Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sdd.hp.com!think.com!spool.mu.edu!cs.umn.edu!talon.UCS.ORST.EDU!usenet!ogicse!intelhf!ichips!iwarp.intel.com!inews!hopi!bhoughto
From: bhoughto@hopi.intel.com (Blair P. Houghton)
Newsgroups: comp.std.c
Subject: Re: tmpfile Question
Keywords: tmpfile, conforming
Message-ID: <3485@inews.intel.com>
Date: 27 Mar 91 06:29:08 GMT
Article-I.D.: inews.3485
References: <2537.03.91@drdhh.hanse.de>
Sender: news@inews.intel.com
Organization: Intel Corp, Chandler, AZ
Lines: 30

In article <2537.03.91@drdhh.hanse.de> bjoern@drdhh.hanse.de writes:
>My question is: 
>May a conforming program call tmpfile() more than once ?

You can call it, but if you get NULL as the return value
or end up clobbering the same file, the implementation
still conforms...

(Btw, mode "wb+" truncates the file to 0 length then
allows fseek()/rewind() operations on it; the 'b'
indicates that when opened the access pointer can be
positioned past the end of the data due to null padding,
but if the file is truncated to 0 length... ? )

>The MW Book doesn't say anything about it and my clib
>always creates the same file (brain-damaged at all). 

This is conformant behavior, nonetheless.

>I would like to have a definitive statement before 
>I go and complain to my vendor.

Complain, anyway.  They weren't prevented from having it
open uniquely-named files, but there's nothing there
that says it should.  Better yet, tell them how to write
it and they might just do it that way.

				--Blair
				  "...for the next 'release'...in six
				   months...for Eastern Borneo only..."