Make #'floor, #'ceiling, #'round, #'truncate conform to Common Lisp. lisp/ChangeLog addition: 2009-08-11 Aidan Kehoe <kehoea@parhasard.net> * cl-extra.el (ceiling*, floor*, round*, truncate*): Implement these in terms of the C functions; mark them as obsolete. (mod*, rem*): Use #'nth-value with the C functions, not #'nth with the CL emulation functions. man/ChangeLog addition: 2009-08-11 Aidan Kehoe <kehoea@parhasard.net> * lispref/numbers.texi (Bigfloat Basics): Correct this documentation (ignoring for the moment that it breaks off in mid-sentence). tests/ChangeLog addition: 2009-08-11 Aidan Kehoe <kehoea@parhasard.net> * automated/lisp-tests.el: Test the new Common Lisp-compatible rounding functions available in C. (generate-rounding-output): Provide a function useful for generating the data for the rounding functions tests. src/ChangeLog addition: 2009-08-11 Aidan Kehoe <kehoea@parhasard.net> * floatfns.c (ROUNDING_CONVERT, CONVERT_WITH_NUMBER_TYPES) (CONVERT_WITHOUT_NUMBER_TYPES, MAYBE_TWO_ARGS_BIGNUM) (MAYBE_ONE_ARG_BIGNUM, MAYBE_TWO_ARGS_RATIO) (MAYBE_ONE_ARG_RATIO, MAYBE_TWO_ARGS_BIGFLOAT) (MAYBE_ONE_ARG_BIGFLOAT, MAYBE_EFF, MAYBE_CHAR_OR_MARKER): New macros, used in the implementation of the rounding functions. (ceiling_two_fixnum, ceiling_two_bignum, ceiling_two_ratio) (ceiling_two_bigfloat, ceiling_one_ratio, ceiling_one_bigfloat) (ceiling_two_float, ceiling_one_float, ceiling_one_mundane_arg) (floor_two_fixnum, floor_two_bignum, floor_two_ratio) (floor_two_bigfloat, floor_one_ratio, floor_one_bigfloat) (floor_two_float, floor_one_mundane_arg, round_two_fixnum) (round_two_bignum_1, round_two_bignum, round_two_ratio) (round_one_bigfloat_1, round_two_bigfloat, round_one_ratio) (round_one_bigfloat, round_two_float, round_one_float) (round_one_mundane_arg, truncate_two_fixnum) (truncate_two_bignum, truncate_two_ratio, truncate_two_bigfloat) (truncate_one_ratio, truncate_one_bigfloat, truncate_two_float) (truncate_one_float, truncate_one_mundane_arg): New functions, used in the implementation of the rounding functions. (Fceiling, Ffloor, Fround, Ftruncate, Ffceiling, Fffloor) (Ffround, Fftruncate): Revise to fully support Common Lisp conventions. This means: -- All functions have optional DIVISOR arguments -- All functions return multiple values; see #'values -- All functions do their arithmetic with the correct number types according to the contamination rules. -- #'round and #'fround always round towards the even number in ambiguous cases. * doprnt.c (emacs_doprnt_1): * number.c (internal_coerce_number): Call Ftruncate with two arguments, not one. * floatfns.c (Ffloat): Correct this, if NUMBER is a bignum. * lisp.h: Declare Ftruncate as taking two arguments. * number.c: Provide scratch_ratio2, init it appropriately. * number.h: Make scratch_ratio2 available. * number.h (BIGFLOAT_ARITH_RETURN): * number.h (BIGFLOAT_ARITH_RETURN1): Correct these functions.

;;; Test geometry settings for frames
(defmacro check-frame-geometry (xx yy)
  `(loop for frame in (list nil (selected-frame))
	 do
	 (assert (eq (frame-property frame 'top)  ,yy))
	 (assert (eq (frame-property frame 'left) ,xx))
	 (assert (eq (frame-property frame 'top)  ,yy))
	 (assert (eq (frame-property frame 'left) ,xx))
	 (loop for plist in
	       (list (frame-properties)
		     (frame-properties nil)
		     (frame-properties (selected-frame)))
	       do
	       (assert (eq (plist-get plist 'top)  ,yy))
	       (assert (eq (plist-get plist 'left) ,xx)))))

(loop for (x y) in '((0 0) (1 1) (3 3) (9 9) (10 20) (20 40) (40 80) (100 200))
      do
      (loop for frame in (list nil (selected-frame))
	    do
	    (set-frame-properties frame `(left ,x top ,y))
	    (check-frame-geometry x y)
	    (set-frame-property frame 'top (+ y 3))
	    (check-frame-geometry x (+ y 3))
	    (set-frame-property frame 'left (+ x 3))
	    (check-frame-geometry (+ x 3) (+ y 3))))