codeobject.c

#include <stdbool.h>

#include "Python.h"
#include "opcode.h"
#include "structmember.h"         // PyMemberDef
#include "pycore_code.h"          // _PyCodeConstructor
#include "pycore_interp.h"        // PyInterpreterState.co_extra_freefuncs
#include "pycore_opcode.h"        // _PyOpcode_Deopt
#include "pycore_pystate.h"       // _PyInterpreterState_GET()
#include "pycore_tuple.h"         // _PyTuple_ITEMS()
#include "clinic/codeobject.c.h"


/******************
 * generic helpers
 ******************/

/* all_name_chars(s): true iff s matches [a-zA-Z0-9_]* */
static int
all_name_chars(PyObject *o)
{
    const unsigned char *s, *e;

    if (!PyUnicode_IS_ASCII(o))
        return 0;

    s = PyUnicode_1BYTE_DATA(o);
    e = s + PyUnicode_GET_LENGTH(o);
    for (; s != e; s++) {
        if (!Py_ISALNUM(*s) && *s != '_')
            return 0;
    }
    return 1;
}

static int
intern_strings(PyObject *tuple)
{
    Py_ssize_t i;

    for (i = PyTuple_GET_SIZE(tuple); --i >= 0; ) {
        PyObject *v = PyTuple_GET_ITEM(tuple, i);
        if (v == NULL || !PyUnicode_CheckExact(v)) {
            PyErr_SetString(PyExc_SystemError,
                            "non-string found in code slot");
            return -1;
        }
        PyUnicode_InternInPlace(&_PyTuple_ITEMS(tuple)[i]);
    }
    return 0;
}

/* Intern selected string constants */
static int
intern_string_constants(PyObject *tuple, int *modified)
{
    for (Py_ssize_t i = PyTuple_GET_SIZE(tuple); --i >= 0; ) {
        PyObject *v = PyTuple_GET_ITEM(tuple, i);
        if (PyUnicode_CheckExact(v)) {
            if (PyUnicode_READY(v) == -1) {
                return -1;
            }

            if (all_name_chars(v)) {
                PyObject *w = v;
                PyUnicode_InternInPlace(&v);
                if (w != v) {
                    PyTuple_SET_ITEM(tuple, i, v);
                    if (modified) {
                        *modified = 1;
                    }
                }
            }
        }
        else if (PyTuple_CheckExact(v)) {
            if (intern_string_constants(v, NULL) < 0) {
                return -1;
            }
        }
        else if (PyFrozenSet_CheckExact(v)) {
            PyObject *w = v;
            PyObject *tmp = PySequence_Tuple(v);
            if (tmp == NULL) {
                return -1;
            }
            int tmp_modified = 0;
            if (intern_string_constants(tmp, &tmp_modified) < 0) {
                Py_DECREF(tmp);
                return -1;
            }
            if (tmp_modified) {
                v = PyFrozenSet_New(tmp);
                if (v == NULL) {
                    Py_DECREF(tmp);
                    return -1;
                }

                PyTuple_SET_ITEM(tuple, i, v);
                Py_DECREF(w);
                if (modified) {
                    *modified = 1;
                }
            }
            Py_DECREF(tmp);
        }
    }
    return 0;
}

/* Return a shallow copy of a tuple that is
   guaranteed to contain exact strings, by converting string subclasses
   to exact strings and complaining if a non-string is found. */
static PyObject*
validate_and_copy_tuple(PyObject *tup)
{
    PyObject *newtuple;
    PyObject *item;
    Py_ssize_t i, len;

    len = PyTuple_GET_SIZE(tup);
    newtuple = PyTuple_New(len);
    if (newtuple == NULL)
        return NULL;

