Re-exported variant is incorrectly marked dead
Opened this issue · 1 comments
sim642 commented
Running reanalyze 389dd68 on Goblint goblint/analyzer@a544002 revealed the following.
In MyCFG
module we have the following re-exporting type definition:
(** Re-exported [Edge.t] with constructors. See [Edge.t] for documentation. *)
type edge = Edge.t =
| Assign of CilType.Lval.t * CilType.Exp.t
| Proc of CilType.Lval.t option * CilType.Exp.t * CilType.Exp.t list
| Entry of CilType.Fundec.t
| Ret of CilType.Exp.t option * CilType.Fundec.t
| Test of CilType.Exp.t * bool
| ASM of string list * Edge.asm_out * Edge.asm_in
| VDecl of CilType.Varinfo.t
| Skip
and in Edge
module we have the original type definition:
type t =
| Assign of CilType.Lval.t * CilType.Exp.t
(** Assignments lval = exp *)
| Proc of CilType.Lval.t option * CilType.Exp.t * CilType.Exp.t list
(** Function calls of the form lva = fexp (e1, e2, ...) *)
| Entry of CilType.Fundec.t
(** Entry edge that relates function declaration to function body. You can use
* this to initialize the local variables. *)
| Ret of CilType.Exp.t option * CilType.Fundec.t
(** Return edge is between the return statement, which may optionally contain
* a return value, and the function. The result of the call is then
* transferred to the function node! *)
| Test of CilType.Exp.t * bool
(** The true-branch or false-branch of a conditional exp *)
| ASM of string list * asm_out * asm_in
(** Inline assembly statements, and the annotations for output and input
* variables. *)
| VDecl of CilType.Varinfo.t
(** VDecl edge for the variable in varinfo. Whether such an edge is there for all
* local variables or only when it is not possible to pull the declaration up, is
* determined by alwaysGenerateVarDecl in cabs2cil.ml in CIL. One case in which a VDecl
* is always there is for VLA. If there is a VDecl edge, it is where the declaration originally
* appeared *)
| Skip [@dead "t.Skip"]
(** This is here for historical reasons. I never use Skip edges! *)
Despite the last documentation comment, we use MyCFG.Skip
in various places and it's not marked dead. But the corresponding Edge.Skip
variant is marked dead.
This is incorrect, because due to the type re-exporting, it wouldn't be possible to remove Edge.Skip
as the two type manifests must match for the code to compile. I suppose reanalyze isn't accounting for this fact.
cristianoc commented
I think there's no attempt in the implementation to handle re-exported variants at the moment.