[Sbcl-devel] [patch] ccase-and-ecase-error-on-t-and-otherwise

[Christophe R. <cs...@ca...>]

Hi,

Attached is a patch to fix our handling of CCASE, ECASE, CTYPECASE and
ETYPECASE.  Somewhat surprisingly, it's shown up some other issues:
firstly, in handling of warnings in CAREFULLY-MACROEXPAND; I hope that
this patch doesn't break building under ancient compilers (it certainly
doesn't under CMUCL 3.x/CMUCL 18d) but it might be worth checking;
secondly, in EVAL-IN-LEXENV:
  (%eval original-exp lexenv)
[ this is after original-exp has been macroexpanded into exp ] is there
a reason why we pass the original-exp to %EVAL, rather than EXP? The
situation where this displays some strangeness is as follows (at the
REPL, with the attached patch):

* (ecase 1 (1 nil) (t t))
STYLE-WARNING:
   Treating bare T in ECASE as introducing a normal-clause, not an
   otherwise-clause
; in: LAMBDA NIL
;     (ECASE 1 (1 NIL) (T T))
; 
; caught STYLE-WARNING:
;   (in macroexpansion of (ECASE 1 # ...))
;   (hint: For more precise location, try *BREAK-ON-SIGNALS*.)
;   Treating bare T in ECASE as introducing a normal-clause, not an otherwise-clause

(note the double style-warning, one from the first macroexpansion, and
one as a result of passing the original expression to %EVAL, rather than
the macroexpanded expression). Can I just change original-exp to exp, or
was there a reason for doing it this way?

Cheers,

Christophe
-- 
http://www-jcsu.jesus.cam.ac.uk/~csr21/       +44 1223 510 299/+44 7729 383 757
(set-pprint-dispatch 'number (lambda (s o) (declare (special b)) (format s b)))
(defvar b "~&Just another Lisp hacker~%")    (pprint #36rJesusCollegeCambridge)