    for (i = 0; i < len; i++) {
        item = PyTuple_GET_ITEM(tup, i);
        if (PyUnicode_CheckExact(item)) {
            Py_INCREF(item);
        }
        else if (!PyUnicode_Check(item)) {
            PyErr_Format(
                PyExc_TypeError,
                "name tuples must contain only "
                "strings, not '%.500s'",
                Py_TYPE(item)->tp_name);
            Py_DECREF(newtuple);
            return NULL;
        }
        else {
            item = _PyUnicode_Copy(item);
            if (item == NULL) {
                Py_DECREF(newtuple);
                return NULL;
            }
        }
        PyTuple_SET_ITEM(newtuple, i, item);
    }

    return newtuple;
}


/******************
 * _PyCode_New()
 ******************/

// This is also used in compile.c.
void
_Py_set_localsplus_info(int offset, PyObject *name, _PyLocals_Kind kind,
                        PyObject *names, PyObject *kinds)
{
    Py_INCREF(name);
    PyTuple_SET_ITEM(names, offset, name);
    _PyLocals_SetKind(kinds, offset, kind);
}

static void
get_localsplus_counts(PyObject *names, PyObject *kinds,
                      int *pnlocals, int *pnplaincellvars, int *pncellvars,
                      int *pnfreevars)
{
    int nlocals = 0;
    int nplaincellvars = 0;
    int ncellvars = 0;
    int nfreevars = 0;
    Py_ssize_t nlocalsplus = PyTuple_GET_SIZE(names);
    for (int i = 0; i < nlocalsplus; i++) {
        _PyLocals_Kind kind = _PyLocals_GetKind(kinds, i);
        if (kind & CO_FAST_LOCAL) {
            nlocals += 1;
            if (kind & CO_FAST_CELL) {
                ncellvars += 1;
            }
        }
        else if (kind & CO_FAST_CELL) {
            ncellvars += 1;
            nplaincellvars += 1;
        }
        else if (kind & CO_FAST_FREE) {
            nfreevars += 1;
        }
    }
    if (pnlocals != NULL) {
        *pnlocals = nlocals;
    }
    if (pnplaincellvars != NULL) {
        *pnplaincellvars = nplaincellvars;
    }
    if (pncellvars != NULL) {
        *pncellvars = ncellvars;
    }
    if (pnfreevars != NULL) {
        *pnfreevars = nfreevars;
    }
}

static PyObject *
get_localsplus_names(PyCodeObject *co, _PyLocals_Kind kind, int num)
{
    PyObject *names = PyTuple_New(num);
    if (names == NULL) {
        return NULL;
    }
    int index = 0;
    for (int offset = 0; offset < co->co_nlocalsplus; offset++) {
        _PyLocals_Kind k = _PyLocals_GetKind(co->co_localspluskinds, offset);
        if ((k & kind) == 0) {
            continue;
        }
        assert(index < num);
        PyObject *name = PyTuple_GET_ITEM(co->co_localsplusnames, offset);
        Py_INCREF(name);
        PyTuple_SET_ITEM(names, index, name);
        index += 1;
    }
    assert(index == num);
    return names;
}

int
_PyCode_Validate(struct _PyCodeConstructor *con)
{
    /* Check argument types */
    if (con->argcount < con->posonlyargcount || con->posonlyargcount < 0 ||
        con->kwonlyargcount < 0 ||
        con->stacksize < 0 || con->flags < 0 ||
        con->code == NULL || !PyBytes_Check(con->code) ||
        con->consts == NULL || !PyTuple_Check(con->consts) ||
        con->names == NULL || !PyTuple_Check(con->names) ||
        con->localsplusnames == NULL || !PyTuple_Check(con->localsplusnames) ||
        con->localspluskinds == NULL || !PyBytes_Check(con->localspluskinds) ||
        PyTuple_GET_SIZE(con->localsplusnames)
            != PyBytes_GET_SIZE(con->localspluskinds) ||
        con->name == NULL || !PyUnicode_Check(con->name) ||
        con->qualname == NULL || !PyUnicode_Check(con->qualname) ||
        con->filename == NULL || !PyUnicode_Check(con->filename) ||
        con->linetable == NULL || !PyBytes_Check(con->linetable) ||
        con->exceptiontable == NULL || !PyBytes_Check(con->exceptiontable)
        ) {
        PyErr_BadInternalCall();
        return -1;
    }

