/* zlibmodule.c -- gzip-compatible data compression */
/* See http://zlib.net/ */

/* Windows users:  read Python's PCbuild\readme.txt */

#ifndef Py_BUILD_CORE_BUILTIN
#  define Py_BUILD_CORE_MODULE 1
#endif

#include "Python.h"
#include "pycore_pyatomic_ft_wrappers.h" // FT_ATOMIC_STORE_CHAR_RELAXED

#include "zlib.h"
#include "stdbool.h"
#include <stddef.h>               // offsetof()

#if defined(ZLIB_VERNUM) && ZLIB_VERNUM < 0x1221
#error "At least zlib version 1.2.2.1 is required"
#endif

#if (SIZEOF_OFF_T == SIZEOF_SIZE_T)
#  define convert_to_z_off_t  PyLong_AsSsize_t
#elif (SIZEOF_OFF_T == SIZEOF_LONG_LONG)
#  define convert_to_z_off_t  PyLong_AsLongLong
#elif (SIZEOF_OFF_T == SIZEOF_LONG)
#  define convert_to_z_off_t  PyLong_AsLong
#else
#  error off_t does not match either size_t, long, or long long!
#endif

// Blocks output buffer wrappers
#include "pycore_blocks_output_buffer.h"

#if OUTPUT_BUFFER_MAX_BLOCK_SIZE > UINT32_MAX
    #error "The maximum block size accepted by zlib is UINT32_MAX."
#endif

/* On success, return value >= 0
   On failure, return -1 */
static inline Py_ssize_t
OutputBuffer_InitAndGrow(_BlocksOutputBuffer *buffer, Py_ssize_t max_length,
                         Bytef **next_out, uint32_t *avail_out)
{
    Py_ssize_t allocated;

    allocated = _BlocksOutputBuffer_InitAndGrow(
                    buffer, max_length, (void**) next_out);
    *avail_out = (uint32_t) allocated;
    return allocated;
}

/* On success, return value >= 0
   On failure, return -1 */
static inline Py_ssize_t
OutputBuffer_Grow(_BlocksOutputBuffer *buffer,
                  Bytef **next_out, uint32_t *avail_out)
{
    Py_ssize_t allocated;

    allocated = _BlocksOutputBuffer_Grow(
                    buffer, (void**) next_out, (Py_ssize_t) *avail_out);
    *avail_out = (uint32_t) allocated;
    return allocated;
}

static inline Py_ssize_t
OutputBuffer_GetDataSize(_BlocksOutputBuffer *buffer, uint32_t avail_out)
{
    return _BlocksOutputBuffer_GetDataSize(buffer, (Py_ssize_t) avail_out);
}

static inline PyObject *
OutputBuffer_Finish(_BlocksOutputBuffer *buffer, uint32_t avail_out)
{
    return _BlocksOutputBuffer_Finish(buffer, (Py_ssize_t) avail_out);
}

static inline void
OutputBuffer_OnError(_BlocksOutputBuffer *buffer)
{
    _BlocksOutputBuffer_OnError(buffer);
}

/* The max buffer size accepted by zlib is UINT32_MAX, the initial buffer size
   `init_size` may > it in 64-bit build. These wrapper functions maintain an
   UINT32_MAX sliding window for the first block:
    1. OutputBuffer_WindowInitWithSize()
    2. OutputBuffer_WindowGrow()
    3. OutputBuffer_WindowFinish()
    4. OutputBuffer_WindowOnError()

   ==== is the sliding window:
    1. ====------
           ^ next_posi, left_bytes is 6
    2. ----====--
               ^ next_posi, left_bytes is 2
    3. --------==
                 ^ next_posi, left_bytes is 0  */
typedef struct {
    Py_ssize_t left_bytes;
    Bytef *next_posi;
} _Uint32Window;

/* Initialize the buffer with an initial buffer size.

   On success, return value >= 0
   On failure, return value < 0 */
static inline Py_ssize_t
OutputBuffer_WindowInitWithSize(_BlocksOutputBuffer *buffer, _Uint32Window *window,
                                Py_ssize_t init_size,
                                Bytef **next_out, uint32_t *avail_out)
{
    Py_ssize_t allocated = _BlocksOutputBuffer_InitWithSize(
                               buffer, init_size, (void**) next_out);

    if (allocated >= 0) {
        // the UINT32_MAX sliding window
        Py_ssize_t window_size = Py_MIN((size_t)allocated, UINT32_MAX);
        *avail_out = (uint32_t) window_size;

        window->left_bytes = allocated - window_size;
        window->next_posi = *next_out + window_size;
    }
    return allocated;
}

/* Grow the buffer.

   On success, return value >= 0
   On failure, return value < 0 */
static inline Py_ssize_t
OutputBuffer_WindowGrow(_BlocksOutputBuffer *buffer, _Uint32Window *window,
                        Bytef **next_out, uint32_t *avail_out)
{
    Py_ssize_t allocated;

    /* ensure no gaps in the data.
       if inlined, this check could be optimized away.*/
    if (*avail_out != 0) {
        PyErr_SetString(PyExc_SystemError,
                        "*avail_out != 0 in OutputBuffer_WindowGrow().");
        return -1;
    }

    // slide the UINT32_MAX sliding window
    if (window->left_bytes > 0) {
        Py_ssize_t window_size = Py_MIN((size_t)window->left_bytes, UINT32_MAX);

        *next_out = window->next_posi;
        *avail_out = (uint32_t) window_size;

        window->left_bytes -= window_size;
        window->next_posi += window_size;

        return window_size;
    }
    assert(window->left_bytes == 0);

    // only the first block may > UINT32_MAX
    allocated = _BlocksOutputBuffer_Grow(
                    buffer, (void**) next_out, (Py_ssize_t) *avail_out);
    *avail_out = (uint32_t) allocated;
    return allocated;
}

/* Finish the buffer.

   On success, return a bytes object
   On failure, return NULL */
static inline PyObject *
OutputBuffer_WindowFinish(_BlocksOutputBuffer *buffer, _Uint32Window *window,
                          uint32_t avail_out)
{
    Py_ssize_t real_avail_out = (Py_ssize_t) avail_out + window->left_bytes;
    return _BlocksOutputBuffer_Finish(buffer, real_avail_out);
}

static inline void
OutputBuffer_WindowOnError(_BlocksOutputBuffer *buffer, _Uint32Window *window)
{
    _BlocksOutputBuffer_OnError(buffer);
}


/* The following parameters are copied from zutil.h, version 0.95 */
#define DEFLATED   8
#if MAX_MEM_LEVEL >= 8
#  define DEF_MEM_LEVEL 8
#else
#  define DEF_MEM_LEVEL  MAX_MEM_LEVEL
#endif

/* Initial buffer size. */
#define DEF_BUF_SIZE (16*1024)
#define DEF_MAX_INITIAL_BUF_SIZE (16 * 1024 * 1024)

static PyModuleDef zlibmodule;

typedef struct {
    PyTypeObject *Comptype;
    PyTypeObject *Decomptype;
    PyTypeObject *ZlibDecompressorType;
    PyObject *ZlibError;
} zlibstate;

static inline zlibstate*
get_zlib_state(PyObject *module)
{
    void *state = PyModule_GetState(module);
    assert(state != NULL);
    return (zlibstate *)state;
}

typedef struct
{
    PyObject_HEAD
    z_stream zst;
    PyObject *unused_data;
    PyObject *unconsumed_tail;
    char eof;
    bool is_initialised;
    PyObject *zdict;
    PyMutex mutex;
} compobject;

#define _compobject_CAST(op)    ((compobject *)op)

static void
zlib_error(zlibstate *state, z_stream zst, int err, const char *msg)
{
    const char *zmsg = Z_NULL;
    /* In case of a version mismatch, zst.msg won't be initialized.
       Check for this case first, before looking at zst.msg. */
    if (err == Z_VERSION_ERROR)
        zmsg = "library version mismatch";
    if (zmsg == Z_NULL)
        zmsg = zst.msg;
    if (zmsg == Z_NULL) {
        switch (err) {
        case Z_BUF_ERROR:
            zmsg = "incomplete or truncated stream";
            break;
        case Z_STREAM_ERROR:
            zmsg = "inconsistent stream state";
            break;
        case Z_DATA_ERROR:
            zmsg = "invalid input data";
            break;
        }
    }
    if (zmsg == Z_NULL)
        PyErr_Format(state->ZlibError, "Error %d %s", err, msg);
    else
        PyErr_Format(state->ZlibError, "Error %d %s: %.200s", err, msg, zmsg);
}

/*[clinic input]
module zlib
class zlib.Compress "compobject *" "&Comptype"
class zlib.Decompress "compobject *" "&Decomptype"
[clinic start generated code]*/
/*[clinic end generated code: output=da39a3ee5e6b4b0d input=093935115c3e3158]*/

static compobject *
newcompobject(PyTypeObject *type)
{
    compobject *self;
    assert(type != NULL);
    assert(type->tp_alloc != NULL);
    self = _compobject_CAST(type->tp_alloc(type, 0));
    if (self == NULL)
        return NULL;
    self->eof = 0;
    self->is_initialised = 0;
    self->zdict = NULL;
    self->unused_data = Py_GetConstant(Py_CONSTANT_EMPTY_BYTES);
    if (self->unused_data == NULL) {
        Py_DECREF(self);
        return NULL;
    }
    self->unconsumed_tail = Py_GetConstant(Py_CONSTANT_EMPTY_BYTES);
    if (self->unconsumed_tail == NULL) {
        Py_DECREF(self);
        return NULL;
    }
    self->mutex = (PyMutex){0};
    return self;
}

static void*
PyZlib_Malloc(voidpf ctx, uInt items, uInt size)
{
    if (size != 0 && items > (size_t)PY_SSIZE_T_MAX / size)
        return NULL;
    /* PyMem_Malloc() cannot be used: the GIL is not held when
       inflate() and deflate() are called */
    return PyMem_RawMalloc((size_t)items * (size_t)size);
}

static void
PyZlib_Free(voidpf ctx, void *ptr)
{
    PyMem_RawFree(ptr);
}

static void
arrange_input_buffer(z_stream *zst, Py_ssize_t *remains)
{
    zst->avail_in = (uInt)Py_MIN((size_t)*remains, UINT_MAX);
    *remains -= zst->avail_in;
}

/*[clinic input]
zlib.compress

    data: Py_buffer
        Binary data to be compressed.
    /
    level: int(c_default="Z_DEFAULT_COMPRESSION") = Z_DEFAULT_COMPRESSION
        Compression level, in 0-9 or -1.
    wbits: int(c_default="MAX_WBITS") = MAX_WBITS
        The window buffer size and container format.

