[Bug c/52560] New: if (r == -1) causes 'assuming signed overflow does not occur when simplifying conditional to constant'

[Bug c/52560] New: if (r == -1) causes 'assuming signed overflow does not occur when simplifying conditional to constant'

[rjones at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52560

             Bug #: 52560
           Summary: if (r == -1) causes 'assuming signed overflow does not
                    occur when simplifying conditional to constant'
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: rjones@redhat.com


Here is the reproducer:

wget 'http://oirase.annexia.org/strict-overflow-warning.i.xz'
unxz strict-overflow-warning.i.xz
gcc -std=gnu99 -O2 -Wstrict-overflow -c strict-overflow-warning.i

The warning is:

  inspect_fs_unix.c: In function ‘check_fstab’:
  inspect_fs_unix.c:1075:6: warning: assuming signed overflow does not occur
when simplifying conditional to constant [-Wstrict-overflow]

However the code doesn't look like anything should be simplified,
or a warning:

  n_app_md_devices = map_app_md_devices (g, &app_map);
  if (n_app_md_devices == -1) goto error;

where map_app_md_devices is a function that returns an int:

  static int map_app_md_devices (guestfs_h *g, Hash_table **map);

and n_app_md_devices is also an int.

I've tried this on several versions of gcc:

gcc (GCC) 4.7.0 20120308 (Red Hat 4.7.0-0.19)
Copyright (C) 2012 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

(for further build info, go to:
http://koji.fedoraproject.org/koji/buildinfo?buildID=305760
and click 'build logs')

Same thing with this gcc from Ubuntu 11.10:

$ gcc --version
gcc (Ubuntu/Linaro 4.6.1-9ubuntu3) 4.6.1
Copyright (C) 2011 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

We think this first started happening in gcc 4.5.1.

[Bug c/52560] if (r == -1) causes 'assuming signed overflow does not occur when simplifying conditional to constant'

[jim at meyering dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52560

--- Comment #1 from jim at meyering dot net 2012-03-12 12:30:20 UTC ---
Created attachment 26877
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=26877
50-line reproducer

[Bug tree-optimization/52560] if (r == -1) causes 'assuming signed overflow does not occur when simplifying conditional to constant'

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52560

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-03-12
          Component|c                           |tree-optimization
     Ever Confirmed|0                           |1

--- Comment #2 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-03-12 13:22:32 UTC ---
It certainly inlines the function, does not figures out the loop does not run
and then computes n's value-range as is [0, +INF(OVF)] and thus when
simplifying
the return value comparison against -1 it says it assumes that n++ does not
wrap.

Looks ok to me, though it is all because of dead code and thus a missed
VRP of some sort.

[Bug tree-optimization/52560] if (r == -1) causes 'assuming signed overflow does not occur when simplifying conditional to constant'

[rjones at redhat dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52560

--- Comment #3 from Richard W.M. Jones <rjones at redhat dot com> 2012-03-12 16:30:45 UTC ---
I see that this is actually a bug in our code.  I pushed
the following fix to libguestfs:
https://github.com/libguestfs/libguestfs/commit/d66dd2260c724bdfe57a8595aac37c8e9173cee5