c/7696: Spurious shift warning

[Christian Ehrhardt]

The following reply was made to PR c/7696; it has been noted by GNATS.

From: "Christian Ehrhardt" <ehrhardt@mathematik.uni-ulm.de>
To: gcc-gnats@gcc.gnu.org, schwab@suse.de, gcc-bugs@gcc.gnu.org,
  nobody@gcc.gnu.org, gcc-prs@gcc.gnu.org
Cc:  
Subject: Re: c/7696: Spurious shift warning
Date: Sun, 6 Apr 2003 14:17:56 +0200

 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=7696
 
 PR 7696 started out as a spurious warning on ia64 but it turns out that
 this is probably a wrong code generation bug on any 64bit arch. The
 source code in question is this:
 $ cat 7696.c
                struct ia64_psr {
                  unsigned long cpl : 2;
                } x = {1};
 		
                main ()
                {
                  unsigned long y =  ((unsigned long) x.cpl) << 40;
 		 printf ("%lx %d\n", y, sizeof (y));
 		 return 0;
                }
 $ gcc -m64 7696.c 
 7696.c: In function `main':
 7696.c:7: warning: left shift count >= width of type
 This program will print 
 $ a.out
 100 8
 
 According to gdb the cast is thrown away early (build_binary_op doesn't see
 it at all), probably because the bitfield is of type unsigned long. But
 later on default_conversion sees that the precision is only two and converts
 x.cpl to an integer (instead of a long). This oviouly hurts on arches where
 int has 32 bit but long has 64.
 
 Two different work arounds can be used to cure the problem: either
 declare the bitfield as int or cast x.cpl to long long instead of long.
 Obviously none of these is a fix. I'll see if I can come up with a
 patch.
 
      regards   Christian
 
 -- 
 THAT'S ALL FOLKS!