On Mon, May 08, 2000 at 11:56:51AM -0500, Martin Atzmueller wrote:
> Under sbcl-0.6.3
>
> when I define a function toplevel, and try to disassemble it, it fails.
(I hope I'm not violating internet etiquette too badly by cc'ing this
reply to the sbcl-help list.)
Thank you, this is completely broken and I didn't notice for a long
time. As it happens, I stumbled across this problem myself last week
and fixed it, so I can tell you the problem, and if you're very
impatient I can send you a patched version of the system which
includes the fix, or you can wait for 0.6.4 (within a week, I hope)
which will also have the fix.
According to my notes, there were two parts to the fix..
I screwed up when removing the CMU CL BACKEND and BACKEND-related
structures, neglecting to remove one use of DSTATE-PARAMS. So:
* Replace (GET-INST-SPACE (DSTATE-PARAMS DSTATE)) with
(GET-INST-SPACE), and remove the PARAMS argument
from DEFUN GET-INST-SPACE.
I didn't test DISASSEMBLE when I set up the system to work
as a true cross-compiler at bootstrap time, and so I overlooked
a problem with a macroexpansion defining things into the host
Lisp instead of the target Lisp. So:
* Change DEFMACRO to SB!XC:DEFMACRO in the code generated
by GEN-ARG-ACCESS-MACRO-FORM.
> BTW, I have one unrelated comment:
> sbcl.1 talks about vsource, but this is obviously not implemented any more.
Yes. When I rewrote the handling of printer-control variables in the
debugger (*PRINT-LEVEL*, etc.), I got rid of VSOURCE; I think I
decided it wasn't particularly necessary any more and was in any case
too messy to document properly. This is one of a number of ways that
the CMU CL manual has drifted out of date for SBCL. The good news is,
I have written a partial user manual for sbcl-0.6.4. The bad news is,
I don't have a chapter on the debugger in there. The only real
documentation for the debugger currently is
* (invoke-debugger)
0] help
--
William Harold Newman <william.newman@...>
software consultant
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C B9 25 FB EE E0 C3 E5 7C
|