I'm trying to put together proofs for my two functions into a proof for a whole program:
...
Definition Vprog : varspecs := nil.
Definition Gprog : funspecs := get_spec :: set_spec :: nil.
Lemma body_get: semax_body Vprog Gprog f_get get_spec.
Proof.
...
Qed.
Lemma body_set: semax_body Vprog Gprog f_set set_spec.
Proof.
...
Qed.
Existing Instance NullExtension.Espec.
Theorem all_funcs_correct:
semax_func Vprog Gprog (prog_funct prog) Gprog.
Proof.
semax_func_skipn.
semax_func_cons body_get.
Before application of semax_func_cons
tactic, I have the following goal:
1 subgoals, subgoal 1 (ID 3951)
============================
semax_func Vprog [get_spec; set_spec]
[(_get, Internal f_get); (_set, Internal f_set)]
[(_get,
WITH x : share * Z * (Z -> val) * val * val PRE [
(_key, tint), (_rez, tint), (_arr, tptr tint)]
(let (p, varr) := x in
let (p0, vk) := p in
let (p1, arr) := p0 in
let (sh, k) := p1 in
PROP (0 <= k < 100; forall i : Z, 0 <= i < 100 -> is_int (arr i);
repr k vk)
LOCAL (`(eq vk) (eval_id _key); `(eq varr) (eval_id _arr);
`isptr (eval_id _arr))
SEP (`(array_at tint sh arr 0 100) (eval_id _arr))) POST [tint]
(let (p, varr) := x in
let (p0, _) := p in
let (p1, arr) := p0 in
let (sh, k) := p1 in
`(array_at tint sh arr 0 100 varr) && local (`(eq (arr k)) retval)));
set_spec]
So it seems reasonable to eliminate f_get
with my proven lemma body_get
. Why the tactic fails?
The message does not help:
Toplevel input, characters 0-24:
Error: No applicable tactic.
The problem is in the argument list specification.
It must contain only function arguments (no locals)
It must have arguments in the same order as in the C function prototype.
BTW, the order of function in Gprog must be the same as the order of function definitions in the c-file.