from scratch. In this case
the first copy of the system will be built with no-longer-valid code.

What the user needs to do to change algebra which has BOOTSTRAP code is:

build the system
change the .spad file
recompile the .spad file
replace the BOOTSTRAP section of code with the newly generated .lsp

now the BOOTSTRAP code mirrors the algebra code and all is well.

I don't ever expect a user to reach that far down into the algebra.
Making BOOTSTRAP files for all of the algebra opens this hole everywhere.

Tim




reply via email to

[Prev in Thread] Current Thread [Next in Thread]