    /* Make sure that code is indexable with an int, this is
       a long running assumption in ceval.c and many parts of
       the interpreter. */
    if (PyBytes_GET_SIZE(con->code) > INT_MAX) {
        PyErr_SetString(PyExc_OverflowError,
                        "code: co_code larger than INT_MAX");
        return -1;
    }
    if (PyBytes_GET_SIZE(con->code) % sizeof(_Py_CODEUNIT) != 0 ||
        !_Py_IS_ALIGNED(PyBytes_AS_STRING(con->code), sizeof(_Py_CODEUNIT))
        ) {
        PyErr_SetString(PyExc_ValueError, "code: co_code is malformed");
        return -1;
    }

    /* Ensure that the co_varnames has enough names to cover the arg counts.
     * Note that totalargs = nlocals - nplainlocals.  We check nplainlocals
     * here to avoid the possibility of overflow (however remote). */
    int nlocals;
    get_localsplus_counts(con->localsplusnames, con->localspluskinds,
                          &nlocals, NULL, NULL, NULL);
    int nplainlocals = nlocals -
                       con->argcount -
                       con->kwonlyargcount -
                       ((con->flags & CO_VARARGS) != 0) -
                       ((con->flags & CO_VARKEYWORDS) != 0);
    if (nplainlocals < 0) {
        PyErr_SetString(PyExc_ValueError, "code: co_varnames is too small");
        return -1;
    }

    return 0;
}

static void
init_code(PyCodeObject *co, struct _PyCodeConstructor *con)
{
    int nlocalsplus = (int)PyTuple_GET_SIZE(con->localsplusnames);
    int nlocals, nplaincellvars, ncellvars, nfreevars;
    get_localsplus_counts(con->localsplusnames, con->localspluskinds,
                          &nlocals, &nplaincellvars, &ncellvars, &nfreevars);

    Py_INCREF(con->filename);
    co->co_filename = con->filename;
    Py_INCREF(con->name);
    co->co_name = con->name;
    Py_INCREF(con->qualname);
    co->co_qualname = con->qualname;
    co->co_flags = con->flags;

    co->co_firstlineno = con->firstlineno;
    Py_INCREF(con->linetable);
    co->co_linetable = con->linetable;

    Py_INCREF(con->consts);
    co->co_consts = con->consts;
    Py_INCREF(con->names);
    co->co_names = con->names;

    Py_INCREF(con->localsplusnames);
    co->co_localsplusnames = con->localsplusnames;
    Py_INCREF(con->localspluskinds);
    co->co_localspluskinds = con->localspluskinds;

    co->co_argcount = con->argcount;
    co->co_posonlyargcount = con->posonlyargcount;
    co->co_kwonlyargcount = con->kwonlyargcount;

    co->co_stacksize = con->stacksize;

    Py_INCREF(con->exceptiontable);
    co->co_exceptiontable = con->exceptiontable;

    /* derived values */
    co->co_nlocalsplus = nlocalsplus;
    co->co_nlocals = nlocals;
    co->co_nplaincellvars = nplaincellvars;
    co->co_ncellvars = ncellvars;
    co->co_nfreevars = nfreevars;

