--- layout: fc_discuss_archives title: Message 1 from Frama-C-discuss on January 2020 ---
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Frama-c-discuss] Strange results when verifying bitshifts using frama-c v19.1 with z3



Hi,
I know v20 is out and I should probably update but I've got some questions
about the behaviour of v19.1 that I would like to have the answer to.

The problem is that I get some strange result when using frama-c v19.1 with
why3 v1.2.1 and z3 v4.8.6 (running on macos 10.14.1).
It seems to starts when I have bit shift in the code I'm trying to verify,
as seen in the following example:

$ cat test.c
//@ ensures \result == 0; ensures \result != 0;
unsigned test(unsigned value) {
    return value >> 1;
}
$ frama-c -wp -wp-prover=z3 test.c
[kernel] Parsing test.c (with preprocessing)
[wp] Warning: Missing RTE guards
[wp] 2 goals scheduled
[wp] Proved goals:    2 / 2
  Qed:             0  (0.38ms-0.82ms)
  Z3:              2  (140ms)

1. Maybe I'm missing something, but to me, \result should neither be always
0 nor never 0 and it should especially not be both at the same time, right?

2. Any idea what could be causing this?

3. Is there a way to read the output of why3 and z3 to try to debug what is
happening without diving into the source code of frama-c or the
configuration files of why3?

I've put off upgrading to v20 as I need to verify quite a lot of bit
manipulation of integers and I've got the impression that the new version
is a bit buggy when it comes to that (even though workarounds seems to
exist).
4. Should I upgrade to v20 anyway or am I better off with 19.1 for that?

Regards,
Benjamin Nauck
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gforge.inria.fr/pipermail/frama-c-discuss/attachments/20200109/cb93477b/attachment.html>