Returns a bytes object containing compressed data.
[clinic start generated code]*/

static PyObject *
zlib_compress_impl(PyObject *module, Py_buffer *data, int level, int wbits)
/*[clinic end generated code: output=46bd152fadd66df2 input=c4d06ee5782a7e3f]*/
{
    PyObject *return_value;
    int flush;
    z_stream zst;
    _BlocksOutputBuffer buffer = {.writer = NULL};

    zlibstate *state = get_zlib_state(module);

    Byte *ibuf = data->buf;
    Py_ssize_t ibuflen = data->len;

    if (OutputBuffer_InitAndGrow(&buffer, -1, &zst.next_out, &zst.avail_out) < 0) {
        goto error;
    }

    zst.opaque = NULL;
    zst.zalloc = PyZlib_Malloc;
    zst.zfree = PyZlib_Free;
    zst.next_in = ibuf;
    int err = deflateInit2(&zst, level, DEFLATED, wbits, DEF_MEM_LEVEL,
                           Z_DEFAULT_STRATEGY);

    switch (err) {
    case Z_OK:
        break;
    case Z_MEM_ERROR:
        PyErr_SetString(PyExc_MemoryError,
                        "Out of memory while compressing data");
        goto error;
    case Z_STREAM_ERROR:
        PyErr_SetString(state->ZlibError, "Bad compression level");
        goto error;
    default:
        deflateEnd(&zst);
        zlib_error(state, zst, err, "while compressing data");
        goto error;
    }

    do {
        arrange_input_buffer(&zst, &ibuflen);
        flush = ibuflen == 0 ? Z_FINISH : Z_NO_FLUSH;

        do {
            if (zst.avail_out == 0) {
                if (OutputBuffer_Grow(&buffer, &zst.next_out, &zst.avail_out) < 0) {
                    deflateEnd(&zst);
                    goto error;
                }
            }

            Py_BEGIN_ALLOW_THREADS
            err = deflate(&zst, flush);
            Py_END_ALLOW_THREADS

            if (err == Z_STREAM_ERROR) {
                deflateEnd(&zst);
                zlib_error(state, zst, err, "while compressing data");
                goto error;
            }

        } while (zst.avail_out == 0);
        assert(zst.avail_in == 0);

    } while (flush != Z_FINISH);
    assert(err == Z_STREAM_END);

    err = deflateEnd(&zst);
    if (err == Z_OK) {
        return_value = OutputBuffer_Finish(&buffer, zst.avail_out);
        if (return_value == NULL) {
            goto error;
        }
        return return_value;
    }
    else
        zlib_error(state, zst, err, "while finishing compression");
 error:
    OutputBuffer_OnError(&buffer);
    return NULL;
}

/*[clinic input]
zlib.decompress

    data: Py_buffer
        Compressed data.
    /
    wbits: int(c_default="MAX_WBITS") = MAX_WBITS
        The window buffer size and container format.
    bufsize: Py_ssize_t(c_default="DEF_BUF_SIZE", allow_negative=False) = DEF_BUF_SIZE
        The initial output buffer size.

Returns a bytes object containing the uncompressed data.
[clinic start generated code]*/

static PyObject *
zlib_decompress_impl(PyObject *module, Py_buffer *data, int wbits,
                     Py_ssize_t bufsize)
/*[clinic end generated code: output=77c7e35111dc8c42 input=530077065b3a2233]*/
{
    PyObject *return_value;
    Byte *ibuf;
    Py_ssize_t ibuflen;
    int err, flush;
    z_stream zst;
    _BlocksOutputBuffer buffer = {.writer = NULL};
    _Uint32Window window;  // output buffer's UINT32_MAX sliding window

    zlibstate *state = get_zlib_state(module);

    if (bufsize == 0) {
        bufsize = 1;
    }

    if (OutputBuffer_WindowInitWithSize(&buffer, &window, bufsize,
                                        &zst.next_out, &zst.avail_out) < 0) {
        goto error;
    }

    ibuf = data->buf;
    ibuflen = data->len;

    zst.opaque = NULL;
    zst.zalloc = PyZlib_Malloc;
    zst.zfree = PyZlib_Free;
    zst.avail_in = 0;
    zst.next_in = ibuf;
    err = inflateInit2(&zst, wbits);

    switch (err) {
    case Z_OK:
        break;
    case Z_MEM_ERROR:
        PyErr_SetString(PyExc_MemoryError,
                        "Out of memory while decompressing data");
        goto error;
    default:
        inflateEnd(&zst);
        zlib_error(state, zst, err, "while preparing to decompress data");
        goto error;
    }

    do {
        arrange_input_buffer(&zst, &ibuflen);
        flush = ibuflen == 0 ? Z_FINISH : Z_NO_FLUSH;

        do {
            if (zst.avail_out == 0) {
                if (OutputBuffer_WindowGrow(&buffer, &window,
                                            &zst.next_out, &zst.avail_out) < 0) {
                    inflateEnd(&zst);
                    goto error;
                }
            }

            Py_BEGIN_ALLOW_THREADS
            err = inflate(&zst, flush);
            Py_END_ALLOW_THREADS

            switch (err) {
            case Z_OK: _Py_FALLTHROUGH;
            case Z_BUF_ERROR: _Py_FALLTHROUGH;
            case Z_STREAM_END:
                break;
            case Z_MEM_ERROR:
                inflateEnd(&zst);
                PyErr_SetString(PyExc_MemoryError,
                                "Out of memory while decompressing data");
                goto error;
            default:
                inflateEnd(&zst);
                zlib_error(state, zst, err, "while decompressing data");
                goto error;
            }

        } while (zst.avail_out == 0);

    } while (err != Z_STREAM_END && ibuflen != 0);


    if (err != Z_STREAM_END) {
        inflateEnd(&zst);
        zlib_error(state, zst, err, "while decompressing data");
        goto error;
    }

    err = inflateEnd(&zst);
    if (err != Z_OK) {
        zlib_error(state, zst, err, "while finishing decompression");
        goto error;
    }

    return_value = OutputBuffer_WindowFinish(&buffer, &window, zst.avail_out);
    if (return_value != NULL) {
        return return_value;
    }

 error:
    OutputBuffer_WindowOnError(&buffer, &window);
    return NULL;
}

/*[clinic input]
zlib.compressobj

    level: int(c_default="Z_DEFAULT_COMPRESSION") = Z_DEFAULT_COMPRESSION
        The compression level (an integer in the range 0-9 or -1; default is
        currently equivalent to 6).  Higher compression levels are slower,
        but produce smaller results.
    method: int(c_default="DEFLATED") = DEFLATED
        The compression algorithm.  If given, this must be DEFLATED.
    wbits: int(c_default="MAX_WBITS") = MAX_WBITS
        +9 to +15: The base-two logarithm of the window size.  Include a zlib
            container.
        -9 to -15: Generate a raw stream.
        +25 to +31: Include a gzip container.
    memLevel: int(c_default="DEF_MEM_LEVEL") = DEF_MEM_LEVEL
        Controls the amount of memory used for internal compression state.
        Valid values range from 1 to 9.  Higher values result in higher memory
        usage, faster compression, and smaller output.
    strategy: int(c_default="Z_DEFAULT_STRATEGY") = Z_DEFAULT_STRATEGY
        Used to tune the compression algorithm.  Possible values are
        Z_DEFAULT_STRATEGY, Z_FILTERED, and Z_HUFFMAN_ONLY.
    zdict: Py_buffer = NULL
        The predefined compression dictionary - a sequence of bytes
        containing subsequences that are likely to occur in the input data.

Return a compressor object.
[clinic start generated code]*/

static PyObject *
zlib_compressobj_impl(PyObject *module, int level, int method, int wbits,
                      int memLevel, int strategy, Py_buffer *zdict)
/*[clinic end generated code: output=8b5bed9c8fc3814d input=1a6f61d8a8885c0d]*/
{
    zlibstate *state = get_zlib_state(module);
    if (zdict->buf != NULL && (size_t)zdict->len > UINT_MAX) {
        PyErr_SetString(PyExc_OverflowError,
                        "zdict length does not fit in an unsigned int");
        return NULL;
    }

    compobject *self = newcompobject(state->Comptype);
    if (self == NULL)
        goto error;
    self->zst.opaque = NULL;
    self->zst.zalloc = PyZlib_Malloc;
    self->zst.zfree = PyZlib_Free;
    self->zst.next_in = NULL;
    self->zst.avail_in = 0;
    int err = deflateInit2(&self->zst, level, method, wbits, memLevel, strategy);
    switch (err) {
    case Z_OK:
        self->is_initialised = 1;
        if (zdict->buf == NULL) {
            goto success;
        } else {
            err = deflateSetDictionary(&self->zst,
                                       zdict->buf, (unsigned int)zdict->len);
            switch (err) {
            case Z_OK:
                goto success;
            case Z_STREAM_ERROR:
                PyErr_SetString(PyExc_ValueError, "Invalid dictionary");
                goto error;
            default:
                PyErr_SetString(PyExc_ValueError, "deflateSetDictionary()");
                goto error;
            }
       }
    case Z_MEM_ERROR:
        PyErr_SetString(PyExc_MemoryError,
                        "Can't allocate memory for compression object");
        goto error;
    case Z_STREAM_ERROR:
        PyErr_SetString(PyExc_ValueError, "Invalid initialization option");
        goto error;
    default:
        zlib_error(state, self->zst, err, "while creating compression object");
        goto error;
    }

 error:
    Py_CLEAR(self);
 success:
    return (PyObject *)self;
}

static int
set_inflate_zdict(zlibstate *state, compobject *self)
{
    Py_buffer zdict_buf;
    if (PyObject_GetBuffer(self->zdict, &zdict_buf, PyBUF_SIMPLE) == -1) {
        return -1;
    }
    if ((size_t)zdict_buf.len > UINT_MAX) {
        PyErr_SetString(PyExc_OverflowError,
                        "zdict length does not fit in an unsigned int");
        PyBuffer_Release(&zdict_buf);
        return -1;
    }
    int err;
    err = inflateSetDictionary(&self->zst,
                               zdict_buf.buf, (unsigned int)zdict_buf.len);
    PyBuffer_Release(&zdict_buf);
    if (err != Z_OK) {
        zlib_error(state, self->zst, err, "while setting zdict");
        return -1;
    }
    return 0;
}

/*[clinic input]
zlib.decompressobj

    wbits: int(c_default="MAX_WBITS") = MAX_WBITS
        The window buffer size and container format.
    zdict: object(c_default="NULL") = b''
        The predefined compression dictionary.  This must be the same
        dictionary as used by the compressor that produced the input data.