    /* not set */
    co->co_weakreflist = NULL;
    co->co_extra = NULL;
    co->_co_code = NULL;

    co->co_warmup = QUICKENING_INITIAL_WARMUP_VALUE;
    co->_co_linearray_entry_size = 0;
    co->_co_linearray = NULL;
    memcpy(_PyCode_CODE(co), PyBytes_AS_STRING(con->code),
           PyBytes_GET_SIZE(con->code));
    int entry_point = 0;
    while (entry_point < Py_SIZE(co) &&
        _Py_OPCODE(_PyCode_CODE(co)[entry_point]) != RESUME) {
        entry_point++;
    }
    co->_co_firsttraceable = entry_point;
}

static int
scan_varint(const uint8_t *ptr)
{
    unsigned int read = *ptr++;
    unsigned int val = read & 63;
    unsigned int shift = 0;
    while (read & 64) {
        read = *ptr++;
        shift += 6;
        val |= (read & 63) << shift;
    }
    return val;
}

static int
scan_signed_varint(const uint8_t *ptr)
{
    unsigned int uval = scan_varint(ptr);
    if (uval & 1) {
        return -(int)(uval >> 1);
    }
    else {
        return uval >> 1;
    }
}

static int
get_line_delta(const uint8_t *ptr)
{
    int code = ((*ptr) >> 3) & 15;
    switch (code) {
        case PY_CODE_LOCATION_INFO_NONE:
            return 0;
        case PY_CODE_LOCATION_INFO_NO_COLUMNS:
        case PY_CODE_LOCATION_INFO_LONG:
            return scan_signed_varint(ptr+1);
        case PY_CODE_LOCATION_INFO_ONE_LINE0:
            return 0;
        case PY_CODE_LOCATION_INFO_ONE_LINE1:
            return 1;
        case PY_CODE_LOCATION_INFO_ONE_LINE2:
            return 2;
        default:
            /* Same line */
            return 0;
    }
}

static PyObject *
remove_column_info(PyObject *locations)
{
    int offset = 0;
    const uint8_t *data = (const uint8_t *)PyBytes_AS_STRING(locations);
    PyObject *res = PyBytes_FromStringAndSize(NULL, 32);
    if (res == NULL) {
        PyErr_NoMemory();
        return NULL;
    }
    uint8_t *output = (uint8_t *)PyBytes_AS_STRING(res);
    while (offset < PyBytes_GET_SIZE(locations)) {
        Py_ssize_t write_offset = output - (uint8_t *)PyBytes_AS_STRING(res);
        if (write_offset + 16 >= PyBytes_GET_SIZE(res)) {
            if (_PyBytes_Resize(&res, PyBytes_GET_SIZE(res) * 2) < 0) {
                return NULL;
            }
            output = (uint8_t *)PyBytes_AS_STRING(res) + write_offset;
        }
        int code = (data[offset] >> 3) & 15;
        if (code == PY_CODE_LOCATION_INFO_NONE) {
            *output++ = data[offset];
        }
        else {
            int blength = (data[offset] & 7)+1;
            output += write_location_entry_start(
                output, PY_CODE_LOCATION_INFO_NO_COLUMNS, blength);
            int ldelta = get_line_delta(&data[offset]);
            output += write_signed_varint(output, ldelta);
        }
        offset++;
        while (offset < PyBytes_GET_SIZE(locations) &&
            (data[offset] & 128) == 0) {
            offset++;
        }
    }
    Py_ssize_t write_offset = output - (uint8_t *)PyBytes_AS_STRING(res);
    if (_PyBytes_Resize(&res, write_offset)) {
        return NULL;
    }
    return res;
}

/* The caller is responsible for ensuring that the given data is valid. */

PyCodeObject *
_PyCode_New(struct _PyCodeConstructor *con)
{
    /* Ensure that strings are ready Unicode string */
    if (PyUnicode_READY(con->name) < 0) {
        return NULL;
    }
    if (PyUnicode_READY(con->qualname) < 0) {
        return NULL;
    }
    if (PyUnicode_READY(con->filename) < 0) {
        return NULL;
    }

    if (intern_strings(con->names) < 0) {
        return NULL;
    }
    if (intern_string_constants(con->consts, NULL) < 0) {
        return NULL;
    }
    if (intern_strings(con->localsplusnames) < 0) {
        return NULL;
    }

