[asdf-devel] new asdf does not like matlisp.asd

[Mario S. Mommer]

Robert Goldman <rpgoldman at sift.info>
writes:
> Question: is it possible to get an analogous issue using only lisp files?

Yes, up to one thing. I use cat instead of a linker (and do not load the
resulting file, but I think that is irelevant)

I uploaded a tarball

http://common-lisp.net/~mmommer/matlispasdfissue.tar.gz

which is the same as before, but that includes a definition file called
abug.asd that does not use any alien business. There are no fortran
components, only plain lisp ones and instead of the so, some bogus file
is generated using cat on the components of unix-dso, which is never
loaded. I just wanted to preserve the semantincs of a module that has
components but itself does something with the components once the
components are done. The operation-done-p has this bogus file to check
if it should run the op or not.

With the old asdf, compilation happens only on the first load, while in
the new asdf, files get compiled again.

Interestingly, with the old asdf, the following happens:

CL-USER> (asdf:operate 'asdf:load-op 'abug)
;;; stuff elided.
; $ cat "/home/mommer/borrar/matlisp/nothere.fasl"  > "/home/mommer/borrar/matlisp/libmatlisp.fasalad"
;;; more stuff elided.
CL-USER> (asdf:operate 'asdf:load-op 'abug)
; $ cat "/home/mommer/borrar/matlisp/nothere.fasl"  > "/home/mommer/borrar/matlisp/libmatlisp.fasalad"
;;; whoa. It ran the "linker" again!
NIL
CL-USER> (asdf:operate 'asdf:load-op 'abug)
NIL
CL-USER> 

Any ideas?

Regards, and thanks,
        Mario