Compare fixnums and floats accurately in value<

Make `value<` compare fixnums and floats by value, as `<` does, instead
of coercing the fixnum to a float first which is what C would do.
This matters when the fixnum cannot be represented as a float. For
example, C would evaluate

  72057594037927935 < 72057594037927936.0

to false since the operands are converted to the same floating-point
number.

* src/fns.c (fixnum_float_cmp): New.
(value_cmp): Use it.
* test/src/fns-tests.el (fns-value<-ordered, fns-value<-unordered):
New test cases.

1 files changed, 10 insertions, 0 deletions

diff --git a/test/src/fns-tests.el b/test/src/fns-tests.el
index ca5b10db705..8d39a1a56b8 100644
--- a/test/src/fns-tests.el
+++ b/test/src/fns-tests.el
@@ -1606,6 +1606,13 @@
                    (1.5 . 1.6) (-1.3 . -1.2) (-13.0 . 12.0)
                    ;; floats/fixnums
                    (1 . 1.1) (1.9 . 2) (-2.0 . 1) (-2 . 1.0)
+                   ;; fixnums that can't be represented as floats
+                   (72057594037927935 . 72057594037927936.0)
+                   (72057594037927936.0 . 72057594037927937)
+                   (-72057594037927936.0 . -72057594037927935)
+                   (-72057594037927937 . -72057594037927936.0)
+                   (2305843009213693951 . 2305843009213693952.0)
+
                    ;; floats/bignums
                    (,big . ,(float (* 2 big))) (,(float big) . ,(* 2 big))
                    ;; symbols
@@ -1683,6 +1690,9 @@
                  ;; numbers
                  (0 . 0.0) (0 . -0.0) (0.0 . -0.0)
 
+                 (72057594037927936 . 72057594037927936.0)
+                 (1 . 0.0e+NaN)
+
                  ;; symbols
                  (a . #:a)