Re: incompatible open modes

Re: incompatible open modes

[Zack Brown]

Hi John,

The best place to ask is on the linux-kernel mailing list (CCed).

Good luck,
Zack

On Thu, Jul 31, 2003 at 12:09:14PM -0400, Ata, John wrote:
>    Hi Zach,
> 
>    I don't know if you're the right contact... just wondered how I go about
>    getting information as to the intent of the Linux kernel... the manpage on
>    "open" states that if a file is opened "O_RDONLY|O_TRUNC", the O_TRUNC is
>    either ignored or an error is returned.  The 2.4 kernel appears to
>    cheerfully truncate the file on open.  I wondered which behavior is
>    actually intended.
> 
>    Thanks for your time...
> 
>    Take care,
>    ------
>    John G. Ata
>    DigitalNet, LLC
>    XTS-400 Software Development
>    MailTo:John.Ata@DigitalNet.com
>    Phone:(703) 563-8092
> 
>    O_TRUNC
>                  If the file already exists and is a regular file  and  the 
>    open
>                  mode  allows  writing  (i.e.,  is O_RDWR or O_WRONLY) it
>    will be
>                  truncated to length 0.  If the file is a FIFO or terminal
>    device
>                  file,  the  O_TRUNC  flag  is  ignored.  Otherwise the
>    effect of
>                  O_TRUNC is unspecified.  (On many  Linux  versions  it 
>    will  be
>                  ignored; on other versions it will return an error.)

-- 
Zack Brown

Re: incompatible open modes

[Andries Brouwer]

> On Thu, Jul 31, 2003 at 12:09:14PM -0400, Ata, John wrote:

> >    the manpage on "open" states that if a file is opened "O_RDONLY|O_TRUNC",
> >    the O_TRUNC is either ignored or an error is returned.  The 2.4 kernel
> >    appears to cheerfully truncate the file on open.  I wondered which
> >    behavior is actually intended.
> >
> >    O_TRUNC
> >        If the file already exists and is a regular file  and  the open
> >        mode  allows  writing  (i.e.,  is O_RDWR or O_WRONLY) it will be
> >        truncated to length 0.  
> >        Otherwise the effect of O_TRUNC is unspecified.
> >        (On many Linux versions it will be ignored; on other versions
> >        it will return an error.)

This was just recently discussed, and it became clear that the parenthetical
remark only led to confusion. It has been deleted. Instead

       The (undefined) effect of O_RDONLY | O_TRUNC various among
       implementations.  On  many  systems  the  file is actually
       truncated.

has been added.

Andries

RE: incompatible open modes

[Richard B. Johnson]

On Thu, 31 Jul 2003, Ata, John wrote:

> Hi Andries,
>
> If that's what's been decided... I presume for backwards compatability,
> but it does seem rather odd though.  After all, it seems like O_RDONLY
> is supposed to safeguard someone from accidently overwriting a file.
> Otherwise why not automatically open everything read/write?  Going down
> the same path, what's next: automatically write enabling a file which
> has been openend for O_RDONLY the next time someone performs a write
> operation on it? ;-)
>
> Take care,
> John

Historically, the word "undefined" has become synonymous with
"worst possible thing" under Unix. If some operation is "undefined"
the implementor is free to low-level format your hard disk.

This is not a good thing. For instance, the MS-DOS 'open' has
defaults that are not harmful. Not so with Unix. There are no
defaults! You must be explicit. You can even create a file you
can't delete if you don't set the permissions correctly when
opening O_CREAT. Note you can even create a file called "*" and
"*.*". So, under Unix you gotta be careful. Like somebody's
.sig said; "Unix gives you enough rope to shoot yourself!"

Cheers,
Dick Johnson
Penguin : Linux version 2.4.20 on an i686 machine (797.90 BogoMips).
            Note 96.31% of all statistics are fiction.

RE: incompatible open modes

[Ata, John]

Hi Andries,

If that's what's been decided... I presume for backwards compatability,
but it does seem rather odd though.  After all, it seems like O_RDONLY
is supposed to safeguard someone from accidently overwriting a file.
Otherwise why not automatically open everything read/write?  Going down
the same path, what's next: automatically write enabling a file which
has been openend for O_RDONLY the next time someone performs a write
operation on it? ;-)

Take care,
John

-----Original Message-----
From: Andries Brouwer [mailto:aebr@win.tue.nl] 
Sent: Thursday, July 31, 2003 1:36 PM
To: Zack Brown
Cc: Ata, John; Linux Kernel Mailing List
Subject: Re: incompatible open modes


> On Thu, Jul 31, 2003 at 12:09:14PM -0400, Ata, John wrote:

> >    the manpage on "open" states that if a file is opened
"O_RDONLY|O_TRUNC",
> >    the O_TRUNC is either ignored or an error is returned.  The 2.4
kernel
> >    appears to cheerfully truncate the file on open.  I wondered
which
> >    behavior is actually intended.
> >
> >    O_TRUNC
> >        If the file already exists and is a regular file  and  the
open
> >        mode  allows  writing  (i.e.,  is O_RDWR or O_WRONLY) it will
be
> >        truncated to length 0.  
> >        Otherwise the effect of O_TRUNC is unspecified.
> >        (On many Linux versions it will be ignored; on other versions
> >        it will return an error.)

This was just recently discussed, and it became clear that the
parenthetical
remark only led to confusion. It has been deleted. Instead

       The (undefined) effect of O_RDONLY | O_TRUNC various among
       implementations.  On  many  systems  the  file is actually
       truncated.

has been added.

Andries