Re: [Sbcl-devel] Compiling with CLISP

[William H. N. <wil...@ai...>]

On Thu, Aug 16, 2001 at 12:30:20AM -0400, Dave MacDonald wrote:
> The CLISP maintainers have recently added MAKE-LOAD-FORM, so I've
> tried over the last couple of days to get it to compile SBCL.  I've
> encountered a number of problems, but compilation gets further than
> when Martin Atzmueller last reported his progress.  However, CLISP
> still cannot complete the make-host-1.sh script.

> [various problems and fixes]

I stalled on the CLISP port because I got impatient with some strange
problem building on OpenBSD. But even if I never fix that problem, I
still have an (old and slightly flaky) machine running Linux, so I
could just stop messing with porting CLISP itself and do a little 
work with porting SBCL.

> - CLISP doesn't implement PPRINT-LOGICAL-BLOCK.  So far, I have been
>   able to ignore this problem by setting
>   *{load,compile}-{verbose,print}* to nil, thereby avoiding the need
>   to attempt a call to this undefined function.

So far I've been successful working around unimplemented i/o stuff in
cross-compilation. CMU CL didn't -- and maybe still doesn't -- do
PRINT-OBJECT or READ-SEQUENCE correctly, and it was still manageable
to build under CMU CL by avoiding those operations.
 
> - CLISP generates the following error:
> 
> Compiling file /home/dave/packages/sbcl-work1/src/code/force-delayed-defbangstructs.lisp ...
> 
> Compilation of file /home/dave/packages/sbcl-work1/src/code/force-delayed-defbangstructs.lisp is finished.
> 0 errors, 0 warnings
> ;; Loading file obj/from-host/src/code/force-delayed-defbangstructs.lisp-obj ...
> *** - VECTOR-PUSH-EXTEND works only on adjustable arrays, not on #(#<UNKNOWN-TYPE NULL> #<ARRAY-TYPE SIMPLE-STRING>)
> 
>   From stepping through sb!kernel::force-delayed-def!structs, I know
>   that the error is generated when processing the PRIM-OBJECT-SLOT
>   structure, but I don't know why it happens.  I'm not sure whether
>   SBCL or CLISP is to blame, and I'm at a loss about how to debug
>   further.

I'd guess that the vector that VECTOR-PUSH-EXTEND is failing on was
created by a MAKE-ARRAY form which didn't explicitly give it a fill
pointer, but did give it some other non-simple property. IIRC SBCL,
like CMU CL, has only one flavor of non-simple array, so if you
specify any form of non-simplicity, you get a fill pointer. Perhaps
some code has come to rely on this.

The fixes you describe sound reasonable. Would you like to send them
in as patches?

-- 
William Harold Newman <wil...@ai...>
"Three hours a day will produce as much as a man ought to write."
  -- Anthony Trollope
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C  B9 25 FB EE E0 C3 E5 7C