Frama-C-discuss mailing list archives

This page gathers the archives of the old Frama-C-discuss archives, that was hosted by Inria's gforge before its demise at the end of 2020. To search for mails newer than September 2020, please visit the page of the new mailing list on Renater.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Frama-c-discuss] WP and memcpy question

Loic,

Thanks so much as always this helps a lot. We are writing our own code
generator for the EDSL I have for runtime verification and using Frama-C to
prove the generated code is correct and can not really avoid some of this
kind of messy code, but we have isolated it to one or two generated
functions.

Thanks again this was very helpful.


On Tue, Jun 19, 2018 at 5:03 AM Loïc Correnson <loic.correnson at cea.fr>
wrote:

> Hi Alwyn,
>
> I think the problem is on the granularity of the assigns clause.
> WP is using an over approximation of the assigns condition to prove
> assigns clauses.
> Each side effect must be **fully** included in one of the component of the
> assigns clauses to pass the check.
>
> Here, the memcpy effect on s1_buff will assigns a **range** of s1_buff,
> namely s1_buff[s1_idx][0..2] if my reading is correct.
> But the assigns contract of the function under proof is : assigns
> s1_buff[s1_idx][0], s1_buff[s1_idx][1], etc.
>
> For the other memcpy, I think an assigns clause regarding exerr_cpy is
> missing.
>
> With these two additions, it only remains one unproved goal
> on s0_gen_ensures.
>
> L.
>
> Le 19 juin 2018 à 00:44, Alwyn Goodloe <agoodloe at gmail.com> a écrit :
>
>
> The annotated C file below The last function  is given as follows
>
> /*@ requires 0 <= s0_idx < 1;
>     requires 0 <= s1_idx < 2;
>     requires  \separated(exarr_cpy + (0..2), exarr + (0..2));
>     requires \separated(s1, exarr_cpy+(0..2));
>     requires \separated(s1, exarr+(0..2));
>     requires \separated(&s0, exarr_cpy+(0..2));
>     requires \separated(&s0, exarr+(0..2));
>     assigns  s0;
>     assigns  s1;
>     assigns  s0_buff[s0_idx];
>     assigns  s1_buff[s1_idx][0];
>     assigns  s1_buff[s1_idx][1];
>     assigns  s1_buff[s1_idx][2];
>     assigns  s0_idx;
>     assigns  s1_idx;
>     ensures  0 <= s0_idx < 1;
>     ensures  0 <= s1_idx < 2;
>     ensures  \forall int i; 0 <= i < 1&& i != \old(s0_idx) ==>
>         s0_buff[i] == \old(s0_buff[i]);
>     ensures  \forall int i; 0 <= i < 2&& i != \old(s1_idx) ==>
>         s1_buff[i] == \old(s1_buff[i]);
> */
> static void step () {
>   memcpy(exarr_cpy, exarr, sizeof(exarr_cpy));
>   if (farray_guard()) {farray(farray_arg0(),
>                        farray_arg1(),
>                        farray_arg2());
>                      }
>   s0 = s0_gen();
>   s1 = s1_gen();
>   int8_t s1_tmp[3];
>   memcpy(s1_tmp, s1, sizeof(s1_tmp));
>   /*@ assert 0 <= s0_idx < 1; */
>   s0_buff[s0_idx] = s0;
>   memcpy(s1_buff[s1_idx], s1_tmp, sizeof(s1_tmp));
>   ++(s0_idx);
>   ++(s1_idx);
>   s0_idx = s0_idx % 1;
>   s1_idx = s1_idx % 2;
> }
>
>
> When I run wp with both alt-ergo and Z3 provers all but four VCs are
> discharged Each unproven VC seems related to memcpy and the goal is always
> the assigns s0 in line 117.  Any suggestions on what I might need to do in
> order to would be appreciated.  Here is one of messages  produced for one
> of the unproved VCs.
>
> Goal Assigns (file NewArray.c, line 117) in 'step' (2/14):
> Call Effect at line 133
> Assume {
>   Type: is_uint32(s0_idx_0) /\ is_uint32(s1_idx_0) /\ is_sint32(status_0).
>   (* Heap *)
>   Have: (region(s1_0.base) <= 0) /\ linked(Malloc_0) /\ sconst(Mchar_0).
>   (* Pre-condition *)
>   Have: (s0_idx_0 <= 0) /\ (0 <= s0_idx_0).
>   (* Pre-condition *)
>   Have: (0 <= s1_idx_0) /\ (s1_idx_0 <= 1).
>   (* Pre-condition *)
>   Have: separated(s1_0, 1, shift_sint8(global(G_exarr_cpy_1569), 0), 3).
>   (* Pre-condition *)
>   Have: separated(s1_0, 1, shift_sint8(global(G_exarr_1191), 0), 3).
>   (* Merge *)
>   Either {
>     Case:
>       Let a = shift_sint8(global(G_exarr_cpy_1569), 0).
>       (* Call 'memcpy' *)
>       Have: L_memcmp(Mchar_0, Mchar_1, a,
>               shift_sint8(global(G_exarr_1191), 0), 3) = 0.
>       (* Call Effects *)
>       Have: havoc(Mchar_0, Mchar_1, a, 3).
>     Case: (* Exit Effects *) Have: havoc(Mchar_0, Mchar_2, a, 3).
>   }
> }
> Prove: false.
> Prover Z3 returns Failed
> Why3 exits with status 1.
> Prover Alt-Ergo returns Unknown (Qed:5ms) (271ms)
>
>
>
> --
> Alwyn E. Goodloe, Ph.D.
> agoodloe at gmail.com
>
> Research Computer Engineer
> NASA Langley Research Center
> <NewArray.c>_______________________________________________
> Frama-c-discuss mailing list
> Frama-c-discuss at lists.gforge.inria.fr
> https://lists.gforge.inria.fr/mailman/listinfo/frama-c-discuss
>
>
> _______________________________________________
> Frama-c-discuss mailing list
> Frama-c-discuss at lists.gforge.inria.fr
> https://lists.gforge.inria.fr/mailman/listinfo/frama-c-discuss



-- 
Alwyn E. Goodloe, Ph.D.
agoodloe at gmail.com

Research Computer Engineer
NASA Langley Research Center
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gforge.inria.fr/pipermail/frama-c-discuss/attachments/20180619/4c7375a9/attachment.html>