Return a decompressor object.
[clinic start generated code]*/

static PyObject *
zlib_decompressobj_impl(PyObject *module, int wbits, PyObject *zdict)
/*[clinic end generated code: output=3069b99994f36906 input=d3832b8511fc977b]*/
{
    zlibstate *state = get_zlib_state(module);

    if (zdict != NULL && !PyObject_CheckBuffer(zdict)) {
        PyErr_SetString(PyExc_TypeError,
                        "zdict argument must support the buffer protocol");
        return NULL;
    }

    compobject *self = newcompobject(state->Decomptype);
    if (self == NULL)
        return NULL;
    self->zst.opaque = NULL;
    self->zst.zalloc = PyZlib_Malloc;
    self->zst.zfree = PyZlib_Free;
    self->zst.next_in = NULL;
    self->zst.avail_in = 0;
    if (zdict != NULL) {
        self->zdict = Py_NewRef(zdict);
    }
    int err = inflateInit2(&self->zst, wbits);
    switch (err) {
    case Z_OK:
        self->is_initialised = 1;
        if (self->zdict != NULL && wbits < 0) {
            if (set_inflate_zdict(state, self) < 0) {
                Py_DECREF(self);
                return NULL;
            }
        }
        return (PyObject *)self;
    case Z_STREAM_ERROR:
        Py_DECREF(self);
        PyErr_SetString(PyExc_ValueError, "Invalid initialization option");
        return NULL;
    case Z_MEM_ERROR:
        Py_DECREF(self);
        PyErr_SetString(PyExc_MemoryError,
                        "Can't allocate memory for decompression object");
        return NULL;
    default:
        zlib_error(state, self->zst, err, "while creating decompression object");
        Py_DECREF(self);
        return NULL;
    }
}

static void
compobject_dealloc_impl(PyObject *op, int (*dealloc)(z_streamp))
{
    PyTypeObject *type = Py_TYPE(op);
    PyObject_GC_UnTrack(op);
    compobject *self = _compobject_CAST(op);
    assert(!PyMutex_IsLocked(&self->mutex));
    if (self->is_initialised) {
        (void)dealloc(&self->zst);
    }
    Py_XDECREF(self->unused_data);
    Py_XDECREF(self->unconsumed_tail);
    Py_XDECREF(self->zdict);
    type->tp_free(self);
    Py_DECREF(type);
}

static int
compobject_traverse(PyObject *op, visitproc visit, void *arg)
{
    compobject *self = _compobject_CAST(op);
    Py_VISIT(Py_TYPE(op));
    Py_VISIT(self->zdict);
    return 0;
}

static void
Comp_dealloc(PyObject *op)
{
    compobject_dealloc_impl(op, &deflateEnd);
}

static void
Decomp_dealloc(PyObject *op)
{
    compobject_dealloc_impl(op, &inflateEnd);
}

/*[clinic input]
zlib.Compress.compress

    cls: defining_class
    data: Py_buffer
        Binary data to be compressed.
    /

Returns a bytes object containing compressed data.

After calling this function, some of the input data may still
be stored in internal buffers for later processing.
Call the flush() method to clear these buffers.
[clinic start generated code]*/

static PyObject *
zlib_Compress_compress_impl(compobject *self, PyTypeObject *cls,
                            Py_buffer *data)
/*[clinic end generated code: output=6731b3f0ff357ca6 input=04d00f65ab01d260]*/
{
    PyObject *return_value;
    int err;
    _BlocksOutputBuffer buffer = {.writer = NULL};
    zlibstate *state = PyType_GetModuleState(cls);

    PyMutex_Lock(&self->mutex);

    self->zst.next_in = data->buf;
    Py_ssize_t ibuflen = data->len;

    if (OutputBuffer_InitAndGrow(&buffer, -1, &self->zst.next_out, &self->zst.avail_out) < 0) {
        goto error;
    }

    do {
        arrange_input_buffer(&self->zst, &ibuflen);

        do {
            if (self->zst.avail_out == 0) {
                if (OutputBuffer_Grow(&buffer, &self->zst.next_out, &self->zst.avail_out) < 0) {
                    goto error;
                }
            }

            Py_BEGIN_ALLOW_THREADS
            err = deflate(&self->zst, Z_NO_FLUSH);
            Py_END_ALLOW_THREADS

            if (err == Z_STREAM_ERROR) {
                zlib_error(state, self->zst, err, "while compressing data");
                goto error;
            }

        } while (self->zst.avail_out == 0);
        assert(self->zst.avail_in == 0);

    } while (ibuflen != 0);

    return_value = OutputBuffer_Finish(&buffer, self->zst.avail_out);
    if (return_value != NULL) {
        goto success;
    }

 error:
    OutputBuffer_OnError(&buffer);
    return_value = NULL;
 success:
    PyMutex_Unlock(&self->mutex);
    return return_value;
}

/* Helper for objdecompress() and flush(). Saves any unconsumed input data in
   self->unused_data or self->unconsumed_tail, as appropriate. */
static int
save_unconsumed_input(compobject *self, Py_buffer *data, int err)
{
    if (err == Z_STREAM_END) {
        /* The end of the compressed data has been reached. Store the leftover
           input data in self->unused_data. */
        if (self->zst.avail_in > 0) {
            Py_ssize_t old_size = PyBytes_GET_SIZE(self->unused_data);
            Py_ssize_t left_size;
            left_size = (Byte *)data->buf + data->len - self->zst.next_in;
            if (left_size > (PY_SSIZE_T_MAX - old_size)) {
                PyErr_NoMemory();
                return -1;
            }
            PyBytesWriter *writer = PyBytesWriter_Create(old_size + left_size);
            if (writer == NULL) {
                return -1;
            }
            char *new_data = PyBytesWriter_GetData(writer);
            memcpy(new_data, PyBytes_AS_STRING(self->unused_data), old_size);
            memcpy(new_data + old_size, self->zst.next_in, left_size);
            PyObject *new_unused_data = PyBytesWriter_Finish(writer);
            if (new_unused_data == NULL) {
                return -1;
            }
            Py_SETREF(self->unused_data, new_unused_data);
            self->zst.next_in += left_size;
            self->zst.avail_in = 0;
        }
    }

    if (self->zst.avail_in > 0 || PyBytes_GET_SIZE(self->unconsumed_tail)) {
        /* This code handles two distinct cases:
           1. Output limit was reached. Save leftover input in unconsumed_tail.
           2. All input data was consumed. Clear unconsumed_tail. */
        Py_ssize_t left_size = (Byte *)data->buf + data->len - self->zst.next_in;
        PyObject *new_data = PyBytes_FromStringAndSize(
                (char *)self->zst.next_in, left_size);
        if (new_data == NULL)
            return -1;
        Py_SETREF(self->unconsumed_tail, new_data);
    }

    return 0;
}

/*[clinic input]
@permit_long_summary
zlib.Decompress.decompress

    cls: defining_class
    data: Py_buffer
        The binary data to decompress.
    /
    max_length: Py_ssize_t(allow_negative=False) = 0
        The maximum allowable length of the decompressed data.
        Unconsumed input data will be stored in
        the unconsumed_tail attribute.

Return a bytes object containing the decompressed version of the data.

After calling this function, some of the input data may still be
stored in internal buffers for later processing.
Call the flush() method to clear these buffers.
[clinic start generated code]*/

static PyObject *
zlib_Decompress_decompress_impl(compobject *self, PyTypeObject *cls,
                                Py_buffer *data, Py_ssize_t max_length)
/*[clinic end generated code: output=b024a93c2c922d57 input=9035027c9e4be7fd]*/
{
    int err = Z_OK;
    Py_ssize_t ibuflen;
    PyObject *return_value;
    _BlocksOutputBuffer buffer = {.writer = NULL};

    PyObject *module = PyType_GetModule(cls);
    if (module == NULL)
        return NULL;

    zlibstate *state = get_zlib_state(module);
    if (max_length == 0) {
        max_length = -1;
    }

    PyMutex_Lock(&self->mutex);

    self->zst.next_in = data->buf;
    ibuflen = data->len;

    if (OutputBuffer_InitAndGrow(&buffer, max_length, &self->zst.next_out, &self->zst.avail_out) < 0) {
        goto abort;
    }

    do {
        arrange_input_buffer(&self->zst, &ibuflen);

        do {
            if (self->zst.avail_out == 0) {
                if (OutputBuffer_GetDataSize(&buffer, self->zst.avail_out) == max_length) {
                    goto save;
                }
                if (OutputBuffer_Grow(&buffer, &self->zst.next_out, &self->zst.avail_out) < 0) {
                    goto abort;
                }
            }

            Py_BEGIN_ALLOW_THREADS
            err = inflate(&self->zst, Z_SYNC_FLUSH);
            Py_END_ALLOW_THREADS

            switch (err) {
            case Z_OK: _Py_FALLTHROUGH;
            case Z_BUF_ERROR: _Py_FALLTHROUGH;
            case Z_STREAM_END:
                break;
            default:
                if (err == Z_NEED_DICT && self->zdict != NULL) {
                    if (set_inflate_zdict(state, self) < 0) {
                        goto abort;
                    }
                    else
                        break;
                }
                goto save;
            }

