Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!husc6!panda!genrad!decvax!tektronix!uw-beaver!tikal!bobc
From: bobc@tikal.UUCP
Newsgroups: net.lang.mod2
Subject: Re: One Pass compiler
Message-ID: <508@tikal.UUCP>
Date: Mon, 6-Oct-86 12:49:00 EDT
Article-I.D.: tikal.508
Posted: Mon Oct  6 12:49:00 1986
Date-Received: Tue, 7-Oct-86 23:36:11 EDT
References: <12243907247.41.KURKURE@Sushi.Stanford.EDU>
Reply-To: bobc@tikal.UUCP (Bob Campbell)
Organization: Teltone Corp., Kirkland, WA
Lines: 74
Keywords: FORWARD

>>	Actually, you can note the fact that whatever type the symbol has
>>    must be compatable with how it is being used, and check that against
>>    the actual declaration when you see it.  For example, if I see a
>>    function call:
>
>>  intvar := funct( 5.0, "test", 7 );
>
>>   I can note that funct must return a type that can be assigned to an int,
>>   and accept a floating point, char array, and integer argument;  then,
>>   when I see a declaration of funct, I can check against this usage, and
>>   determine whether the previous usage was correct.
>
>     What happens if the intvar is also declared forward ?
>

I feel that a one pass compiler can't easily handle forward
declarations as stated in the language definition.  I also have used
the "ETHZ" one pass compiler, and feel that the limitations that Wirth
added are reasonable.  The first limitation (and not a unreasonable
one) is that variables must be declared before they are used.  Also
TYPE must be defined before they are used (with the exception of the
normal POINTER TO undefinedtype defines).  Third is that forward
declaritions of PROCEDUREs is required using the FORWARD directive.

At one point I felt that "backpatching" could solve the problem as in:

    intvar := funct( 5.0, "test", 7 );

But since I have done some work with the one pass compiler (and have
seen the source) it turns out that there are some problems.  The
first is you can't tell what type any of the arguments are.  For
example what if the parameters were all "ARRAY OF WORD" (BYTE is
another form of the same thing on the One Pass Compiler).  For example
based on what the ETHZ One Pass Compiler does:

PROCEDURE funct(v1,v2,v3:ARRAY OF BYTE):INTEGER; FORWARD;

....
    intvar := funct(5.0, "test", 7);

Expands to the following code:

    PUSH.W	#3
    PUSH.L	ADR(Const5.0)
    PUSH.W	#3
    PUSH.L	ADR(Const"test")
    PUSH.W	#1
    PUSH.L	ADR(Const7)
    CALL	funct
    POP.W	intvar

Note that arrays are passed by address, and the called function then
copies them.  This code is much different then the expected:

    PUSH.L	#5.0
    PUSH.W	#3
    PUSH.L	ADR(Const"test")
    PUSH.W	7
    CALL	funct
    POP.W	intvar

This means that you would have to do so much backpatching as to be come
almost unreasonable, (not completely impossable but almost)  It would
seem almost as reasonable to make a two pass compiler.  I do however
feel that using variables before they are defined is unreasonable, and
I also find the "FORWARD" directive a small price to pay for a very
fast compiler.

Bob Campbell
Teltone Corporation		18520 - 66th AVE NE
P.O. Box 657			Seattle, WA 98155
Kirkland, WA 98033

{amc,dataio,fluke,hplsla,sunup,uw-beaver}!tikal!bobc