Hello all,
I have just encountered a bug while using SBCL 0.9.0.1 or OpenMCL 0.14.3 and the latest CVS slime (as of 10 minutes ago) on my PowerBook G4 with Mac OS 10.3.9. Evaluating the following commands using C-M-x (except for the in-package, which I have to type at the REPL) works fine:
(defpackage :fast-math (:use :cl))
(in-package :fast-math)
(defmacro f+ (&rest xs) `(the double-float (+ ,@(map 'list #'(lambda (x) `(the double-float ,x)) xs))))
(defmacro f- (&rest xs) `(the double-float (- ,@(map 'list #'(lambda (x) `(the double-float ,x)) xs))))
(defmacro f* (&rest xs) `(the double-float (* ,@(map 'list #'(lambda (x) `(the double-float ,x)) xs))))
(defmacro f/ (&rest xs) `(the double-float (/ ,@(map 'list #'(lambda (x) `(the double-float ,x)) xs))))
(f* 10.d0 30.d0) (f+ 10.d0 30.d0) (f- 10.d0 30.d0) (f/ 30.d0 10.d0)
However, if I then type
(f* 10.d0 30.d0)
at the REPL, I get an error involving the "undefined function f*". Even after this, if I then evaluate (f* 10.d0 30.d0) using C-M-x it works fine! I notice that typing (f* ...) in the REPL doesn't bring up the autocomplete information from SLIME, while typing it in a different slime buffer does bring up the note (f* &rest xs). The problem doesn't occur if the macros are defined in the cl-user package.
I don't really know anything about the internal workings of slime or I'd post a fix, too. Sorry.
Let me know if there's more information I can offer about the problem.
Thanks! Will Farr
On 4/29/05, Will M. Farr farr@mit.edu wrote: < snip definitions >
(f* 10.d0 30.d0) (f+ 10.d0 30.d0) (f- 10.d0 30.d0) (f/ 30.d0 10.d0)
However, if I then type
(f* 10.d0 30.d0)
at the REPL,
At the REPL, type:
(in-package :fast-math)
Then type (f* 10d0 30d0).
Does it work?
The behavior you're seeing is intended. When you're in the lisp file buffer, and you C-M-x something, slime treats that command as being executed in the package specified by the (in-package ...) form at the top of the file. Thus, the macro definitions can be found. In the REPL, you're in CL-USER, where those functions don't exist, and so SBCL/OpenMCL can't find the definitions.
Hope this helps.
-Denis PGP: http://pgp.mit.edu:11371/pks/lookup?search=0xA1B51B4B&op=index
At the REPL, type:
(in-package :fast-math)
Then type (f* 10d0 30d0).
That didn't fix it, but the problem was related; when I evaluated the defmacro commands in a secondary buffer, it didn't have (in-package :fast-math) at the top, so the macros were installed in cl-user. Of course, when in the fast-math package at the REPL, the commands weren't available! I forgot that slime plays that package magic---thanks for the tip!
Will
On Fri, Apr 29, 2005 at 01:43:04PM -0400, Will M. Farr wrote:
I have just encountered a bug while using SBCL 0.9.0.1 or OpenMCL 0.14.3 and the latest CVS slime (as of 10 minutes ago) on my PowerBook G4 with Mac OS 10.3.9. Evaluating the following commands using C-M-x (except for the in-package, which I have to type at the REPL) works fine:
(defpackage :fast-math (:use :cl))
(in-package :fast-math)
[....]
However, if I then type
(f* 10.d0 30.d0)
at the REPL, I get an error involving the "undefined function f*". Even after this, if I then evaluate (f* 10.d0 30.d0) using C-M-x it works fine! I notice that typing (f* ...) in the REPL doesn't bring up the autocomplete information from SLIME, while typing it in a different slime buffer does bring up the note (f* &rest xs). The problem doesn't occur if the macros are defined in the cl-user package.
It sounds like you've got package problems.
The way this works is that each buffer has a separate package (or packages), and the REPL has a current package.
Buffer packages are determined automatically by the last IN-PACKAGE form before point. You can see the buffer package in the modeline:
--:** scratch.lisp All (3,0) (Lisp Slime:common-lisp-user)---------
The REPL package is shown on the prompt:
CL-USER>
It sounds like you defined the functions in a different package from where your REPL is.
You don't have to evaluate IN-PACKAGE forms in buffers at all - they get automatically picked up. You can experiment with this by typing a new one and watching the buffer package change on the modeline on the fly, and by having more than one and watching it change as you move between them.
The REPL package is changed by either running "(in-package :foo)", or by using the change-package REPL command (type ",change-package" or ",!p" at an empty REPL prompt).
In your case, make sure the REPL prompt shows "FAST-MATH> " since that's where your buffer is designated to go.
If this still doesn't work you can, from the repl, type (apropos "F*"), which will print all symbols containing F*. If your macro shows up in the output it will be fully qualified such that typing the name as it shows up will get at it. If you're in the correct package, it should show up simply as "F*", and if it has a package qualifier it will at least tell you where it wound up and you can try to figure it out from there.
-bcd
-
Brian Downing -
Denis Bueno -
Will M. Farr