        } while (self->zst.avail_out == 0 || err == Z_NEED_DICT);

    } while (err != Z_STREAM_END && ibuflen != 0);

 save:
    if (save_unconsumed_input(self, data, err) < 0)
        goto abort;

    if (err == Z_STREAM_END) {
        /* This is the logical place to call inflateEnd, but the old behaviour
           of only calling it on flush() is preserved. */
        FT_ATOMIC_STORE_CHAR_RELAXED(self->eof, 1);
    } else if (err != Z_OK && err != Z_BUF_ERROR) {
        /* We will only get Z_BUF_ERROR if the output buffer was full
           but there wasn't more output when we tried again, so it is
           not an error condition.
        */
        zlib_error(state, self->zst, err, "while decompressing data");
        goto abort;
    }

    return_value = OutputBuffer_Finish(&buffer, self->zst.avail_out);
    if (return_value != NULL) {
        goto success;
    }

 abort:
    OutputBuffer_OnError(&buffer);
    return_value = NULL;
 success:
    PyMutex_Unlock(&self->mutex);
    return return_value;
}

/*[clinic input]
zlib.Compress.flush

    cls: defining_class
    mode: int(c_default="Z_FINISH") = zlib.Z_FINISH
        One of the constants Z_SYNC_FLUSH, Z_FULL_FLUSH, Z_FINISH.
        If mode == Z_FINISH, the compressor object can no longer be
        used after calling the flush() method.  Otherwise, more data
        can still be compressed.
    /

Return a bytes object containing any remaining compressed data.
[clinic start generated code]*/

static PyObject *
zlib_Compress_flush_impl(compobject *self, PyTypeObject *cls, int mode)
/*[clinic end generated code: output=c7efd13efd62add2 input=286146e29442eb6c]*/
{
    int err;
    PyObject *return_value;
    _BlocksOutputBuffer buffer = {.writer = NULL};

    zlibstate *state = PyType_GetModuleState(cls);
    /* Flushing with Z_NO_FLUSH is a no-op, so there's no point in
       doing any work at all; just return an empty string. */
    if (mode == Z_NO_FLUSH) {
        return Py_GetConstant(Py_CONSTANT_EMPTY_BYTES);
    }

    PyMutex_Lock(&self->mutex);

    self->zst.avail_in = 0;

    if (OutputBuffer_InitAndGrow(&buffer, -1, &self->zst.next_out, &self->zst.avail_out) < 0) {
        goto error;
    }

    do {
        if (self->zst.avail_out == 0) {
            if (OutputBuffer_Grow(&buffer, &self->zst.next_out, &self->zst.avail_out) < 0) {
                goto error;
            }
        }

        Py_BEGIN_ALLOW_THREADS
        err = deflate(&self->zst, mode);
        Py_END_ALLOW_THREADS

        if (err == Z_STREAM_ERROR) {
            zlib_error(state, self->zst, err, "while flushing");
            goto error;
        }
    } while (self->zst.avail_out == 0);
    assert(self->zst.avail_in == 0);

    /* If mode is Z_FINISH, we also have to call deflateEnd() to free
       various data structures. Note we should only get Z_STREAM_END when
       mode is Z_FINISH, but checking both for safety*/
    if (err == Z_STREAM_END && mode == Z_FINISH) {
        err = deflateEnd(&self->zst);
        if (err != Z_OK) {
            zlib_error(state, self->zst, err, "while finishing compression");
            goto error;
        }
        else
            self->is_initialised = 0;

        /* We will only get Z_BUF_ERROR if the output buffer was full
           but there wasn't more output when we tried again, so it is
           not an error condition.
        */
    } else if (err != Z_OK && err != Z_BUF_ERROR) {
        zlib_error(state, self->zst, err, "while flushing");
        goto error;
    }

    return_value = OutputBuffer_Finish(&buffer, self->zst.avail_out);
    if (return_value != NULL) {
        goto success;
    }

error:
    OutputBuffer_OnError(&buffer);
    return_value = NULL;
success:
    PyMutex_Unlock(&self->mutex);
    return return_value;
}

#ifdef HAVE_ZLIB_COPY

/*[clinic input]
zlib.Compress.copy

    cls: defining_class

Return a copy of the compression object.
[clinic start generated code]*/

static PyObject *
zlib_Compress_copy_impl(compobject *self, PyTypeObject *cls)
/*[clinic end generated code: output=c4d2cfb4b0d7350b input=235497e482d40986]*/
{
    zlibstate *state = PyType_GetModuleState(cls);

    compobject *return_value = newcompobject(state->Comptype);
    if (!return_value) return NULL;

    /* Copy the zstream state
     * We use mutex to make this thread-safe
     */
    PyMutex_Lock(&self->mutex);
    int err = deflateCopy(&return_value->zst, &self->zst);
    switch (err) {
    case Z_OK:
        break;
    case Z_STREAM_ERROR:
        PyErr_SetString(PyExc_ValueError, "Inconsistent stream state");
        goto error;
    case Z_MEM_ERROR:
        PyErr_SetString(PyExc_MemoryError,
                        "Can't allocate memory for compression object");
        goto error;
    default:
        zlib_error(state, self->zst, err, "while copying compression object");
        goto error;
    }
    Py_XSETREF(return_value->unused_data, Py_NewRef(self->unused_data));
    Py_XSETREF(return_value->unconsumed_tail, Py_NewRef(self->unconsumed_tail));
    Py_XSETREF(return_value->zdict, Py_XNewRef(self->zdict));
    return_value->eof = self->eof;

    /* Mark it as being initialized */
    return_value->is_initialised = 1;

    PyMutex_Unlock(&self->mutex);
    return (PyObject *)return_value;

error:
    PyMutex_Unlock(&self->mutex);
    Py_XDECREF(return_value);
    return NULL;
}

/*[clinic input]
zlib.Compress.__copy__

    cls: defining_class

[clinic start generated code]*/

static PyObject *
zlib_Compress___copy___impl(compobject *self, PyTypeObject *cls)
/*[clinic end generated code: output=074613db332cb668 input=5c0188367ab0fe64]*/
{
    return zlib_Compress_copy_impl(self, cls);
}

/*[clinic input]
zlib.Compress.__deepcopy__

    cls: defining_class
    memo: object
    /

[clinic start generated code]*/

static PyObject *
zlib_Compress___deepcopy___impl(compobject *self, PyTypeObject *cls,
                                PyObject *memo)
/*[clinic end generated code: output=24b3aed785f54033 input=c90347319a514430]*/
{
    return zlib_Compress_copy_impl(self, cls);
}

/*[clinic input]
zlib.Decompress.copy

    cls: defining_class

Return a copy of the decompression object.
[clinic start generated code]*/

static PyObject *
zlib_Decompress_copy_impl(compobject *self, PyTypeObject *cls)
/*[clinic end generated code: output=a7ddc016e1d0a781 input=20ef3aa208282ff2]*/
{
    zlibstate *state = PyType_GetModuleState(cls);

    compobject *return_value = newcompobject(state->Decomptype);
    if (!return_value) return NULL;

    /* Copy the zstream state
     * We use mutex to make this thread-safe
     */
    PyMutex_Lock(&self->mutex);
    int err = inflateCopy(&return_value->zst, &self->zst);
    switch (err) {
    case Z_OK:
        break;
    case Z_STREAM_ERROR:
        PyErr_SetString(PyExc_ValueError, "Inconsistent stream state");
        goto error;
    case Z_MEM_ERROR:
        PyErr_SetString(PyExc_MemoryError,
                        "Can't allocate memory for decompression object");
        goto error;
    default:
        zlib_error(state, self->zst, err, "while copying decompression object");
        goto error;
    }

    Py_XSETREF(return_value->unused_data, Py_NewRef(self->unused_data));
    Py_XSETREF(return_value->unconsumed_tail, Py_NewRef(self->unconsumed_tail));
    Py_XSETREF(return_value->zdict, Py_XNewRef(self->zdict));
    return_value->eof = self->eof;

    /* Mark it as being initialized */
    return_value->is_initialised = 1;

    PyMutex_Unlock(&self->mutex);
    return (PyObject *)return_value;

error:
    PyMutex_Unlock(&self->mutex);
    Py_XDECREF(return_value);
    return NULL;
}

/*[clinic input]
zlib.Decompress.__copy__

    cls: defining_class

[clinic start generated code]*/

static PyObject *
zlib_Decompress___copy___impl(compobject *self, PyTypeObject *cls)
/*[clinic end generated code: output=cf1e6473744f53fa input=cc3143067b622bdf]*/
{
    return zlib_Decompress_copy_impl(self, cls);
}

/*[clinic input]
zlib.Decompress.__deepcopy__

    cls: defining_class
    memo: object
    /

[clinic start generated code]*/

static PyObject *
zlib_Decompress___deepcopy___impl(compobject *self, PyTypeObject *cls,
                                  PyObject *memo)
/*[clinic end generated code: output=34f7b719a0c0d51b input=fc13b9c58622544e]*/
{
    return zlib_Decompress_copy_impl(self, cls);
}

#endif

/*[clinic input]
zlib.Decompress.flush

    cls: defining_class
    length: Py_ssize_t(c_default="DEF_BUF_SIZE") = zlib.DEF_BUF_SIZE
        the initial size of the output buffer.
    /

Return a bytes object containing any remaining decompressed data.
[clinic start generated code]*/

static PyObject *
zlib_Decompress_flush_impl(compobject *self, PyTypeObject *cls,
                           Py_ssize_t length)
/*[clinic end generated code: output=4532fc280bd0f8f2 input=42f1f4b75230e2cd]*/
{
    int err, flush;
    Py_buffer data;
    PyObject *return_value;
    Py_ssize_t ibuflen;
    _BlocksOutputBuffer buffer = {.writer = NULL};
    _Uint32Window window;  // output buffer's UINT32_MAX sliding window

    PyObject *module = PyType_GetModule(cls);
    if (module == NULL) {
        return NULL;
    }

    zlibstate *state = get_zlib_state(module);

    if (length <= 0) {
        PyErr_SetString(PyExc_ValueError, "length must be greater than zero");
        return NULL;
    }

    PyMutex_Lock(&self->mutex);

