From mboxrd@z Thu Jan  1 00:00:00 1970
From: Ian Lance Taylor <ian@zembu.com>
To: Franz.Sirl-kernel@lauterbach.com
Cc: gas2@sourceware.cygnus.com
Subject: Re: strip looses original file ownership and file permissions
Date: Thu, 06 May 1999 11:03:00 -0000
Message-id: <19990506180307.24286.qmail@daffy.airs.com>
References: <4.2.0.37.19990506133808.0363a9f0@mail.lauterbach.com> <4.2.0.37.19990506133808.0363a9f0@mail.lauterbach.com> <4.2.0.37.19990506183739.0372b0e0@mail.lauterbach.com> <4.2.0.37.19990506183739.0372b0e0@mail.lauterbach.com>
X-SW-Source: 1999/msg00091.html

   Date: Thu, 06 May 1999 18:54:16 +0200
   From: Franz Sirl <Franz.Sirl-kernel@lauterbach.com>

   At 18:27 06.05.99 , Ian Lance Taylor wrote:
   >    Date: Thu, 06 May 1999 13:44:47 +0200
   >    From: Franz Sirl <Franz.Sirl-kernel@lauterbach.com>
   >
   >    >I just verified that strip out of gas-990418 looses original 
   > ownership and
   >    >permissions of a file.
   >    >This is on glibc-2.1.1pre2, Linux-2.2.6 (PPC).
   >    >
   >    >Is this platform specific or does anybody else notice this?
   >
   >    After a quick browse through the source I came up with the following
   >    untested patch. Does it look right?
   >
   >It doesn't look right to me.
   >
   >We need to rename the file FROM to TO.  In the normal case of strip,
   >FROM is a temporary file, and TO is the original file which we are
   >stripping.  However, this function is also used in other cases.
   >
   >If TO does not exist, we should just use rename.  This is not the
   >normal case of strip, but it happens in other cases.  Your patch
   >breaks that.  That seems to be only significant change in your patch.
   >Perhaps I am missing something.
   >
   >I think the only way to reliably preserve ownership is to avoid using
   >rename.  Perhaps the code should be changed to call simple_copy when
   >the owner of the file differs from the effective uid.

   I think my patch produces the behavior that the comments are suggesting. 
   The old code was bogus anyway, cause it tried to chmod/chown "to" after 
   rename with values from a non-existing (exists != 0, this is really a 
   bad-named variable, it's named the opposite of it's meaning) file.

You're right.  I applied the appended patch, which I think should fix
the problem.

Ian

Index: rename.c
===================================================================
RCS file: /cvs/binutils/binutils/binutils/rename.c,v
retrieving revision 1.1.1.1
diff -u -r1.1.1.1 rename.c
--- rename.c	1999/05/03 07:29:10	1.1.1.1
+++ rename.c	1999/05/06 18:02:05
@@ -139,17 +139,17 @@
      const char *to;
      int preserve_dates;
 {
-  int exists;
+  boolean exists;
   struct stat s;
   int ret = 0;
 
-  exists = lstat (to, &s);
+  exists = lstat (to, &s) == 0;
 
 #if defined (_WIN32) && !defined (__CYGWIN32__)
   /* Win32, unlike unix, will not erase `to' in `rename(from, to)' but
      fail instead.  Also, chown is not present.  */
 
-  if (exists == 0)
+  if (exists)
     remove (to);
 
   ret = rename (from, to);
@@ -163,7 +163,7 @@
 #else
   /* Use rename only if TO is not a symbolic link and has
      only one hard link.  */
-  if (exists < 0 || (!S_ISLNK (s.st_mode) && s.st_nlink == 1))
+  if (! exists || (!S_ISLNK (s.st_mode) && s.st_nlink == 1))
     {
       ret = rename (from, to);
       if (ret == 0)