Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: notesfiles
Path: utzoo!watmath!clyde!burl!ulysses!mhuxl!ihnp4!zehntel!hplabs!hp-pcd!hp-dcd!donn
From: donn@hp-dcd.UUCP (donn)
Newsgroups: net.lang
Subject: Re: Bad language constructs
Message-ID: <24200007@hp-dcd.UUCP>
Date: Tue, 10-Jul-84 14:29:00 EDT
Article-I.D.: hp-dcd.24200007
Posted: Tue Jul 10 14:29:00 1984
Date-Received: Mon, 16-Jul-84 07:37:29 EDT
References: <2877@ecsvax.UUCP>
Organization: Hewlett-Packard - Fort Collins, CO
Lines: 43
Nf-ID: #R:ecsvax:-287700:hp-dcd:24200007:000:1395
Nf-From: hp-dcd!donn    Jul 13 10:29:00 1984

From what I remember of COBOL, there is a consequence of the ALTER statement
in the langauge that may be worse than the ALTER itself.  That's the semantics
of PERFORM, the sort-of-subroutine call.

It permits specifying the beginning and end of the subroutine: 

	PERFORM a THRU b.

Thus you can create the following mess (where the arrows indicate the range
of two perform statements):

	A.
		......   <----
	B.		     |
		......       |   <----
	C.		     |       |
		......   <----       |
	D.			     |
		......		 <----
	E.
		......

Since PERFORM A THRU B appears to be equivalent to 

	<implicitly place paragraph name B-END at the end of B, and
	paragraph name HERE after the perform clause>
	ALTER B-END TO PROCEED TO HERE.
	GOTO A.
     HERE.
	ALTER B-END TO PROCEED TO C.    (to fix it all up for next time.)

it is possible to have BOTH the ranges above alive simultaneousy;  it
would be legal to have paragraph A contain a PERFORM B THRU D, as long
as control flow did not drop thru the end of C during the inner perform.

Some of the newer standards may limit this, but the standards and 
implementations of the late 60's and early 70's (I date myself) appeared
to permit it.   Did I do it?  I simply couldn't bring myself to it, it
was TOO horrible.  Did I see stuff like that (or worse) used?  Repeatedly;
and they were proud of how clever they were!

Donn Terry
HP Ft. Collins Co.