c. the examples in CLHS 7.6.5.1 (regarding generic function lambda
lists and &KEY arguments) do not signal errors when they should.
-192: "constant-folding DATA-VECTOR-REF"
- (from Einar Floystad Dorum sbcl-devel 14 Aug 2002)
- Typing this
- (defun test ()
- (funcall
- (LAMBDA (TEXT I L )
- (LABELS ((G908 (I)
- (LET ((INDEX
- (OR
- (IF (= I L)
- NIL
- (LET ((S TEXT)
- (E (ELT TEXT I)))
- (DECLARE (IGNORABLE S E))
- (WHEN (EQL #\a E)
- (G909 (1+ I))))))))
- INDEX))
- (G909 (I)
- (OR
- (IF (= I L)
- NIL
- (LET ((S TEXT)
- (E (ELT TEXT I)))
- (DECLARE (IGNORABLE S E))
- (WHEN (EQL #\b E) (G910 (1+ I)))))))
- (G910 (I)
- (LET ((INDEX
- (OR
- (IF NIL
- NIL
- (LET ((S TEXT))
- (DECLARE (IGNORABLE S))
- (WHEN T I))))))
- INDEX)))
- (G908 I))) "abcdefg" 0 (length "abcdefg")))
- at the command prompt in sbcl-0.7.6.23 gives
- (BUG "full call to SB-KERNEL:DATA-VECTOR-REF").
-
DEFUNCT CATEGORIES OF BUGS
IR1-#:
These labels were used for bugs related to the old IR1 interpreter.
* Bug fix: changed encoding of PCL's internal MAKE-INSTANCE
functions so that EXPORTing the name of the class doesn't cause
MAKE-INSTANCE functions from earlier DEFCLASSes to get lost (thanks
- to Antonio Martinez)
+ to Antonio Martinez for reporting this)
+ * Bug 192 fixed: The internal primitive DATA-VECTOR-REF can now be
+ constant-folded without failing an assertion. (thanks to Einar
+ Floystad Dorum for reporting this)
* Minor incompatible change: COMPILE-FILE-PATHNAME now merges its
OUTPUT-FILE argument with its INPUT-FILE argument, resulting in
behaviour analogous to RENAME-FILE. This puts its behaviour more
# "lisp -noinit -batch"
# to use an existing CMU CL binary as a cross-compilation host
# when you have weird things in your .cmucl-init file
-# Someday any sufficiently ANSI Common Lisp, perhaps CLISP and/or
-# OpenMCL should work
-# "clisp"
-# "??"
-# but not yet as of sbcl-0.7.4. (There are still some weird dependencies
-# on idiosyncrasies of the way CMU CL implements Common Lisp.)
+# "openmcl --batch"
+# to use an OpenMCL binary as a cross-compilation host
#
# FIXME: Make a more sophisticated command line parser, probably
# accepting "sh make.sh --xc-host foolisp" instead of the
index))))
*specialized-array-element-types*))))
+;;; (Ordinary DATA-VECTOR-REF usage compiles into a vop, but
+;;; DATA-VECTOR-REF is also FOLDABLE, and this ordinary function
+;;; definition is needed for the compiler to use in constant folding.)
+(defun data-vector-ref (array index)
+ (hairy-data-vector-ref array index))
+
(defun hairy-data-vector-set (array index new-value)
(with-array-data ((vector array) (index index) (end))
(declare (ignore end))
;;;; code to detect whether a package has changed
+;;;;
+;;;; This is really old code which was most useful when first
+;;;; bootstrapping SBCL when only CMU CL was available as an XC host.
+;;;; Its main purpose was to check that legacy code like DEFMACRO
+;;;; DOLIST and DEFUN IR1-OPTIMIZE-UNTIL-DONE was all correctly
+;;;; converted from code which mutated the XC host into code which
+;;;; built things for the target.
+;;;;
+;;;; These days, things like DEFUN IR1-OPTIMIZE-UNTIL-DONE can't very
+;;;; well be mutating the cross-compiler host because we can build
+;;;; successfully under OpenMCL, which shouldn't have the same
+;;;; packages or symbols. So we don't need to worry very much about
+;;;; modifying the XC host's private packages. However, it's still
+;;;; conceivable that something affecting the XC host's CL package
+;;;; (maybe DEFMACRO DOLIST?) could be written in such a way that it
+;;;; would silently compile under SBCL, CMU CL, and even OpenMCL, and
+;;;; still be fundamentally wrong. Since it'd be good to prevent such
+;;;; modifications of the XC host's CL package, this code is still
+;;;; retained despite being a little strange.
;;;; This software is part of the SBCL system. See the README file for
;;;; more information.
(defun load-function-generator (test gensyms generator generator-lambda system)
(store-fgen (make-fgen test gensyms generator generator-lambda system)))
-
(logand a0 a10)
;; a call to prevent the other arguments from being optimized away
(logand a1 a2 a3 a4 a5 a6 a7 a8 a9)))
+
+;;; bug 192, reported by Einar Floystad Dorum: Compiling this in 0.7.6
+;;; caused the compiler to try to constant-fold DATA-VECTOR-REF, which
+;;; is OK, except that there was no non-VOP definition of
+;;; DATA-VECTOR-REF, so it would fail.
+(defun bug192 ()
+ (funcall
+ (LAMBDA (TEXT I L )
+ (LABELS ((G908 (I)
+ (LET ((INDEX
+ (OR
+ (IF (= I L)
+ NIL
+ (LET ((S TEXT)
+ (E (ELT TEXT I)))
+ (DECLARE (IGNORABLE S E))
+ (WHEN (EQL #\a E)
+ (G909 (1+ I))))))))
+ INDEX))
+ (G909 (I)
+ (OR
+ (IF (= I L)
+ NIL
+ (LET ((S TEXT)
+ (E (ELT TEXT I)))
+ (DECLARE (IGNORABLE S E))
+ (WHEN (EQL #\b E) (G910 (1+ I)))))))
+ (G910 (I)
+ (LET ((INDEX
+ (OR
+ (IF NIL
+ NIL
+ (LET ((S TEXT))
+ (DECLARE (IGNORABLE S))
+ (WHEN T I))))))
+ INDEX)))
+ (G908 I))) "abcdefg" 0 (length "abcdefg")))
\f
;;; BUG 48a. and b. (symbol-macrolet handling), fixed by Eric Marsden
;;; and Raymond Toy for CMUCL, fix ported for sbcl-0.7.6.18.
;;; for internal versions, especially for internal versions off the
;;; main CVS branch, it gets hairier, e.g. "0.pre7.14.flaky4.13".)
-"0.7.6.25"
+"0.7.6.26"
projects / sbcl.git / commitdiff