Re: 664 vs. touch -d now

[Paul Eggert]

address@hidden writes:

> Well at least: Ah ha, isn't "now" the current time? Bug! Muhahaha.

Thanks for the bug report.  Here is a patch.  It's not that easy to
write up a test case, due to the permissions issues, so I omitted
that.

2008-01-03  Paul Eggert  <address@hidden>

        * src/touch.c (main): Treat "-d now" as if it were absent, if
        neither -a nor -m is specified.  Problem reported by Dan Jacobson in:
        http://lists.gnu.org/archive/html/bug-coreutils/2008-01/msg00010.html

diff --git a/src/touch.c b/src/touch.c
index 2540558..205fbf6 100644
--- a/src/touch.c
+++ b/src/touch.c
@@ -368,9 +368,29 @@ main (int argc, char **argv)
     {
       if (flex_date)
        {
-         get_reldate (&newtime[0], flex_date, NULL);
+         struct timespec now;
+         gettime (&now);
+         get_reldate (&newtime[0], flex_date, &now);
          newtime[1] = newtime[0];
          date_set = true;
+
+         /* If neither -a nor -m is specified, treat "-d now" as if
+            it were absent; this lets "touch" succeed more often in
+            the presence of restrictive permissions.  */
+         if (change_times == (CH_ATIME | CH_MTIME)
+             && newtime[0].tv_sec == now.tv_sec
+             && newtime[0].tv_nsec == now.tv_nsec)
+           {
+             /* Check that it really was "-d now", and not a time
+                stamp that just happens to be the current time.  */
+             struct timespec notnow, notnow1;
+             notnow.tv_sec = now.tv_sec ^ 1;
+             notnow.tv_nsec = now.tv_nsec;
+             get_reldate (&notnow1, flex_date, &notnow);
+             if (notnow1.tv_sec == notnow.tv_sec
+                 && notnow1.tv_nsec == notnow.tv_nsec)
+               date_set = false;
+           }
        }
     }