From: William Harold Newman <william.newman@ai...> - 2001-05-28 14:57:27
I'm thinking of making the fasl file version number independent of
the CPU. Instead of
in backend.lisp and
(setf *backend-fasl-file-version* 2)
in e.g. src/compiler/alpha/backend-parms.lisp, it could just be
(defconstant +fasl-file-version+ 11)
somewhere, perhaps still in in backend.lisp for now, and eventually
in the shared-load/dump-machinery package I keep thinking about.
So far, most of the changes which have required changes in the fasl
file version number have been independent of CPU. (See the comments in
src/compiler/x86/backend-parms.lisp.) Making the backend fasl file
version independent of CPU would make it easier to keep track of such
The only drawback I can see is that when a change is made which needs
only impact one CPU, it will still cause a backwards compatibility
problem for all CPUs. Since I expect that most fasl file changes will
continue to affect all CPUs, I don't expect this will be a big problem.
William Harold Newman <william.newman@...>
"The beatings will continue until morale improves." -- ??
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C B9 25 FB EE E0 C3 E5 7C
From: Daniel Barlow <dan@te...> - 2001-06-01 12:48:27
William Harold Newman <william.newman@...> writes:
> I'm thinking of making the fasl file version number independent of
> the CPU.
Sounds reasonable to me.
On the general subject of Once And Only Once, you might also look at
the alien declarations for os_context_register_addr and friends.
There's a note on my TODO list suggesting that the code in
alpha-vm.lisp is mostly cut & paste from the x86 equivalent and could
stand to be moved somewhere arch-independent.
http://ww.telent.net/cliki/ - Link farm for free CL-on-Unix resources