    /* A previous flush() already reached the end of the stream and freed the
       decompression state, so there is nothing left to process. */
    if (!self->is_initialised) {
        PyMutex_Unlock(&self->mutex);
        return Py_GetConstant(Py_CONSTANT_EMPTY_BYTES);
    }

    if (PyObject_GetBuffer(self->unconsumed_tail, &data, PyBUF_SIMPLE) == -1) {
        PyMutex_Unlock(&self->mutex);
        return NULL;
    }

    self->zst.next_in = data.buf;
    ibuflen = data.len;

    if (OutputBuffer_WindowInitWithSize(&buffer, &window, length,
                                        &self->zst.next_out, &self->zst.avail_out) < 0) {
        goto abort;
    }

    do {
        arrange_input_buffer(&self->zst, &ibuflen);
        flush = ibuflen == 0 ? Z_FINISH : Z_NO_FLUSH;

        do {
            if (self->zst.avail_out == 0) {
                if (OutputBuffer_WindowGrow(&buffer, &window,
                                            &self->zst.next_out, &self->zst.avail_out) < 0) {
                    goto abort;
                }
            }

            Py_BEGIN_ALLOW_THREADS
            err = inflate(&self->zst, flush);
            Py_END_ALLOW_THREADS

            switch (err) {
            case Z_OK: _Py_FALLTHROUGH;
            case Z_BUF_ERROR: _Py_FALLTHROUGH;
            case Z_STREAM_END:
                break;
            default:
                goto save;
            }

        } while (self->zst.avail_out == 0 || err == Z_NEED_DICT);

    } while (err != Z_STREAM_END && ibuflen != 0);

 save:
    if (save_unconsumed_input(self, &data, err) < 0) {
        goto abort;
    }

    /* If at end of stream, clean up any memory allocated by zlib. */
    if (err == Z_STREAM_END) {
        FT_ATOMIC_STORE_CHAR_RELAXED(self->eof, 1);
        self->is_initialised = 0;
        err = inflateEnd(&self->zst);
        if (err != Z_OK) {
            zlib_error(state, self->zst, err, "while finishing decompression");
            goto abort;
        }
    }
    else if (err != Z_OK && err != Z_BUF_ERROR) {
        zlib_error(state, self->zst, err, "while decompressing data");
        goto abort;
    }

    return_value = OutputBuffer_WindowFinish(&buffer, &window, self->zst.avail_out);
    if (return_value != NULL) {
        goto success;
    }

 abort:
    OutputBuffer_WindowOnError(&buffer, &window);
    return_value = NULL;
 success:
    PyBuffer_Release(&data);
    PyMutex_Unlock(&self->mutex);
    return return_value;
}


typedef struct {
    PyObject_HEAD
    z_stream zst;
    PyObject *zdict;
    PyMutex mutex;
    PyObject *unused_data;
    uint8_t *input_buffer;
    Py_ssize_t input_buffer_size;
    /* zst>avail_in is only 32 bit, so we store the true length
       separately. Conversion and looping is encapsulated in
       decompress_buf() */
    Py_ssize_t avail_in_real;
    bool is_initialised;
    char eof;           /* Py_T_BOOL expects a char */
    char needs_input;
} ZlibDecompressor;

#define ZlibDecompressor_CAST(op)   ((ZlibDecompressor *)(op))

/*[clinic input]
class zlib._ZlibDecompressor "ZlibDecompressor *" "&ZlibDecompressorType"
[clinic start generated code]*/
/*[clinic end generated code: output=da39a3ee5e6b4b0d input=49151d1d703e6bcc]*/

static void
ZlibDecompressor_dealloc(PyObject *op)
{
    PyTypeObject *type = Py_TYPE(op);
    PyObject_GC_UnTrack(op);
    ZlibDecompressor *self = ZlibDecompressor_CAST(op);
    assert(!PyMutex_IsLocked(&self->mutex));
    if (self->is_initialised) {
        inflateEnd(&self->zst);
    }
    PyMem_Free(self->input_buffer);
    Py_CLEAR(self->unused_data);
    Py_CLEAR(self->zdict);
    type->tp_free(self);
    Py_DECREF(type);
}

static int
ZlibDecompressor_traverse(PyObject *op, visitproc visit, void *arg)
{
    ZlibDecompressor *self = ZlibDecompressor_CAST(op);
    Py_VISIT(Py_TYPE(op));
    Py_VISIT(self->zdict);
    return 0;
}

static int
set_inflate_zdict_ZlibDecompressor(zlibstate *state, ZlibDecompressor *self)
{
    Py_buffer zdict_buf;
    if (PyObject_GetBuffer(self->zdict, &zdict_buf, PyBUF_SIMPLE) == -1) {
        return -1;
    }
    if ((size_t)zdict_buf.len > UINT_MAX) {
        PyErr_SetString(PyExc_OverflowError,
                        "zdict length does not fit in an unsigned int");
        PyBuffer_Release(&zdict_buf);
        return -1;
    }
    int err;
    err = inflateSetDictionary(&self->zst,
                               zdict_buf.buf, (unsigned int)zdict_buf.len);
    PyBuffer_Release(&zdict_buf);
    if (err != Z_OK) {
        zlib_error(state, self->zst, err, "while setting zdict");
        return -1;
    }
    return 0;
}

static Py_ssize_t
arrange_output_buffer_with_maximum(uint32_t *avail_out,
                                   uint8_t **next_out,
                                   PyObject **buffer,
                                   Py_ssize_t length,
                                   Py_ssize_t max_length)
{
    Py_ssize_t occupied;

    if (*buffer == NULL) {
        if (!(*buffer = PyBytes_FromStringAndSize(NULL, length)))
            return -1;
        occupied = 0;
    }
    else {
        occupied = *next_out - (uint8_t *)PyBytes_AS_STRING(*buffer);

        if (length == occupied) {
            Py_ssize_t new_length;
            assert(length <= max_length);
            /* can not scale the buffer over max_length */
            if (length == max_length)
                return -2;
            if (length <= (max_length >> 1))
                new_length = length << 1;
            else
                new_length = max_length;
            if (_PyBytes_Resize(buffer, new_length) < 0)
                return -1;
            length = new_length;
        }
    }

    *avail_out = (uint32_t)Py_MIN((size_t)(length - occupied), UINT32_MAX);
    *next_out = (uint8_t *)PyBytes_AS_STRING(*buffer) + occupied;

    return length;
}

/* Decompress data of length self->avail_in_real in self->state.next_in. The
   output buffer is allocated dynamically and returned. If the max_length is
   of sufficiently low size, max_length is allocated immediately. At most
   max_length bytes are returned, so some of the input may not be consumed.
   self->state.next_in and self->avail_in_real are updated to reflect the
   consumed input. */
static PyObject*
decompress_buf(ZlibDecompressor *self, Py_ssize_t max_length)
{
    /* data_size is strictly positive, but because we repeatedly have to
       compare against max_length and PyBytes_GET_SIZE we declare it as
       signed */
    PyObject *return_value = NULL;
    Py_ssize_t hard_limit;
    Py_ssize_t obuflen;
    zlibstate *state = PyType_GetModuleState(Py_TYPE(self));

    int err = Z_OK;

    /* When sys.maxsize is passed as default use DEF_BUF_SIZE as start buffer.
       In this particular case the data may not necessarily be very big, so
       it is better to grow dynamically.*/
    if ((max_length < 0) || max_length == PY_SSIZE_T_MAX) {
        hard_limit = PY_SSIZE_T_MAX;
        obuflen = DEF_BUF_SIZE;
    } else {
        /* Assume that decompressor is used in file decompression with a fixed
           block size of max_length. In that case we will reach max_length almost
           always (except at the end of the file). So it makes sense to allocate
           max_length. */
        hard_limit = max_length;
        obuflen = max_length;
        if (obuflen > DEF_MAX_INITIAL_BUF_SIZE){
            // Safeguard against memory overflow.
            obuflen = DEF_MAX_INITIAL_BUF_SIZE;
        }
    }

    do {
        arrange_input_buffer(&(self->zst), &(self->avail_in_real));

        do {
            obuflen = arrange_output_buffer_with_maximum(&(self->zst.avail_out),
                                                        &(self->zst.next_out),
                                                        &return_value,
                                                        obuflen,
                                                        hard_limit);
            if (obuflen == -1){
                PyErr_SetString(PyExc_MemoryError,
                                "Insufficient memory for buffer allocation");
                goto error;
            }
            else if (obuflen == -2) {
                break;
            }
            Py_BEGIN_ALLOW_THREADS
            err = inflate(&self->zst, Z_SYNC_FLUSH);
            Py_END_ALLOW_THREADS
            switch (err) {
            case Z_OK:  _Py_FALLTHROUGH;
            case Z_BUF_ERROR: _Py_FALLTHROUGH;
            case Z_STREAM_END:
                break;
            default:
                if (err == Z_NEED_DICT) {
                    goto error;
                }
                else {
                    break;
                }
            }
        } while (self->zst.avail_out == 0);
    } while(err != Z_STREAM_END && self->avail_in_real != 0);

    if (err == Z_STREAM_END) {
        FT_ATOMIC_STORE_CHAR_RELAXED(self->eof, 1);
        self->is_initialised = 0;
        /* Unlike the Decompress object we call inflateEnd here as there are no
           backwards compatibility issues */
        err = inflateEnd(&self->zst);
        if (err != Z_OK) {
            zlib_error(state, self->zst, err, "while finishing decompression");
            goto error;
        }
    } else if (err != Z_OK && err != Z_BUF_ERROR) {
        zlib_error(state, self->zst, err, "while decompressing data");
        goto error;
    }

    self->avail_in_real += self->zst.avail_in;

    if (_PyBytes_Resize(&return_value, self->zst.next_out -
                        (uint8_t *)PyBytes_AS_STRING(return_value)) != 0) {
        goto error;
    }

    goto success;
error:
    Py_CLEAR(return_value);
success:
    return return_value;
}


static PyObject *
decompress(ZlibDecompressor *self, uint8_t *data,
           size_t len, Py_ssize_t max_length)
{
    bool input_buffer_in_use;
    PyObject *result;

