Frama-C API - Sigma
 module F = Lang.FMemory Environments.
The concrete memory is partionned into a vector of abstract data. Each component of the partition is called a memory chunk and holds an abstract representation of some part of the memory.
A sigma assigns a logical variable to each memory chunk, assigned lazily, that represents the contents of this slice of memory at a given program point.
Remark: memory chunks are not required to be independent from each other, provided the memory model implementation is consistent with the chosen representation. Conversely, a given object might be represented by several memory chunks.
Memory chunk types.
The type of chunks is extensible. Each memory model extends this type with of its own memory chunks by functor Make.
Memory chunks at a given program point.
This is a collection of variables indexed by chunks. New chunks are generated from the context pool of Lang.freshvar.
Chunks API
module type ChunkType = sig ... endmodule Heap : Qed.Collection.S with type t = chunkUniform Sets and Maps of chunks.
module Domain = Heap.Settype domain = Domain.tMemory footprint, ie. set of memory model chunk types.
Internal representation of chunks (chunk-kind). This type can only be extended via functor Make.
Sigma API
val create : unit -> sigmaInitially empty environment.
val pretty : Stdlib.Format.formatter -> sigma -> unitFor debugging purpose
Duplicate the environment. Fresh chunks in the copy are not duplicated into the source environment.
Make two environment pairwise equal via the passive form.
Missing chunks in one environment are added with the corresponding variable of the other environment. When both environments don't agree on a chunk, their variables are added to the passive form.
val assigned : pre:sigma -> post:sigma -> domain -> F.pred Frama_c_kernel.Bag.tMake chunks equal outside of some domain.
This is similar to join, but outside the given footprint of an assigns clause. Although, the function returns the equality predicates instead of a passive form.
Like in join, missing chunks are reported from one side to the other one, and common chunks are added to the equality bag.
Make the union of each sigma, choosing the minimal variable in case of conflict. Both initial environments are kept unchanged.
Make the union of each sigma, choosing a new variable for each conflict, and returns the corresponding joins. Both initial environments are kept unchanged.
Same than merge but for a list of sigmas. Much more efficient than folding merge step by step.
Iterates over the chunks and associated variables already accessed so far in the environment.
Same as iter for both environments.
All the chunks in the provided footprint are generated and made fresh.
Existing chunk variables outside the footprint are copied into the new environment. The original environment itself is kept unchanged. More efficient than iterating havoc_chunk over the footprint.
All the chunks are made fresh. As an optimisation, when ~call:true is set, only non-local chunks are made fresh. Local chunks are those for which Chunk.is_frame returns true.
Return a copy of the environment where chunks in the footprint have been removed. Keep the original environment unchanged.
val is_init : chunk -> boolShortcut to Chunk.is_init
val is_framed : chunk -> boolShortcut to Chunk.is_framed
Footprint of a memory environment. That is, the set of accessed chunks so far in the environment.
val empty : domainSame as Domain.empty
writes s indicates which chunks are new in s.post compared to s.pre.
