Re: chown xxx.yyy, POSIX, and LSB

[Paul Eggert]

Jim Meyering <address@hidden> writes:

> If you do that, you should also adjust `chown --help' output
> and coreutils.texi to describe the new default behavior.

But the practice elsewhere is to not document the old-fashioned POSIX
1003.2-1992 behavior in help strings; for example, "sort --help" and
"tail --help" don't.

As I caused this problem originally (sorry!), I looked into the POSIX
spec with tweezers and a fine-toothed comb, and I spotted an escape
hatch.  Here is a simpler (and what I hope is a better) proposal: have
chown support the old syntax regardless of POSIX version of
POSIXLY_CORRECT.  This conforms, is because POSIX allows the "chown
user.group file" behavior as an upward-compatible extension, so long
as it comes into play only where a vanilla chown would report an error
(which is what the code already does).

Here is a proposed patch.

2004-02-23  Paul Eggert  <address@hidden>

        * NEWS: Document the following.
        * doc/coreutils.texi (chown invocation): Likewise.
        * lib/userspec.c: Don't include "posixver.h".
        (parse_user_spec): Fall back on USER.GROUP parsing regardless
        of POSIX version, as POSIX 1003.1-2001 allows that behavior as a
        compatible extension.  Simplify code by removing a boolean int
        that was always nonzero if a string was nonnull.

Index: NEWS
===================================================================
RCS file: /home/meyering/coreutils/cu/NEWS,v
retrieving revision 1.178
diff -p -u -r1.178 NEWS
--- NEWS        21 Feb 2004 09:26:56 -0000      1.178
+++ NEWS        23 Feb 2004 23:02:56 -0000
@@ -14,6 +14,12 @@ GNU coreutils NEWS                      
     The `|' and `&' operators now use short-circuit evaluation, e.g.,
     `expr 1 \| 1 / 0' no longer reports a division by zero.
 
+** New features
+
+  `chown user.group file' now has its traditional meaning even when
+  conforming to POSIX 1003.1-2001, so long as no user has a name
+  containing `.' that happens to equal `user.group'.
+
 
 * Major changes in release 5.2.0 (2004-02-19) [stable]
 
Index: doc/coreutils.texi
===================================================================
RCS file: /home/meyering/coreutils/cu/doc/coreutils.texi,v
retrieving revision 1.161
diff -p -u -r1.161 coreutils.texi
--- doc/coreutils.texi  22 Feb 2004 15:44:32 -0000      1.161
+++ doc/coreutils.texi  23 Feb 2004 23:05:15 -0000
@@ -7712,11 +7712,13 @@ is omitted, only the group of the files 
 @end table
 
 Some older scripts may still use @samp{.} in place of the @samp{:} separator.
address@hidden 1003.1-2001 (@pxref{Standards conformance}) does not allow that,
-but you may be able to work around the compatibility problems by setting
address@hidden in your environment.
-New scripts should avoid the use of @samp{.} because @sc{gnu} @command{chown}
-may fail if @var{owner} contains @samp{.} characters.
address@hidden 1003.1-2001 (@pxref{Standards conformance}) does not
+require support for that, but for backward compatibility @acronym{GNU}
address@hidden supports @samp{.} so long as no ambiguity results.
+New scripts should avoid the use of @samp{.} because it is not
+portable, and because it has undesirable results if the entire
address@hidden@samp{.}group} happens to identify a user whose name
+contains @samp{.}.
 
 Warning: The @command{chown} command may clear the set-user-ID or
 set-group-ID bits on some systems.  The @command{chown} command is
Index: lib/userspec.c
===================================================================
RCS file: /home/meyering/coreutils/cu/lib/userspec.c,v
retrieving revision 1.39
diff -p -u -r1.39 userspec.c
--- lib/userspec.c      3 Nov 2003 14:57:41 -0000       1.39
+++ lib/userspec.c      23 Feb 2004 21:34:10 -0000
@@ -42,7 +42,6 @@
 #endif
 
 #include "userspec.h"
-#include "posixver.h"
 #include "xalloc.h"
 #include "xstrtol.h"
 
@@ -158,7 +157,6 @@ parse_user_spec (const char *spec_arg, u
   struct group *grp;
   char *g, *u, *separator;
   char *groupname;
-  int maybe_retry = 0;
   char *dot = NULL;
 
   error_msg = NULL;
@@ -171,17 +169,14 @@ parse_user_spec (const char *spec_arg, u
   separator = strchr (spec, ':');
 
   /* If there is no colon, then see if there's a `.'.  */
-  if (separator == NULL && posix2_version () < 200112)
+  if (separator == NULL)
     {
       dot = strchr (spec, '.');
       /* If there's no colon but there is a `.', then first look up the
         whole spec, in case it's an OWNER name that includes a dot.
         If that fails, then we'll try again, but interpreting the `.'
-        as a separator.  */
-      /* FIXME: accepting `.' as the separator is contrary to POSIX.
-        someday we should drop support for this.  */
-      if (dot)
-       maybe_retry = 1;
+        as a separator.  This is a compatible extension to POSIX, since
+        the POSIX-required behavior is always tried first.  */
     }
 
  retry:
@@ -311,10 +306,10 @@ parse_user_spec (const char *spec_arg, u
        }
     }
 
-  if (error_msg && maybe_retry)
+  if (error_msg && dot)
     {
-      maybe_retry = 0;
       separator = dot;
+      dot = NULL;
       error_msg = NULL;
       goto retry;
     }