From: Ian D Romanick <idr AT cs DOT pdx DOT edu>
Message-Id: <199612051809.KAA12682@xavier.cs.pdx.edu>
Subject: Re: v2 funnies
To: tehmul%NIITPUN AT iris DOT ernet DOT in
Date: Thu, 5 Dec 1996 10:09:10 -0800 (PST)
Cc: djgpp AT delorie DOT com
In-Reply-To: <9E6C083001E72A00@iris.ernet.in> from "tehmul%NIITPUN@iris.ernet.in" at Dec 5, 96 09:34:00 pm
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

> >If I call open("filename",mode) with two args to create a new file and write
> >to it, even if I use mode O_RDWR, the file is created with its DOS readonly
> >attribute set so it can't be altered or deleted afterwards (unless I state
> >the permission required, as open("filename",mode,S_IWUSR)). This is
> >unexpected and annoying, and did not happen in v1. Why is this?
> 
> 
> I thought that ANSI behaviour specifies that whenever you create a file, the
> third parameter to open (the permission) is _mandatory_, not optional.  So
> not giving the third parameter while creating the file might be a mistake in
> the first place. (Correct me if I'm wrong guys, I haven't looked at any ref
> manual for a loong time.)

Well, ANSI doesn't specify *anything* for open: it's not part of the
standard.  But looking at the man page on a Solaris (POSIX) sytem, it
shows the prototype as:

     int open(const char *path, int oflag, /* mode_t mode */ ...);

It also only mentions the use of the mode field if O_CREAT is set in the
oflag field.  So it seems that this is perfectly valid usage.

-- 
The words I live by:
  "That which does not kill us makes us stronger."
  "There is no knowledge that is not power."         Knowledge and power at:
  "Without struggle, there is no progress."          http://www.cs.pdx.edu/~idr