    PyObject *replacement_locations = NULL;
    // Compact the linetable if we are opted out of debug
    // ranges.
    if (!_Py_GetConfig()->code_debug_ranges) {
        replacement_locations = remove_column_info(con->linetable);
        if (replacement_locations == NULL) {
            return NULL;
        }
        con->linetable = replacement_locations;
    }

    Py_ssize_t size = PyBytes_GET_SIZE(con->code) / sizeof(_Py_CODEUNIT);
    PyCodeObject *co = PyObject_NewVar(PyCodeObject, &PyCode_Type, size);
    if (co == NULL) {
        Py_XDECREF(replacement_locations);
        PyErr_NoMemory();
        return NULL;
    }
    init_code(co, con);
    Py_XDECREF(replacement_locations);
    return co;
}


/******************
 * the legacy "constructors"
 ******************/

PyCodeObject *
PyCode_NewWithPosOnlyArgs(int argcount, int posonlyargcount, int kwonlyargcount,
                          int nlocals, int stacksize, int flags,
                          PyObject *code, PyObject *consts, PyObject *names,
                          PyObject *varnames, PyObject *freevars, PyObject *cellvars,
                          PyObject *filename, PyObject *name,
                          PyObject *qualname, int firstlineno,
                          PyObject *linetable,
                          PyObject *exceptiontable)
{
    PyCodeObject *co = NULL;
    PyObject *localsplusnames = NULL;
    PyObject *localspluskinds = NULL;

    if (varnames == NULL || !PyTuple_Check(varnames) ||
        cellvars == NULL || !PyTuple_Check(cellvars) ||
        freevars == NULL || !PyTuple_Check(freevars)
        ) {
        PyErr_BadInternalCall();
        return NULL;
    }

    // Set the "fast locals plus" info.
    int nvarnames = (int)PyTuple_GET_SIZE(varnames);
    int ncellvars = (int)PyTuple_GET_SIZE(cellvars);
    int nfreevars = (int)PyTuple_GET_SIZE(freevars);
    int nlocalsplus = nvarnames + ncellvars + nfreevars;
    localsplusnames = PyTuple_New(nlocalsplus);
    if (localsplusnames == NULL) {
        goto error;
    }
    localspluskinds = PyBytes_FromStringAndSize(NULL, nlocalsplus);
    if (localspluskinds == NULL) {
        goto error;
    }
    int  offset = 0;
    for (int i = 0; i < nvarnames; i++, offset++) {
        PyObject *name = PyTuple_GET_ITEM(varnames, i);
        _Py_set_localsplus_info(offset, name, CO_FAST_LOCAL,
                               localsplusnames, localspluskinds);
    }
    for (int i = 0; i < ncellvars; i++, offset++) {
        PyObject *name = PyTuple_GET_ITEM(cellvars, i);
        int argoffset = -1;
        for (int j = 0; j < nvarnames; j++) {
            int cmp = PyUnicode_Compare(PyTuple_GET_ITEM(varnames, j),
                                        name);
            assert(!PyErr_Occurred());
            if (cmp == 0) {
                argoffset = j;
                break;
            }
        }
        if (argoffset >= 0) {
            // Merge the localsplus indices.
            nlocalsplus -= 1;
            offset -= 1;
            _PyLocals_Kind kind = _PyLocals_GetKind(localspluskinds, argoffset);
            _PyLocals_SetKind(localspluskinds, argoffset, kind | CO_FAST_CELL);
            continue;
        }
        _Py_set_localsplus_info(offset, name, CO_FAST_CELL,
                               localsplusnames, localspluskinds);
    }
    for (int i = 0; i < nfreevars; i++, offset++) {
        PyObject *name = PyTuple_GET_ITEM(freevars, i);
        _Py_set_localsplus_info(offset, name, CO_FAST_FREE,
                               localsplusnames, localspluskinds);
    }
    // If any cells were args then nlocalsplus will have shrunk.
    if (nlocalsplus != PyTuple_GET_SIZE(localsplusnames)) {
        if (_PyTuple_Resize(&localsplusnames, nlocalsplus) < 0
                || _PyBytes_Resize(&localspluskinds, nlocalsplus) < 0) {
            goto error;
        }
    }

    struct _PyCodeConstructor con = {
        .filename = filename,
        .name = name,
        .qualname = qualname,
        .flags = flags,