    /* Prepend unconsumed input if necessary */
    if (self->zst.next_in != NULL) {
        size_t avail_now, avail_total;

        /* Number of bytes we can append to input buffer */
        avail_now = (self->input_buffer + self->input_buffer_size)
            - (self->zst.next_in + self->avail_in_real);

        /* Number of bytes we can append if we move existing
           contents to beginning of buffer (overwriting
           consumed input) */
        avail_total = self->input_buffer_size - self->avail_in_real;

        if (avail_total < len) {
            size_t offset = self->zst.next_in - self->input_buffer;
            uint8_t *tmp;
            size_t new_size = self->input_buffer_size + len - avail_now;

            /* Assign to temporary variable first, so we don't
               lose address of allocated buffer if realloc fails */
            tmp = PyMem_Realloc(self->input_buffer, new_size);
            if (tmp == NULL) {
                PyErr_SetNone(PyExc_MemoryError);
                return NULL;
            }
            self->input_buffer = tmp;
            self->input_buffer_size = new_size;

            self->zst.next_in = self->input_buffer + offset;
        }
        else if (avail_now < len) {
            memmove(self->input_buffer, self->zst.next_in,
                    self->avail_in_real);
            self->zst.next_in = self->input_buffer;
        }
        memcpy((void*)(self->zst.next_in + self->avail_in_real), data, len);
        self->avail_in_real += len;
        input_buffer_in_use = 1;
    }
    else {
        self->zst.next_in = data;
        self->avail_in_real = len;
        input_buffer_in_use = 0;
    }

    result = decompress_buf(self, max_length);
    if(result == NULL) {
        self->zst.next_in = NULL;
        return NULL;
    }

    if (self->eof) {
        FT_ATOMIC_STORE_CHAR_RELAXED(self->needs_input, 0);

        if (self->avail_in_real > 0) {
            PyObject *unused_data = PyBytes_FromStringAndSize(
                (char *)self->zst.next_in, self->avail_in_real);
            if (unused_data == NULL) {
                goto error;
            }
            Py_XSETREF(self->unused_data, unused_data);
        }
    }
    else if (self->avail_in_real == 0) {
        self->zst.next_in = NULL;
        FT_ATOMIC_STORE_CHAR_RELAXED(self->needs_input, 1);
    }
    else {
        FT_ATOMIC_STORE_CHAR_RELAXED(self->needs_input, 0);

        /* If we did not use the input buffer, we now have
           to copy the tail from the caller's buffer into the
           input buffer */
        if (!input_buffer_in_use) {

            /* Discard buffer if it's too small
               (resizing it may needlessly copy the current contents) */
            if (self->input_buffer != NULL &&
                self->input_buffer_size < self->avail_in_real) {
                PyMem_Free(self->input_buffer);
                self->input_buffer = NULL;
            }

            /* Allocate if necessary */
            if (self->input_buffer == NULL) {
                self->input_buffer = PyMem_Malloc(self->avail_in_real);
                if (self->input_buffer == NULL) {
                    PyErr_SetNone(PyExc_MemoryError);
                    goto error;
                }
                self->input_buffer_size = self->avail_in_real;
            }

            /* Copy tail */
            memcpy(self->input_buffer, self->zst.next_in, self->avail_in_real);
            self->zst.next_in = self->input_buffer;
        }
    }
    return result;

error:
    self->zst.next_in = NULL;
    Py_XDECREF(result);
    return NULL;
}

/*[clinic input]
zlib._ZlibDecompressor.decompress

    data: Py_buffer
    max_length: Py_ssize_t=-1

Decompress *data*, returning uncompressed data as bytes.

If *max_length* is nonnegative, returns at most *max_length* bytes
of decompressed data.  If this limit is reached and further output
can be produced, *self.needs_input* will be set to ``False``.  In
this case, the next call to *decompress()* may provide *data* as b''
to obtain more of the output.

If all of the input data was decompressed and returned (either
because this was less than *max_length* bytes, or because
*max_length* was negative), *self.needs_input* will be set to True.

Attempting to decompress data after the end of stream is reached
raises an EOFError.  Any data found after the end of the stream is
ignored and saved in the unused_data attribute.
[clinic start generated code]*/

static PyObject *
zlib__ZlibDecompressor_decompress_impl(ZlibDecompressor *self,
                                       Py_buffer *data,
                                       Py_ssize_t max_length)
/*[clinic end generated code: output=ac00dcf73e843e99 input=d7862eade3f29d56]*/

{
    PyObject *result = NULL;

    PyMutex_Lock(&self->mutex);
    if (self->eof) {
        PyErr_SetString(PyExc_EOFError, "End of stream already reached");
    }
    else {
        result = decompress(self, data->buf, data->len, max_length);
    }
    PyMutex_Unlock(&self->mutex);
    return result;
}

/*[clinic input]
@classmethod
zlib._ZlibDecompressor.__new__

    wbits: int(c_default='MAX_WBITS') = MAX_WBITS
    zdict: object(c_default='NULL') = b''
        The predefined compression dictionary. This is a sequence of bytes
        (such as a bytes object) containing subsequences that are expected
        to occur frequently in the data that is to be compressed. Those
        subsequences that are expected to be most common should come at the
        end of the dictionary. This must be the same dictionary as used by the
        compressor that produced the input data.

Create a decompressor object for decompressing data incrementally.
[clinic start generated code]*/

static PyObject *
zlib__ZlibDecompressor_impl(PyTypeObject *type, int wbits, PyObject *zdict)
/*[clinic end generated code: output=1065607df0d33baa input=9ebad0be6de226e2]*/
{
    assert(type != NULL && type->tp_alloc != NULL);
    zlibstate *state = PyType_GetModuleState(type);
    ZlibDecompressor *self = ZlibDecompressor_CAST(type->tp_alloc(type, 0));
    if (self == NULL) {
        return NULL;
    }
    self->eof = 0;
    self->needs_input = 1;
    self->avail_in_real = 0;
    self->input_buffer = NULL;
    self->input_buffer_size = 0;
    self->zdict = Py_XNewRef(zdict);
    self->zst.opaque = NULL;
    self->zst.zalloc = PyZlib_Malloc;
    self->zst.zfree = PyZlib_Free;
    self->zst.next_in = NULL;
    self->zst.avail_in = 0;
    self->unused_data = Py_GetConstant(Py_CONSTANT_EMPTY_BYTES);
    self->mutex = (PyMutex){0};
    int err = inflateInit2(&(self->zst), wbits);
    switch (err) {
        case Z_OK:
        self->is_initialised = 1;
        if (self->zdict != NULL && wbits < 0) {
            if (set_inflate_zdict_ZlibDecompressor(state, self) < 0) {
                Py_DECREF(self);
                return NULL;
            }
        }
        return (PyObject *)self;
    case Z_STREAM_ERROR:
        Py_DECREF(self);
        PyErr_SetString(PyExc_ValueError, "Invalid initialization option");
        return NULL;
    case Z_MEM_ERROR:
        Py_DECREF(self);
        PyErr_SetString(PyExc_MemoryError,
                        "Can't allocate memory for decompression object");
        return NULL;
    default:
        zlib_error(state, self->zst, err, "while creating decompression object");
        Py_DECREF(self);
        return NULL;
    }
}

#include "clinic/zlibmodule.c.h"

static PyMethodDef comp_methods[] =
{
    ZLIB_COMPRESS_COMPRESS_METHODDEF
    ZLIB_COMPRESS_FLUSH_METHODDEF
    ZLIB_COMPRESS_COPY_METHODDEF
    ZLIB_COMPRESS___COPY___METHODDEF
    ZLIB_COMPRESS___DEEPCOPY___METHODDEF
    {NULL, NULL}
};

static PyMethodDef Decomp_methods[] =
{
    ZLIB_DECOMPRESS_DECOMPRESS_METHODDEF
    ZLIB_DECOMPRESS_FLUSH_METHODDEF
    ZLIB_DECOMPRESS_COPY_METHODDEF
    ZLIB_DECOMPRESS___COPY___METHODDEF
    ZLIB_DECOMPRESS___DEEPCOPY___METHODDEF
    {NULL, NULL}
};

static PyMethodDef ZlibDecompressor_methods[] = {
    ZLIB__ZLIBDECOMPRESSOR_DECOMPRESS_METHODDEF
    {NULL}
};

static PyObject *
Decomp_unused_data_get(PyObject *op, void *Py_UNUSED(ignored))
{
    compobject *self = _compobject_CAST(op);
    PyMutex_Lock(&self->mutex);
    assert(self->unused_data != NULL);
    PyObject *result = Py_NewRef(self->unused_data);
    PyMutex_Unlock(&self->mutex);
    return result;
}

static PyObject *
Decomp_unconsumed_tail_get(PyObject *op, void *Py_UNUSED(ignored))
{
    compobject *self = _compobject_CAST(op);
    PyMutex_Lock(&self->mutex);
    assert(self->unconsumed_tail != NULL);
    PyObject *result = Py_NewRef(self->unconsumed_tail);
    PyMutex_Unlock(&self->mutex);
    return result;
}

static PyGetSetDef Decomp_getset[] = {
    {"unused_data", Decomp_unused_data_get, NULL, NULL},
    {"unconsumed_tail", Decomp_unconsumed_tail_get, NULL, NULL},
    {NULL},
};

#define COMP_OFF(x) offsetof(compobject, x)
static PyMemberDef Decomp_members[] = {
    {"eof",             Py_T_BOOL,   COMP_OFF(eof), Py_READONLY},
    {NULL},
};

PyDoc_STRVAR(ZlibDecompressor_eof__doc__,
"True if the end-of-stream marker has been reached.");

PyDoc_STRVAR(ZlibDecompressor_unused_data__doc__,
"Data found after the end of the compressed stream.");

PyDoc_STRVAR(ZlibDecompressor_needs_input_doc,
"True if more input is needed before more decompressed data can be produced.");

static PyObject *
ZlibDecompressor_unused_data_get(PyObject *op, void *Py_UNUSED(ignored))
{
    ZlibDecompressor *self = ZlibDecompressor_CAST(op);
    PyMutex_Lock(&self->mutex);
    assert(self->unused_data != NULL);
    PyObject *result = Py_NewRef(self->unused_data);
    PyMutex_Unlock(&self->mutex);
    return result;
}

static PyGetSetDef ZlibDecompressor_getset[] = {
    {"unused_data", ZlibDecompressor_unused_data_get, NULL,
     ZlibDecompressor_unused_data__doc__},
    {NULL},
};

static PyMemberDef ZlibDecompressor_members[] = {
    {"eof", Py_T_BOOL, offsetof(ZlibDecompressor, eof),
     Py_READONLY, ZlibDecompressor_eof__doc__},
    {"needs_input", Py_T_BOOL, offsetof(ZlibDecompressor, needs_input), Py_READONLY,
     ZlibDecompressor_needs_input_doc},
    {NULL},
};


/*[clinic input]
zlib.adler32

    data: Py_buffer
    value: unsigned_int(bitwise=True) = 1
        Starting value of the checksum.
    /

Compute an Adler-32 checksum of data.

