inconsistent definition of "sort" (and what about stability?)

I've pushed the following changes to the implementation and documentation
of sort in an attempt to address the questions raised a few days ago.
Hope this helps. If further discussion is needed, well, have at it.

best

Robert Dodier

PS.
diff --git a/doc/info/Lists.texi b/doc/info/Lists.texi
index 861f280..81d25bc 100644
--- a/doc/info/Lists.texi
+++ b/doc/info/Lists.texi
@@ -889,7 +889,7 @@ See @mref{first} for more details.
 @deffnx {Function} sort (@var{L})

 Sorts a list @var{L} according to a predicate @code{P} of two arguments, such
-that @code{@var{P} (@var{L}[k], @var{L}[k + 1])} is @code{true} for any two
+that @code{@var{P}(@var{L}[k + 1], @var{L}[k])} is @code{false} for any two
 successive elements.  The predicate may be specified as the name of a function
 or binary infix operator, or as a @code{lambda} expression.  If specified as
 the name of an operator, the name is enclosed in "double quotes".
@@ -900,11 +900,16 @@ the elements of @var{L}.
 @c DUNNO IF WE NEED TO GO INTO THE IMPLICATIONS OF SHALLOW COPY HERE ...

 @c MIGHT CONSIDER A REF FOR TOTAL ORDER HERE
-If the predicate @var{P} is not a total order on the elements of @var{L}, then
- at code{sort} might run to completion without error, but the result is undefined.
+It is assumed the predicate @var{P} is a strict total order on the
elements of @var{L}.
+If not, @code{sort} might run to completion without error, but the
result is undefined.
 @code{sort} complains if the predicate evaluates to something other than
 @code{true} or @code{false}.

+ at code{sort} is a stable sort:
+if two elements @var{x} and @var{y} are equivalent in the sense that
+ at code{@var{P}(@var{x}, @var{y})} and @code{@var{P}(@var{y}, @var{x})}
are both @code{false},
+then the relative order of @var{x} and @var{y} in @var{L} is
preserved by @code{sort}.
+
 @code{sort (@var{L})} is equivalent to @code{sort (@var{L}, orderlessp)}.
 That is, the default sorting order is ascending, as determined by
 @mrefdot{orderlessp}  All Maxima atoms and expressions are comparable under
diff --git a/src/mstuff.lisp b/src/mstuff.lisp
index 2dbba30..82b4184 100644
--- a/src/mstuff.lisp
+++ b/src/mstuff.lisp
@@ -21,7 +21,7 @@
 	  (mfunction1 (setq bfun (getopr f))))
     (when (member bfun '(lessthan great) :test #'eq)
       (setq llist (mapcar #'ratdisrep llist)))
-    (cons '(mlist) (sort llist comparfun))))
+    (cons '(mlist) (stable-sort llist comparfun))))

 ;; cmulisp does not like the closure version.  Clisp insists on the
 ;; closure version.  Gcl likes either...  For the moment we will