--* From smwatt@spadserv.watson.ibm.com  Thu Oct 21 07:24:06 1993
--* Received: from spadserv.watson.ibm.com by radical.watson.ibm.com (AIX 3.2/UCB 5.64/900524)
--*           id AA25179; Thu, 21 Oct 1993 07:24:06 -0400
--* Received: by spadserv.watson.ibm.com (AIX 3.2/UCB 5.64/900524)
--*           id AA20514; Thu, 21 Oct 1993 07:32:05 -0400
--* Date: Thu, 21 Oct 1993 07:32:05 -0400
--* From: smwatt@spadserv.watson.ibm.com (Stephen Watt)
--* X-External-Networks: yes
--* Message-Id: <9310211132.AA20514@spadserv.watson.ibm.com>
--* To: asbugs@radical.watson.ibm.com, bronstei@inf.ethz.ch, peteb@yorktown,
--*         santas@inf.ethz.ch, smwatt@watson.ibm.com
--* Subject: Multiple value assignment bug

--@ Fixed  by:  PI   Fri Jan 13 09:07:50 EST 1995 
--@ Tested by:  none 
--@ Summary:    Fixed the code generation for multiple assignments. Now prints '1,2,2,4'. 


> From: Philip Santas <santas@inf.ethz.ch>
> Date: Wed, 20 Oct 93 23:57:22 +0100
> To: bronstei@inf.ethz.ch, smwatt@watson.ibm.com
> Subject: Re: A# code
> 
> 
> Does the code
> (a, b) := (b, ncrem(a, b))
> have the same semantics as:
> 
> c:=a
> a:=b
> b:=ncrem(c,b)
> ?
> 
> Are tuples evaluted from left to right?
> What if we had:
> 
> (a,b) := ((a,b):=ncrem(a, b);b, ncrem(a, b))
> ?
> a
> I ask, because the code:
> 
> -------------------------------------------------------
> (a:SI, b:SI) := (1,2)
> 
> print(a)()(b)()
> 
> (a,b) := ({{(a,b):=(a+1, b+1)}; a}, {{(a,b):=(a+1, b+1)}; b})
> print(a)()(b)()
> ------------------------------------------------------
> returns:
> 
> 1,2,3,4
> 
> where I was initially expecting it to return:
> 1,2,2,4
> 
> Philip

The order of evaluation in multiple values is implementation dependent.
I.e. it could be left-to-right or right-to-left or some other order.
In a multiple value assignment, all rhs are evaluated before any lhs
are updated.

Nevertheless, your code does illustrate a different bug:  
It should print either  1,2,2,4  or  1,2,3,3  but not 1,2,3,4.