The returned checksum is an integer.
[clinic start generated code]*/

static PyObject *
zlib_adler32_impl(PyObject *module, Py_buffer *data, unsigned int value)
/*[clinic end generated code: output=422106f5ca8c92c0 input=6ff4557872160e88]*/
{
    /* Releasing the GIL for very small buffers is inefficient
       and may lower performance */
    if (data->len > 1024*5) {
        unsigned char *buf = data->buf;
        Py_ssize_t len = data->len;

        Py_BEGIN_ALLOW_THREADS
        /* Avoid truncation of length for very large buffers. adler32() takes
           length as an unsigned int, which may be narrower than Py_ssize_t. */
        while ((size_t)len > UINT_MAX) {
            value = adler32(value, buf, UINT_MAX);
            buf += (size_t) UINT_MAX;
            len -= (size_t) UINT_MAX;
        }
        value = adler32(value, buf, (unsigned int)len);
        Py_END_ALLOW_THREADS
    } else {
        value = adler32(value, data->buf, (unsigned int)data->len);
    }
    return PyLong_FromUnsignedLong(value & 0xffffffffU);
}

/*[clinic input]
zlib.adler32_combine -> unsigned_int

    adler1: unsigned_int(bitwise=True)
        Adler-32 checksum for sequence A

    adler2: unsigned_int(bitwise=True)
        Adler-32 checksum for sequence B

    len2: object(subclass_of='&PyLong_Type')
        Length of sequence B
    /

Combine two Adler-32 checksums into one.

Given the Adler-32 checksum 'adler1' of a sequence A and the
Adler-32 checksum 'adler2' of a sequence B of length 'len2',
return the Adler-32 checksum of A and B concatenated.
[clinic start generated code]*/

static unsigned int
zlib_adler32_combine_impl(PyObject *module, unsigned int adler1,
                          unsigned int adler2, PyObject *len2)
/*[clinic end generated code: output=61842cefb16afb1b input=51bb045c95130c6f]*/
{
#if defined(Z_WANT64)
    z_off64_t len = convert_to_z_off_t(len2);
#else
    z_off_t len = convert_to_z_off_t(len2);
#endif
    if (len < 0) {
        if (!PyErr_Occurred()) {
            PyErr_SetString(PyExc_ValueError, "len2 must be non-negative");
        }
        return (unsigned int)-1;
    }
    return adler32_combine(adler1, adler2, len);
}


/*[clinic input]
zlib.crc32 -> unsigned_int

    data: Py_buffer
    value: unsigned_int(bitwise=True) = 0
        Starting value of the checksum.
    /

Compute a CRC-32 checksum of data.

The returned checksum is an integer.
[clinic start generated code]*/

static unsigned int
zlib_crc32_impl(PyObject *module, Py_buffer *data, unsigned int value)
/*[clinic end generated code: output=b217562e4fe6d6a6 input=1229cb2fb5ea948a]*/
{
    /* Releasing the GIL for very small buffers is inefficient
       and may lower performance */
    if (data->len > 1024*5) {
        unsigned char *buf = data->buf;
        Py_ssize_t len = data->len;

        Py_BEGIN_ALLOW_THREADS
        /* Avoid truncation of length for very large buffers. crc32() takes
           length as an unsigned int, which may be narrower than Py_ssize_t.
           We further limit size due to bugs in Apple's macOS zlib.
           See https://github.com/python/cpython/issues/105967.
         */
#define ZLIB_CRC_CHUNK_SIZE 0x40000000
#if ZLIB_CRC_CHUNK_SIZE > INT_MAX
# error "unsupported less than 32-bit platform?"
#endif
        while ((size_t)len > ZLIB_CRC_CHUNK_SIZE) {
            value = crc32(value, buf, ZLIB_CRC_CHUNK_SIZE);
            buf += (size_t) ZLIB_CRC_CHUNK_SIZE;
            len -= (size_t) ZLIB_CRC_CHUNK_SIZE;
        }
#undef ZLIB_CRC_CHUNK_SIZE
        value = crc32(value, buf, (unsigned int)len);
        Py_END_ALLOW_THREADS
    } else {
        value = crc32(value, data->buf, (unsigned int)data->len);
    }
    return value;
}

/*[clinic input]
zlib.crc32_combine -> unsigned_int

    crc1: unsigned_int(bitwise=True)
        CRC-32 checksum for sequence A

    crc2: unsigned_int(bitwise=True)
        CRC-32 checksum for sequence B

    len2: object(subclass_of='&PyLong_Type')
        Length of sequence B
    /

Combine two CRC-32 checksums into one.

Given the CRC-32 checksum 'crc1' of a sequence A and the
CRC-32 checksum 'crc2' of a sequence B of length 'len2',
return the CRC-32 checksum of A and B concatenated.
[clinic start generated code]*/

static unsigned int
zlib_crc32_combine_impl(PyObject *module, unsigned int crc1,
                        unsigned int crc2, PyObject *len2)
/*[clinic end generated code: output=c4def907c602e6eb input=9c8a065d9040dc66]*/
{
#if defined(Z_WANT64)
    z_off64_t len = convert_to_z_off_t(len2);
#else
    z_off_t len = convert_to_z_off_t(len2);
#endif
    if (len < 0) {
        if (!PyErr_Occurred()) {
            PyErr_SetString(PyExc_ValueError, "len2 must be non-negative");
        }
        return (unsigned int)-1;
    }
    return crc32_combine(crc1, crc2, len);
}

static PyObject *
zlib_getattr(PyObject *self, PyObject *args)
{
    PyObject *name;
    if (!PyArg_UnpackTuple(args, "__getattr__", 1, 1, &name)) {
        return NULL;
    }

    if (PyUnicode_Check(name) && PyUnicode_EqualToUTF8(name, "__version__")) {
        if (PyErr_WarnEx(PyExc_DeprecationWarning,
                         "'__version__' is deprecated and slated for removal in Python 3.20",
                         1) < 0) {
            return NULL;
        }
        return PyUnicode_FromString("1.0");
    }

    PyErr_Format(PyExc_AttributeError, "module 'zlib' has no attribute %R", name);
    return NULL;
}

PyDoc_STRVAR(zlib_version_info__doc__,
"zlib.zlib_version_info\n\
\n\
Zlib version information as a named tuple.");

static PyStructSequence_Field zlib_version_info_fields[] = {
    {"major", "Major release number"},
    {"minor", "Minor release number"},
    {"revision", "Revision release number"},
    {"subversion", "Subversion release number"},
    {0}
};

static PyStructSequence_Desc zlib_version_info_desc = {
    "zlib.zlib_version_info",        /* name */
    zlib_version_info__doc__,        /* doc */
    zlib_version_info_fields,        /* fields */
    4
};

#ifdef ZLIBNG_VERSION
PyDoc_STRVAR(zlibng_version_info__doc__,
"zlib.zlibng_version_info\n\
\n\
Zlib-ng version information as a named tuple.");

static PyStructSequence_Field zlibng_version_info_fields[] = {
    {"major", "Major release number"},
    {"minor", "Minor release number"},
    {"revision", "Revision release number"},
    {0}
};

static PyStructSequence_Desc zlibng_version_info_desc = {
    "zlib.zlibng_version_info",        /* name */
    zlibng_version_info__doc__,        /* doc */
    zlibng_version_info_fields,        /* fields */
    3
};
#endif // ZLIBNG_VERSION

/* Create a named tuple from the first *size* components of a version string
   like "1.2.11" or "1.2.11.1".  sscanf() is expected to fail on trailing
   garbage and on versions with fewer components (for example "1.2.0.f" or
   "1.3.1.zlib-ng"); the components which were not parsed are left zero.
   This is deliberate -- a zero is more useful here than a hard error. */
static PyObject *
make_version_info(PyTypeObject *type, const char *string, Py_ssize_t size)
{
    unsigned int components[4] = {0, 0, 0, 0};
    assert(size <= (Py_ssize_t)Py_ARRAY_LENGTH(components));

    sscanf(string, "%u.%u.%u.%u",
           &components[0], &components[1], &components[2], &components[3]);

