Menu
▾
▴
[Sbcl-commits] CVS: sbcl BUGS,1.380,1.381 NEWS,1.527,1.528 version.lisp-expr,1.1587,1.1588
|
From: Christophe R. <cr...@us...> - 2004-05-11 18:30:19
|
Update of /cvsroot/sbcl/sbcl In directory sc8-pr-cvs1.sourceforge.net:/tmp/cvs-serv3976 Modified Files: BUGS NEWS version.lisp-expr Log Message: 0.8.10.19: Fix floating point read/print consistency, with a slightly tidied up version of Burger and Dybvig from the one CSR posted to sbcl-devel ca. end 2004-04. ... no more recursion, yay! ... still two versions of float printing algorithms ... test case ... fix the reader too. (The original workaround was probably a CMUCLism from having :underflow traps enabled; since SBCL has never been distributed with underflow traps, we can remove the workaround). Log all remaining undisputed bugs from Bruno Haible. Index: BUGS =================================================================== RCS file: /cvsroot/sbcl/sbcl/BUGS,v retrieving revision 1.380 retrieving revision 1.381 diff -u -d -r1.380 -r1.381 --- BUGS 2 May 2004 18:12:42 -0000 1.380 +++ BUGS 11 May 2004 18:29:50 -0000 1.381 @@ -382,6 +382,20 @@ Raymond Toy comments that this is tricky on the X86 since its FPU uses 80-bit precision internally. + Bruno Haible comments: + The values are those that are expected for an IEEE double-float + arithmetic. The problem appears to be that the rounding is not + IEEE on x86 compliant: namely, values are first rounded to 64 + bits mantissa precision, then only to 53 bits mantissa + precision. This gives different results than rounding to 53 bits + mantissa precision in a single step. + + The quick "fix", to permanently change the FPU control word from + 0x037f to 0x027f, will give problems with the fdlibm code that is + used for computing transcendental functions like sinh() etc. + so maybe we need to change the FPU control word to that for Lisp + code, and adjust it to the safe 0x037f for calls to C? + 124: As of version 0.pre7.14, SBCL's implementation of MACROLET makes the entire lexical environment at the point of MACROLET available @@ -1310,38 +1324,6 @@ around the same time regarding a call to LIST on sparc with 1000 arguments) and other implementation limit constants. -310: "Floating point printing inaccuracy" - (reported by Bruno Haible sbcl-devel "print-read consistency for - floating point numbers" 2004-04-19) - (let ((x (/ -9.349640046247849d-21 -9.381494249123696d-11))) - (let ((y (read-from-string (write-to-string x :readably t)))) - (eql x y))) - should return T but, as of sbcl-0.8.9.51, returns NIL. - - That this is a bug in the printer is demonstrated by - (setq x1 (float -5496527/100000000000000000)) - (setq x2 (float -54965272/1000000000000000000)) - (integer-decode-float x1) => 15842660 -58 -1 - (integer-decode-float x2) => 15842661 -58 -1 - (prin1-to-string x1) => "-5.496527e-11" - (prin1-to-string x2) => "-5.496527e-11" ; should be different! - - Note also the following comment from src/code/print.lisp: - ;;; NOTE: When a number is to be printed in exponential format, it is - ;;; scaled in floating point. Since precision may be lost in this - ;;; process, the guaranteed accuracy properties of FLONUM-TO-STRING - ;;; are lost. The difficulty is that FLONUM-TO-STRING performs - ;;; extensive computations with integers of similar magnitude to that - ;;; of the number being printed. For large exponents, the bignums - ;;; really get out of hand. If bignum arithmetic becomes reasonably - ;;; fast and the exponent range is not too large, then it might become - ;;; attractive to handle exponential notation with the same accuracy - ;;; as non-exponential notation, using the method described in the - ;;; Steele and White paper. - - See also CSR sbcl-devel with an implementation of Berger and - Dybvig's algorithm for printing and a fix for reading. - 311: "Tokeniser not thread-safe" (see also Robert Marlow sbcl-help "Multi threaded read chucking a spak" 2004-04-19) @@ -1363,3 +1345,121 @@ 313: "source-transforms are Lisp-1" (fixed in 0.8.10.2) + +314: "LOOP :INITIALLY clauses and scope of initializers" + reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP + test suite, originally by Thomas F. Burdick. + ;; <http://www.lisp.org/HyperSpec/Body/sec_6-1-7-2.html> + ;; According to the HyperSpec 6.1.2.1.4, in for-as-equals-then, var is + ;; initialized to the result of evaluating form1. 6.1.7.2 says that + ;; initially clauses are evaluated in the loop prologue, which precedes all + ;; loop code except for the initial settings provided by with, for, or as. + (loop :for x = 0 :then (1+ x) + :for y = (1+ x) :then (ash y 1) + :for z :across #(1 3 9 27 81 243) + :for w = (+ x y z) + :initially (assert (zerop x)) :initially (assert (= 2 w)) + :until (>= w 100) :collect w) + Expected: (2 6 15 38) + Got: ERROR + +315: "no bounds check for access to displaced array" + reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP + test suite. + (locally (declare (optimize (safety 3) (speed 0))) + (let* ((x (make-array 10 :fill-pointer 4 :element-type 'character + :initial-element #\space :adjustable t)) + (y (make-array 10 :fill-pointer 4 :element-type 'character + :displaced-to x))) + (adjust-array