adler32
Update an Adler-32 checksum using a 32-bit byte count.
Full chapter and shared rules ↗uLong ZEXPORT adler32(uLong adler, const Bytef *buf, uInt len);
Update an Adler-32 checksum using a 32-bit byte count.
Return values and next calls
Return type. uLong — 32-bit checksum
Returns the updated Adler-32, or the initial value for a new checksum.
adler32()| Returned value | Meaning and conditions | Caller’s next step |
|---|---|---|
1Initial value |
buf == NULL and len == 0, with any incoming checksum; no bytes are read. | Start a new sequence at 1. |
reduce(adler)Empty range reduces the seed |
buf is non-NULL and len == 0; a one-past pointer is also allowed for the empty range. reduce takes the low 32 bits and reduces each 16-bit half modulo 65521. The result equals low32(adler) when adler is a genuine Adler-32, because both halves of a genuine Adler-32 are below 65521. |
Keep this value for the next update. |
Adler-32(A || buf)Checksum after append |
buf covers len readable bytes, len > 0, and the low 32 bits of adler are the Adler-32 of A. | Use this value in later updates or compare it with the expected checksum. |
Reading the result.
- len is uInt, at most 2^32 - 1 bytes; a larger count converted to uInt loses its high bits before the call. Use the _z form or multiple updates for larger counts.
Calling conditions and effects
Domain Restricted domain. Effects: read-only when buf is non-NULL.
Pre Either buf is NULL and len == 0, or it points to len readable
bytes. For the checksum postcondition, the low 32 bits of adler are the Adler-32 of A.
Effects Preserve all memory and ownership. Read only the given range.
The forged-seed scope of adler32_z applies here too.
Proof goals.
- Goal.
Goal.adler32_safety(Safety/Checksums.lean); required. Theorem:ZlibF.Spec.Public.StatelessSafetyProof.adler32_safety(public-stateless-safety.lean). - Applies when.
bufis NULL withlen == 0, orbufpoints tolenreadable bytes (AdlerBufferBytes). Forlen == 0a pointer one past the end of an object is allowed. - Covered without condition. Every
adler, also a value that is not a checksum; every content of the bytes; every heap, with or without the library's tables. - Guarantees.
- Returns. The call returns.
- State, writes, allocations. None: the heap after the call is the heap before it.
- Not in this goal. NULL
bufwith a nonzerolen: outside every goal; this is narrower thanzlib.h(deviations, row "Adler initialization queries"). The value returned: functional goalGoal.adler32(not planned; a theorem proves it).
Complete return conditions and boundaries
Post Return exactly the result of adler32_z(adler, buf, len).
A NULL, zero-length query returns 1. A non-NULL empty range returns the low 32 bits of
adler with each 16-bit half reduced modulo 65521, which preserves a genuine checksum.
Shared conditions and scope
Compute a checksum a buffer at a time, or combine the checksums of independently processed byte sequences without rereading them. The byte sequences A and B below are ordinary data; the contracts use the Adler-32 or CRC-32 of those sequences. For example, after processing A and then B, the checksum of A can be the starting value of the next update; if A and B were processed independently, the combination functions give the checksum of A followed by B.
In the update functions, buf points to this batch of input, len is its byte count, and the
incoming adler or crc is the value to continue from. In the combination functions, len2 is
the byte count of the second sequence B only, not the total length of the two.
| Task | Entries |
|---|---|
| Update Adler-32 | Size-type length, 32-bit length |
| Combine Adler-32 | Ordinary offset, explicit 64-bit offset |
| Update CRC-32 | Size-type length, 32-bit length |
| Combine CRC-32 | Ordinary offset, explicit 64-bit offset |
| Reuse a CRC combination operator | Generate, generate with 64-bit length, apply |
| Inspect the byte table | get_crc_table |
The contracts follow the pinned adler32.c, crc32.c and zlib.h in the default LP64 Linux
static build. No patch touches these two C files (patches).
Dynamic table generation, hardware CRC hooks and generated table executables are not covered
(build and platform). No
checksum entry writes state, allocates, frees or performs I/O. Exceptional
numeric results are ordinary values, rather than error codes. Domain and effects remain
independent as explained in contract conventions.
These are focused contracts: Pre describes ordinary correct use and is deliberately narrower than
what the function accepts. A call outside Pre is not thereby undefined, and it need not return an
error code.
What is proved. Safety only (safety): each safety goal (Safety/Checksums.lean) says that the call returns and that the heap after the call is the heap before it. It says nothing about the value. Each entry gives its goal under Proof goals, in the five items and with the fixed phrases that Reading a Proof goals paragraph defines. The proof assumes the conditions listed there, not the documentary Pre. Post, the return tables and the notes describe functional behaviour. They are read from the C, are to confirm where no theorem states them, and are not required for the safety results. Progress says which goals are done.
The CRC goals have a premise on the model's heap: its two constant CRC tables, crc_table and
x2n_table, equal the tables of the translated crc32.h (CrcGlobals). The premise is assumed
of the heap the call starts on; no theorem derives it for an arbitrary heap. It is a condition on
the model, not a duty of the caller
(model limits). Two facts are
proved about it. A library heap, which is what the
stream goals start from, meets it (crc_globals_of_library). Each checksum goal returns the heap
unchanged, so the premise still holds after the call.
The practical checksum contracts use actual byte sequences A and B. Each
checksum is computed independently, starting at 1 for Adler-32 or 0 for CRC-32.
These contracts deliberately omit functional promises for forged checksums.
The functional goals in Functional/Checksums.lean
also describe arbitrary seeds and exact scalar results. Both the Adler entries and their Lean
goals, safety and functional, require zero length for NULL initialization queries
(AdlerBufferBytes in Common/Checksums.lean).
See deviations.
In this build, uLong and z_size_t are unsigned 64-bit types; uInt is
unsigned 32-bit. z_off_t and z_off64_t are signed 64-bit. Input checksum bits
above bit 31 are ignored; returned checksums fit in 32 bits. The explicit
*64 declarations depend on the header's large-file feature macros.
Safety goal
ZlibF.Spec.Public.Goal.adler32_safety
State: done · Status of the results
/-- `uLong adler32(uLong adler, const Bytef *buf, uInt len)`.
**Safety of `adler32`**: on a readable buffer of `len` bytes, or `NULL` with `len = 0`
(`AdlerBufferBytes`; narrower than zlib.h, `docs/en/deviations.md`), the call returns and the heap
is unchanged. -/
def Goal.adler32_safety : Prop :=
∀ (h : Heap) (seed : UInt64) (pointer : BPtr) (length : UInt32) (bytes : List UInt8),
AdlerBufferBytes h pointer length.toNat bytes →
∃ value, C.run (Zlib.Adler32.adler32 seed pointer length) h = .ok (value, h)Related definitions and source ↗Functional contract (beyond safety)
ZlibF.Spec.Public.Goal.adler32
State: done · Status of the results
/-- `uLong adler32(uLong adler, const Bytef *buf, uInt len)`.
The public 32-bit length entry has the same buffer domain and exact result as `_z`.
NULL queries require zero length, as recorded in `docs/en/deviations.md`. -/
def Goal.adler32 : Prop :=
∀ (h : Heap) (seed : UInt64) (pointer : BPtr) (length : UInt32) (bytes : List UInt8),
AdlerBufferBytes h pointer length.toNat bytes →
C.run (Zlib.Adler32.adler32 seed pointer length) h =
.ok ((if pointer = .null then 1 else adlerValue seed bytes), h)Related definitions and source ↗