    PyObject *version = PyStructSequence_New(type);
    if (version == NULL) {
        return NULL;
    }
    for (Py_ssize_t i = 0; i < size; i++) {
        PyObject *item = PyLong_FromUnsignedLong(components[i]);
        if (item == NULL) {
            Py_DECREF(version);
            return NULL;
        }
        PyStructSequence_SET_ITEM(version, i, item);
    }
    return version;
}


static PyMethodDef zlib_methods[] =
{
    ZLIB_ADLER32_METHODDEF
    ZLIB_ADLER32_COMBINE_METHODDEF
    ZLIB_COMPRESS_METHODDEF
    ZLIB_COMPRESSOBJ_METHODDEF
    ZLIB_CRC32_METHODDEF
    ZLIB_CRC32_COMBINE_METHODDEF
    ZLIB_DECOMPRESS_METHODDEF
    ZLIB_DECOMPRESSOBJ_METHODDEF
    {"__getattr__", zlib_getattr, METH_VARARGS, "Module __getattr__"},
    {NULL, NULL}
};

static PyType_Slot Comptype_slots[] = {
    {Py_tp_dealloc, Comp_dealloc},
    {Py_tp_traverse, compobject_traverse},
    {Py_tp_methods, comp_methods},
    {0, 0},
};

static PyType_Spec Comptype_spec = {
    .name = "zlib.Compress",
    .basicsize = sizeof(compobject),
    .flags = (
        Py_TPFLAGS_DEFAULT
        | Py_TPFLAGS_DISALLOW_INSTANTIATION
        | Py_TPFLAGS_IMMUTABLETYPE
        | Py_TPFLAGS_HAVE_GC
    ),
    .slots= Comptype_slots,
};

static PyType_Slot Decomptype_slots[] = {
    {Py_tp_dealloc, Decomp_dealloc},
    {Py_tp_traverse, compobject_traverse},
    {Py_tp_methods, Decomp_methods},
    {Py_tp_members, Decomp_members},
    {Py_tp_getset, Decomp_getset},
    {0, 0},
};

static PyType_Spec Decomptype_spec = {
    .name = "zlib.Decompress",
    .basicsize = sizeof(compobject),
    .flags = (
        Py_TPFLAGS_DEFAULT
        | Py_TPFLAGS_DISALLOW_INSTANTIATION
        | Py_TPFLAGS_IMMUTABLETYPE
        | Py_TPFLAGS_HAVE_GC
    ),
    .slots = Decomptype_slots,
};

static PyType_Slot ZlibDecompressor_type_slots[] = {
    {Py_tp_dealloc, ZlibDecompressor_dealloc},
    {Py_tp_traverse, ZlibDecompressor_traverse},
    {Py_tp_members, ZlibDecompressor_members},
    {Py_tp_getset, ZlibDecompressor_getset},
    {Py_tp_new, zlib__ZlibDecompressor},
    {Py_tp_doc, (char *)zlib__ZlibDecompressor__doc__},
    {Py_tp_methods, ZlibDecompressor_methods},
    {0, 0},
};

static PyType_Spec ZlibDecompressor_type_spec = {
    .name = "zlib._ZlibDecompressor",
    .basicsize = sizeof(ZlibDecompressor),
    // Calling PyType_GetModuleState() on a subclass is not safe.
    // ZlibDecompressor_type_spec does not have Py_TPFLAGS_BASETYPE flag
    // which prevents to create a subclass.
    // So calling PyType_GetModuleState() in this file is always safe.
    .flags = (
        Py_TPFLAGS_DEFAULT
        | Py_TPFLAGS_IMMUTABLETYPE
        | Py_TPFLAGS_HAVE_GC
    ),
    .slots = ZlibDecompressor_type_slots,
};

PyDoc_STRVAR(zlib_module_documentation,
"The functions in this module allow compression and decompression using the\n"
"zlib library, which is based on GNU zip.\n"
"\n"
"adler32(string[, start]) -- Compute an Adler-32 checksum.\n"
"adler32_combine(adler1, adler2, len2, /) -- Combine two Adler-32 checksums.\n"
"compress(data[, level]) -- Compress data, with compression level 0-9 or -1.\n"
"compressobj([level[, ...]]) -- Return a compressor object.\n"
"crc32(string[, start]) -- Compute a CRC-32 checksum.\n"
"crc32_combine(crc1, crc2, len2, /) -- Combine two CRC-32 checksums.\n"
"decompress(string,[wbits],[bufsize]) -- Decompresses a compressed string.\n"
"decompressobj([wbits[, zdict]]) -- Return a decompressor object.\n"
"\n"
"'wbits' is window buffer size and container format.\n"
"Compressor objects support compress() and flush() methods; decompressor\n"
"objects support decompress() and flush().");

static int
zlib_clear(PyObject *mod)
{
    zlibstate *state = get_zlib_state(mod);
    Py_CLEAR(state->Comptype);
    Py_CLEAR(state->Decomptype);
    Py_CLEAR(state->ZlibDecompressorType);
    Py_CLEAR(state->ZlibError);
    return 0;
}

static int
zlib_traverse(PyObject *mod, visitproc visit, void *arg)
{
    zlibstate *state = get_zlib_state(mod);
    Py_VISIT(state->Comptype);
    Py_VISIT(state->Decomptype);
    Py_VISIT(state->ZlibDecompressorType);
    Py_VISIT(state->ZlibError);
    return 0;
}

static void
zlib_free(void *mod)
{
    zlib_clear((PyObject *)mod);
}

static int
zlib_exec(PyObject *mod)
{
    zlibstate *state = get_zlib_state(mod);

    state->Comptype = (PyTypeObject *)PyType_FromModuleAndSpec(
        mod, &Comptype_spec, NULL);
    if (state->Comptype == NULL) {
        return -1;
    }

    state->Decomptype = (PyTypeObject *)PyType_FromModuleAndSpec(
        mod, &Decomptype_spec, NULL);
    if (state->Decomptype == NULL) {
        return -1;
    }

    state->ZlibDecompressorType = (PyTypeObject *)PyType_FromModuleAndSpec(
        mod, &ZlibDecompressor_type_spec, NULL);
    if (state->ZlibDecompressorType == NULL) {
        return -1;
    }

    state->ZlibError = PyErr_NewException("zlib.error", NULL, NULL);
    if (PyModule_AddObjectRef(mod, "error", state->ZlibError) < 0) {
        return -1;
    }
    if (PyModule_AddObjectRef(mod, "_ZlibDecompressor",
                              (PyObject *)state->ZlibDecompressorType) < 0) {
        return -1;
    }

#define ZLIB_ADD_INT_MACRO(c)                           \
    do {                                                \
        if ((PyModule_AddIntConstant(mod, #c, c)) < 0) {  \
            return -1;                                  \
        }                                               \
    } while(0)

    ZLIB_ADD_INT_MACRO(MAX_WBITS);
    ZLIB_ADD_INT_MACRO(DEFLATED);
    ZLIB_ADD_INT_MACRO(DEF_MEM_LEVEL);
    ZLIB_ADD_INT_MACRO(DEF_BUF_SIZE);
    // compression levels
    ZLIB_ADD_INT_MACRO(Z_NO_COMPRESSION);
    ZLIB_ADD_INT_MACRO(Z_BEST_SPEED);
    ZLIB_ADD_INT_MACRO(Z_BEST_COMPRESSION);
    ZLIB_ADD_INT_MACRO(Z_DEFAULT_COMPRESSION);
    // compression strategies
    ZLIB_ADD_INT_MACRO(Z_FILTERED);
    ZLIB_ADD_INT_MACRO(Z_HUFFMAN_ONLY);
#ifdef Z_RLE // 1.2.0.1
    ZLIB_ADD_INT_MACRO(Z_RLE);
#endif
#ifdef Z_FIXED // 1.2.2.2
    ZLIB_ADD_INT_MACRO(Z_FIXED);
#endif
    ZLIB_ADD_INT_MACRO(Z_DEFAULT_STRATEGY);
    // allowed flush values
    ZLIB_ADD_INT_MACRO(Z_NO_FLUSH);
    ZLIB_ADD_INT_MACRO(Z_PARTIAL_FLUSH);
    ZLIB_ADD_INT_MACRO(Z_SYNC_FLUSH);
    ZLIB_ADD_INT_MACRO(Z_FULL_FLUSH);
    ZLIB_ADD_INT_MACRO(Z_FINISH);
#ifdef Z_BLOCK // 1.2.0.5 for inflate, 1.2.3.4 for deflate
    ZLIB_ADD_INT_MACRO(Z_BLOCK);
#endif
#ifdef Z_TREES // 1.2.3.4, only for inflate
    ZLIB_ADD_INT_MACRO(Z_TREES);
#endif

    /* zlib_version */
    if (PyModule_Add(mod, "ZLIB_VERSION",
                     PyUnicode_FromString(ZLIB_VERSION)) < 0) {
        return -1;
    }
    if (PyModule_Add(mod, "zlib_version",
                     PyUnicode_FromString(zlibVersion())) < 0) {
        return -1;
    }
    if (PyModule_Add(mod, "ZLIB_RUNTIME_VERSION",
                     PyUnicode_FromString(zlibVersion())) < 0) {
        return -1;
    }
#ifdef ZLIBNG_VERSION
    if (PyModule_Add(mod, "ZLIBNG_VERSION",
                     PyUnicode_FromString(ZLIBNG_VERSION)) < 0) {
        return -1;
    }
#endif
    PyTypeObject *version_type;
    version_type = PyStructSequence_NewType(&zlib_version_info_desc);
    if (version_type == NULL) {
        return -1;
    }
    if (PyModule_Add(mod, "ZLIB_VERSION_INFO",
            make_version_info(version_type, ZLIB_VERSION, 4)) < 0)
    {
        Py_DECREF(version_type);
        return -1;
    }
    if (PyModule_Add(mod, "zlib_version_info",
            make_version_info(version_type, zlibVersion(), 4)) < 0)
    {
        Py_DECREF(version_type);
        return -1;
    }
    Py_DECREF(version_type);
#ifdef ZLIBNG_VERSION
    version_type = PyStructSequence_NewType(&zlibng_version_info_desc);
    if (version_type == NULL) {
        return -1;
    }
    if (PyModule_Add(mod, "ZLIBNG_VERSION_INFO",
            make_version_info(version_type, ZLIBNG_VERSION, 3)) < 0)
    {
        Py_DECREF(version_type);
        return -1;
    }
    Py_DECREF(version_type);
#endif
    return 0;
}

static PyModuleDef_Slot zlib_slots[] = {
    _Py_ABI_SLOT,
    {Py_mod_exec, zlib_exec},
    {Py_mod_multiple_interpreters, Py_MOD_PER_INTERPRETER_GIL_SUPPORTED},
    {Py_mod_gil, Py_MOD_GIL_NOT_USED},
    {0, NULL}
};

static struct PyModuleDef zlibmodule = {
    PyModuleDef_HEAD_INIT,
    .m_name = "zlib",
    .m_doc = zlib_module_documentation,
    .m_size = sizeof(zlibstate),
    .m_methods = zlib_methods,
    .m_slots = zlib_slots,
    .m_traverse = zlib_traverse,
    .m_clear = zlib_clear,
    .m_free = zlib_free,
};

PyMODINIT_FUNC
PyInit_zlib(void)
{
    return PyModuleDef_Init(&zlibmodule);
}