        .code = code,
        .firstlineno = firstlineno,
        .linetable = linetable,

        .consts = consts,
        .names = names,

        .localsplusnames = localsplusnames,
        .localspluskinds = localspluskinds,

        .argcount = argcount,
        .posonlyargcount = posonlyargcount,
        .kwonlyargcount = kwonlyargcount,

        .stacksize = stacksize,

        .exceptiontable = exceptiontable,
    };

    if (_PyCode_Validate(&con) < 0) {
        goto error;
    }
    assert(PyBytes_GET_SIZE(code) % sizeof(_Py_CODEUNIT) == 0);
    assert(_Py_IS_ALIGNED(PyBytes_AS_STRING(code), sizeof(_Py_CODEUNIT)));
    if (nlocals != PyTuple_GET_SIZE(varnames)) {
        PyErr_SetString(PyExc_ValueError,
                        "code: co_nlocals != len(co_varnames)");
        goto error;
    }

    co = _PyCode_New(&con);
    if (co == NULL) {
        goto error;
    }

error:
    Py_XDECREF(localsplusnames);
    Py_XDECREF(localspluskinds);
    return co;
}

PyCodeObject *
PyCode_New(int argcount, int kwonlyargcount,
           int nlocals, int stacksize, int flags,
           PyObject *code, PyObject *consts, PyObject *names,
           PyObject *varnames, PyObject *freevars, PyObject *cellvars,
           PyObject *filename, PyObject *name, PyObject *qualname,
           int firstlineno,
           PyObject *linetable,
           PyObject *exceptiontable)
{
    return PyCode_NewWithPosOnlyArgs(argcount, 0, kwonlyargcount, nlocals,
                                     stacksize, flags, code, consts, names,
                                     varnames, freevars, cellvars, filename,
                                     name, qualname, firstlineno,
                                     linetable,
                                     exceptiontable);
}

static const char assert0[6] = {
    RESUME, 0,
    LOAD_ASSERTION_ERROR, 0,
    RAISE_VARARGS, 1
};

PyCodeObject *
PyCode_NewEmpty(const char *filename, const char *funcname, int firstlineno)
{
    PyObject *nulltuple = NULL;
    PyObject *filename_ob = NULL;
    PyObject *funcname_ob = NULL;
    PyObject *code_ob = NULL;
    PyCodeObject *result = NULL;

    nulltuple = PyTuple_New(0);
    if (nulltuple == NULL) {
        goto failed;
    }
    funcname_ob = PyUnicode_FromString(funcname);
    if (funcname_ob == NULL) {
        goto failed;
    }
    filename_ob = PyUnicode_DecodeFSDefault(filename);
    if (filename_ob == NULL) {
        goto failed;
    }
    code_ob = PyBytes_FromStringAndSize(assert0, 6);
    if (code_ob == NULL) {
        goto failed;
    }

#define emptystring (PyObject *)&_Py_SINGLETON(bytes_empty)
    struct _PyCodeConstructor con = {
        .filename = filename_ob,
        .name = funcname_ob,
        .qualname = funcname_ob,
        .code = code_ob,
        .firstlineno = firstlineno,
        .linetable = emptystring,
        .consts = nulltuple,
        .names = nulltuple,
        .localsplusnames = nulltuple,
        .localspluskinds = emptystring,
        .exceptiontable = emptystring,
        .stacksize = 1,
    };
    result = _PyCode_New(&con);

failed:
    Py_XDECREF(nulltuple);
    Py_XDECREF(funcname_ob);
    Py_XDECREF(filename_ob);
    Py_XDECREF(code_ob);
    return result;
}


