NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/lang/fortran/2791

Text File | 1992-07-21 | 1.6 KB | 39 lines

Newsgroups: comp.lang.fortran
Path: sparky!uunet!mcsun!dxcern!dscomsa.desy.de!dscomsa!ohl
From: [e-mail] (Thorsten Ohl)
Subject: Re: recursive I/O
In-Reply-To: [e-mail]'s message of Tue, 21 Jul 1992 22:31:57 GMT
Message-ID: <[e-mail]>
Sender: [e-mail] (USENET News System)
Nntp-Posting-Host: ips103.desy.de
Organization: Deutsches Elektronen-Synchrotron, Hamburg, Germany.
References: <[e-mail]>
Date: Wed, 22 Jul 1992 01:55:44 GMT
Lines: 25

>>>>> On Tue, 21 Jul 1992 22:31:57 GMT, [e-mail] (Thomas Koenig) said:

Thomas> Is the following [ = recursive I/O ] legal F77?

No!

Thomas> It works on the UNIX box I'm on at the moment, but I can
Thomas> imagine trouble with architectures which do not use a stack
Thomas> for argument passing, for example /370.

It will certainly fail on S/370 running MVS, but I seem to remember to
have seen a runtime error also on a RS/6000 running AIX 3.1.  The
problem is not only the missing stack, but also that several FORTRAN
I/O subsystems are _not_ reentrant (they seem to use statically
allocated buffers and other nasty things).

I try to avoid function calls in I/O lists.   Yes, that costs
additional temporaries, but in large codes it is easy to overlook that
the (n+1)-th level of a nested function call might call the I/O
subsystem.

--
////// Thorsten Ohl, DESY (-T-), Notkestr. 85, W-2000 Hamburg 52, F.R.G.
///////// net: [e-mail], [e-mail]
//////////////////////// voice: [phone],  fax: [phone]