guile-commits
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Guile-commits] 10/55: Fix 'round-ash' of negative integers by huge righ


From: Andy Wingo
Subject: [Guile-commits] 10/55: Fix 'round-ash' of negative integers by huge right shift counts.
Date: Thu, 23 May 2019 11:52:37 -0400 (EDT)

wingo pushed a commit to branch master
in repository guile.

commit e6100f64bbab1b2bac085c0ea5250163197df4a9
Author: Mark H Weaver <address@hidden>
Date:   Sun Oct 14 05:24:14 2018 -0400

    Fix 'round-ash' of negative integers by huge right shift counts.
    
    This is a followup to commit 011aec7e240ef987931548d90c53e6692c85d01c.
    
    When rounding, right shifting a negative integer by a huge shift count
    results in 0, not -1.
    
    * libguile/numbers.c: Add top-level 'verify' to ensure that the
    assumptions in 'scm_ash' and 'scm_round_ash' are valid.
    (scm_round_ash): In the case that handles huge right shifts, require
    that the shift count _exceeds_ the integer length, and return 0 instead
    of -1.
    * test-suite/tests/numbers.test: Adjust tests accordingly.
---
 libguile/numbers.c            | 16 ++++++++++------
 test-suite/tests/numbers.test | 10 +++++-----
 2 files changed, 15 insertions(+), 11 deletions(-)

diff --git a/libguile/numbers.c b/libguile/numbers.c
index 67295f9..21dcdb6 100644
--- a/libguile/numbers.c
+++ b/libguile/numbers.c
@@ -91,6 +91,11 @@ verify (FLT_RADIX == 2);
 /* Make sure that scm_t_inum fits within a SCM value.  */
 verify (sizeof (scm_t_inum) <= sizeof (scm_t_bits));
 
+/* Several functions below assume that fixnums fit within a long, and
+   furthermore that there is some headroom to spare for other operations
+   without overflowing. */
+verify (SCM_I_FIXNUM_BIT <= SCM_LONG_BIT - 2);
+
 #define scm_from_inum(x) (scm_from_signed_integer (x))
 
 /* Test an inum to see if it can be converted to a double without loss
@@ -5127,12 +5132,11 @@ SCM_DEFINE (scm_round_ash, "round-ash", 2, 0, 0,
         bits_to_shift = SCM_I_INUM (count);
       else if (scm_is_signed_integer (count, LONG_MIN, LONG_MAX))
         bits_to_shift = scm_to_long (count);
-      else if (scm_is_false (scm_positive_p (scm_sum (scm_integer_length (n),
-                                                      count))))
-        /* Huge right shift that eliminates all but the sign bit */
-        return scm_is_false (scm_negative_p (n))
-          ? SCM_INUM0 : SCM_I_MAKINUM (-1);
-      else if (scm_is_true (scm_zero_p (n)))
+      else if (scm_is_true (scm_negative_p (scm_sum (scm_integer_length (n),
+                                                     count)))
+               || scm_is_true (scm_zero_p (n)))
+        /* If N is zero, or the right shift count exceeds the integer
+           length, the result is zero. */
         return SCM_INUM0;
       else
         scm_num_overflow ("round-ash");
diff --git a/test-suite/tests/numbers.test b/test-suite/tests/numbers.test
index 4e0bc82..8cecb06 100644
--- a/test-suite/tests/numbers.test
+++ b/test-suite/tests/numbers.test
@@ -5377,7 +5377,7 @@
 ;;;
 
 (let ()
-  (define (test-ash-variant name ash-variant round-variant)
+  (define (test-ash-variant name ash-variant round-variant rounded?)
     (with-test-prefix name
       (define (test n count)
         (pass-if (list n count)
@@ -5434,17 +5434,17 @@
           0
         (ash-variant 123 (- (expt 2 1000))))
       (pass-if-equal "Huge right shift of negative integer"
-          -1
+          (if rounded? 0 -1)
         (ash-variant -123 (- (expt 2 1000))))
       (pass-if-equal "Huge right shift of -1"
-          -1
+          (if rounded? 0 -1)
         (ash-variant -1 (- (expt 2 1000))))
       (pass-if-exception "Huge left shift of non-zero => numerical overflow"
           exception:numerical-overflow
         (ash-variant 123 (expt 2 1000)))))
 
-  (test-ash-variant       'ash       ash floor)
-  (test-ash-variant 'round-ash round-ash round))
+  (test-ash-variant       'ash       ash floor #f)
+  (test-ash-variant 'round-ash round-ash round #t))
 
 ;;;
 ;;; regressions



reply via email to

[Prev in Thread] Current Thread [Next in Thread]