/******************
 * source location tracking (co_lines/co_positions)
 ******************/

/* Use co_linetable to compute the line number from a bytecode index, addrq.  See
   lnotab_notes.txt for the details of the lnotab representation.
*/

int
_PyCode_CreateLineArray(PyCodeObject *co)
{
    assert(co->_co_linearray == NULL);
    PyCodeAddressRange bounds;
    int size;
    int max_line = 0;
    _PyCode_InitAddressRange(co, &bounds);
    while(_PyLineTable_NextAddressRange(&bounds)) {
        if (bounds.ar_line > max_line) {
            max_line = bounds.ar_line;
        }
    }
    if (max_line < (1 << 15)) {
        size = 2;
    }
    else {
        size = 4;
    }
    co->_co_linearray = PyMem_Malloc(Py_SIZE(co)*size);
    if (co->_co_linearray == NULL) {
        PyErr_NoMemory();
        return -1;
    }
    co->_co_linearray_entry_size = size;
    _PyCode_InitAddressRange(co, &bounds);
    while(_PyLineTable_NextAddressRange(&bounds)) {
        int start = bounds.ar_start / sizeof(_Py_CODEUNIT);
        int end = bounds.ar_end / sizeof(_Py_CODEUNIT);
        for (int index = start; index < end; index++) {
            assert(index < (int)Py_SIZE(co));
            if (size == 2) {
                assert(((int16_t)bounds.ar_line) == bounds.ar_line);
                ((int16_t *)co->_co_linearray)[index] = bounds.ar_line;
            }
            else {
                assert(size == 4);
                ((int32_t *)co->_co_linearray)[index] = bounds.ar_line;
            }
        }
    }
    return 0;
}

int
PyCode_Addr2Line(PyCodeObject *co, int addrq)
{
    if (addrq < 0) {
        return co->co_firstlineno;
    }
    assert(addrq >= 0 && addrq < _PyCode_NBYTES(co));
    if (co->_co_linearray) {
        return _PyCode_LineNumberFromArray(co, addrq / sizeof(_Py_CODEUNIT));
    }
    PyCodeAddressRange bounds;
    _PyCode_InitAddressRange(co, &bounds);
    return _PyCode_CheckLineNumber(addrq, &bounds);
}

void
_PyLineTable_InitAddressRange(const char *linetable, Py_ssize_t length, int firstlineno, PyCodeAddressRange *range)
{
    range->opaque.lo_next = (const uint8_t *)linetable;
    range->opaque.limit = range->opaque.lo_next + length;
    range->ar_start = -1;
    range->ar_end = 0;
    range->opaque.computed_line = firstlineno;
    range->ar_line = -1;
}

int
_PyCode_InitAddressRange(PyCodeObject* co, PyCodeAddressRange *bounds)
{
    assert(co->co_linetable != NULL);
    const char *linetable = PyBytes_AS_STRING(co->co_linetable);
    Py_ssize_t length = PyBytes_GET_SIZE(co->co_linetable);
    _PyLineTable_InitAddressRange(linetable, length, co->co_firstlineno, bounds);
    return bounds->ar_line;
}

