Re: Coreutils-gotchas

[Assaf Gordon]

On 12/16/2015 01:52 PM, Pádraig Brady wrote:
> On 16/12/15 18:45, Assaf Gordon wrote:
> > The recent "touch -d yesterday" might be a 'gotcha'
> > ( http://lists.gnu.org/archive/html/bug-coreutils/2015-12/msg00041.html ) .

<...>

> That's a Linux gotcha if anything.
> I fear the page will get too noisy/large if including such system policies.

I agree that we should not focus on system-specific policies, but allow me to 
present two counter points:

1. The '-d now' usage is a feature by gnu coreutils which specifically passes 
NULL to the syscall (instead of a non-null timespec containing the current 
time) - as such isn't instinctively clear to the user that there's a different 
mechanism in place (and the comment in touch.c:157 indicates that this is a 
special case, a feature which doesn't exist in other 'touch' implementations). 
And it passes NULL to the syscall regardless of which operating system is used.

2. The "NULL = current-time" feature seems to have the same relaxed permission 
requirements on other common operating systems, and in POSIX:

From man page of FreeBSD,OpenBSD futimens(2) and Mac OS X futimes(2):
===
  If times is NULL, the access and modification times are set to the cur-
  rent time.  The caller must be the owner of the file, have permission to
  write the file, or be the super-user.

  If times is non-NULL, [...] The caller must be the owner of the file or be
  the super-user.
===


From man page of OpenSolaris 5.11 futimens(2) and POSIX ( 
http://pubs.opengroup.org/onlinepubs/9699919799/functions/futimens.html ):
===
     Only a process with the effective user ID equal to the  user
     ID  of  the  file, or with write access to the file, or with
     appropriate privileges may  use  futimens()  or  utimensat()
     with  a  null  pointer  as  the  times argument [...]

     Only  a process  with  the effective user ID equal to the user ID of
     the file or with appropriate privileges may  use  futimens()
     or  utimensat() with a non-null times argument [...]
===

So one could say that "time=NULL" requiring only write-permission while 
"time!=NULL" requires ownership is not linux-specific,
and because GNU coreutils' 'touch' is one of the few user-land utilities that 
exposes this ability,
it might be useful to mention it somewhere.


just my 2 cents,
 - assaf