help with maxima/sbcl/alpha/linux



Raymond Toy (RT/EUS) wrote:

>It would have nice if you had just cut out a few relevant lines instead 
>of making people wade through the entire log about stuff unrelated to 
>maxima.

Err, sorry about that. Post in hurry, regret at leisure.

>In any case, I think if you just load up f2cl-lib.lisp and slatec.lisp, 
>and then compile zasyi.lisp, you'll reproduce the error.

So, I tried some variations on that, and got a different error.

debugger invoked on a SIMPLE-ERROR in thread #<THREAD "initial thread" RUNNING {A834819}>:
  Error during processing of --eval option (LOAD
                                            #P"/home/bremner/tmp/test.lisp"):

    The value 3.1415927 is not of type DOUBLE-FLOAT.

Where test.lisp has

    (load "numerical/f2cl-package.lisp")
    (load "numerical/f2cl-lib.lisp")
    (load "numerical/slatec.lisp")
    (load "numerical/slatec/zasyi.lisp")

Since I get this same error on i386, I figured it was not the bug I
was hunting.  But now I try with clisp and ecl, and it seems to load
ok, so maybe that is an sbcl bug, if not necessarily the same one.

>My guess that there are enough local variables in that function to cause 
>problems when the compiler needs to save some on the stack.  The error 
>messages hints at that since it seems to want an 8-bit offset or a TN 
>(internal compiler Temporary Name).

Thanks for the feedback,

David