x '(5)) + (char y 5))) + + SBCL 0.8.10 elides the bounds check somewhere along the line, and + returns #\Nul (where an error would be much preferable, since a test + of that form but with (setf (char y 5) #\Space) potentially corrupts + the heap and certainly confuses the world if that string is used by + C code. + +316: "SHIFTF and multiple values" + reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP + test suite. + (shiftf (values x y) (values y x)) + gives an error in sbcl-0.8.10. + + Parts of the explanation of SHIFTF in ANSI CL talk about multiple + store variables, and the X3J13 vote + SETF-MULTIPLE-STORE-VARIABLES:ALLOW also says that SHIFTF should + support multiple value places. + +317: "FORMAT of floating point numbers" + reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP + test suite. + (format nil "~1F" 10) => "0." ; "10." expected + (format nil "~0F" 10) => "0." ; "10." expected + (format nil "~2F" 1234567.1) => "1000000." ; "1234567." expected + it would be nice if whatever fixed this also untangled the two + competing implementations of floating point printing (Steele and + White, and Burger and Dybvig) present in src/code/print.lisp + +318: "stack overflow in compiler warning with redefined class" + reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP + test suite. + (setq *print-pretty* nil) + (defstruct foo a) + (setf (find-class 'foo) nil) + (defstruct foo slot-1) + gives + ...#<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTURE-CLASSOID #<SB-KERNEL:STRUCTUREControl stack guard page temporarily disabled: proceed with caution + (it's not really clear what it should give: is (SETF FIND-CLASS) + meant to be enough to delete structure classes from the system? + Giving a stack overflow is definitely suboptimal, though.) + +319: "backquote with comma inside array" + reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP + test suite. + (read-from-string "`#1A(1 2 ,(+ 2 2) 4)") + gives + #(1 2 ((SB-IMPL::|,|) + 2 2) 4) + which probably isn't intentional. + +320: "shared to local slot in class redefinition" + reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP + test suite. + ;; Shared slot becomes local. + ;; 4.3.6.1.: "The value of a slot that is specified as shared in + ;; the old class and as local in the new class is retained." + (multiple-value-bind (value condition) + (ignore-errors + (defclass foo85a () + ((size :initarg :size :initform 1 :allocation :class))) + (defclass foo85b (foo85a) ()) + (setq i (make-instance 'foo85b)) + (defclass foo85a () ((size :initarg :size :initform 2) (other))) + (slot-value i 'size)) + (list value (type-of condition))) + should return (1 NULL) but returns (2 NULL) in sbcl-0.8.10. See + ensuing discussion sbcl-devel for how to deal with this. + +321: "DEFINE-METHOD-COMBINATION lambda list parsing" + reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP + test suite. + (define-method-combination w-args () + ((method-list *)) + (:arguments arg1 arg2 &aux (extra :extra)) + `(progn ,@(mapcar (lambda (method) `(call-method ,method)) method-list))) + gives a (caught) compile-time error, which can be exposed by + (defgeneric mc-test-w-args (p1 p2 s) + (:method-combination w-args) + (:method ((p1 number) (p2 t) s) + (vector-push-extend (list 'number p1 p2) s)) + (:method ((p1 string) (p2 t) s) + (vector-push-extend (list 'string p1 p2) s)) + (:method ((p1 t) (p2 t) s) (vector-push-extend (list t p1 p2) s))) + +322: "DEFSTRUCT :TYPE LIST predicate and improper lists" + reported by Bruno Haible sbcl-devel "various SBCL bugs" from CLISP + test suite. + (defstruct (a (:type list) (:initial-offset 5) :named)) + (defstruct (b (:type list) (:initial-offset 2) :named (:include a))) + (b-p (list* nil nil nil nil nil 'foo73 nil 'tail)) + gives an error in sbcl-0.8.10 Index: NEWS =================================================================== RCS file: /cvsroot/sbcl/sbcl/NEWS,v retrieving revision 1.527 retrieving revision 1.528 diff -u -d -r1.527 -r1.528 --- NEWS 11 May 2004 07:31:36 -0000 1.527 +++ NEWS 11 May 2004 18:29:50 -0000 1.528 @@ -2432,7 +2432,10 @@ * fixed some bugs revealed by Paul Dietz' test suite: ** FILE-POSITION works as specified on BROADCAST-STREAMs. ** CAST optimizer forgot to flush argument derived type. - + ** print/read consistency on floats is now orders of magnitude + more likely. (thanks also to Bruno Haible for a similar report + and discussions) + planned incompatible changes in 0.8.x: * (not done yet, but planned:) When the profiling interface settles down, it might impact TRACE. They both encapsulate functions, and Index: version.lisp-expr =================================================================== RCS file: /cvsroot/sbcl/sbcl/version.lisp-expr,v retrieving revision 1.1587 retrieving revision 1.1588 diff -u -d -r1.1587 -r1.1588 --- version.lisp-expr 11 May 2004 07:31:37 -0000 1.1587 +++ version.lisp-expr 11 May 2004 18:29:51 -0000 1.1588 @@ -17,4 +17,4 @@ ;;; checkins which aren't released. (And occasionally for internal ;;; versions, especially for internal versions off the main CVS ;;; branch, it gets hairier, e.g. "0.pre7.14.flaky4.13".) -"0.8.10.18" +"0.8.10.19" |