/* Update *bounds to describe the first and one-past-the-last instructions in
   the same line as lasti.  Return the number of that line, or -1 if lasti is out of bounds. */
int
_PyCode_CheckLineNumber(int lasti, PyCodeAddressRange *bounds)
{
    while (bounds->ar_end <= lasti) {
        if (!_PyLineTable_NextAddressRange(bounds)) {
            return -1;
        }
    }
    while (bounds->ar_start > lasti) {
        if (!_PyLineTable_PreviousAddressRange(bounds)) {
            return -1;
        }
    }
    return bounds->ar_line;
}

static int
is_no_line_marker(uint8_t b)
{
    return (b >> 3) == 0x1f;
}


#define ASSERT_VALID_BOUNDS(bounds) \
    assert(bounds->opaque.lo_next <=  bounds->opaque.limit && \
        (bounds->ar_line == -1 || bounds->ar_line == bounds->opaque.computed_line) && \
        (bounds->opaque.lo_next == bounds->opaque.limit || \
        (*bounds->opaque.lo_next) & 128))

static int
next_code_delta(PyCodeAddressRange *bounds)
{
    assert((*bounds->opaque.lo_next) & 128);
    return (((*bounds->opaque.lo_next) & 7) + 1) * sizeof(_Py_CODEUNIT);
}

static int
previous_code_delta(PyCodeAddressRange *bounds)
{
    if (bounds->ar_start == 0) {
        // If we looking at the first entry, the
        // "previous" entry has an implicit length of 1.
        return 1;
    }
    const uint8_t *ptr = bounds->opaque.lo_next-1;
    while (((*ptr) & 128) == 0) {
        ptr--;
    }
    return (((*ptr) & 7) + 1) * sizeof(_Py_CODEUNIT);
}

static int
read_byte(PyCodeAddressRange *bounds)
{
    return *bounds->opaque.lo_next++;
}

static int
read_varint(PyCodeAddressRange *bounds)
{
    unsigned int read = read_byte(bounds);
    unsigned int val = read & 63;
    unsigned int shift = 0;
    while (read & 64) {
        read = read_byte(bounds);
        shift += 6;
        val |= (read & 63) << shift;
    }
    return val;
}

static int
read_signed_varint(PyCodeAddressRange *bounds)
{
    unsigned int uval = read_varint(bounds);
    if (uval & 1) {
        return -(int)(uval >> 1);
    }
    else {
        return uval >> 1;
    }
}

static void
retreat(PyCodeAddressRange *bounds)
{
    ASSERT_VALID_BOUNDS(bounds);
    assert(bounds->ar_start >= 0);
    do {
        bounds->opaque.lo_next--;
    } while (((*bounds->opaque.lo_next) & 128) == 0);
    bounds->opaque.computed_line -= get_line_delta(bounds->opaque.lo_next);
    bounds->ar_end = bounds->ar_start;
    bounds->ar_start -= previous_code_delta(bounds);
    if (is_no_line_marker(bounds->opaque.lo_next[-1])) {
        bounds->ar_line = -1;
    }
    else {
        bounds->ar_line = bounds->opaque.computed_line;
    }
    ASSERT_VALID_BOUNDS(bounds);
}

static void
advance(PyCodeAddressRange *bounds)
{
    ASSERT_VALID_BOUNDS(bounds);
    bounds->opaque.computed_line += get_line_delta(bounds->opaque.lo_next);
    if (is_no_line_marker(*bounds->opaque.lo_next)) {
        bounds->ar_line = -1;
    }
    else {
        bounds->ar_line = bounds->opaque.computed_line;
    }
    bounds->ar_start = bounds->ar_end;
    bounds->ar_end += next_code_delta(bounds);
    do {
        bounds->opaque.lo_next++;
    } while (bounds->opaque.lo_next < bounds->opaque.limit &&
        ((*bounds->opaque.lo_next) & 128) == 0);
    ASSERT_VALID_BOUNDS(bounds);
}

static void
advance_with_locations(PyCodeAddressRange *bounds, int *endline, int *column, int *endcolumn)
{
    ASSERT_VALID_BOUNDS(bounds);
    int first_byte = read_byte(bounds);
    int code = (first_byte >> 3) & 15;
    bounds->ar_start = bounds->ar_end;
    bounds->ar_end = bounds->ar_start + ((first_byte & 7) + 1) * sizeof(_Py_CODEUNIT);
    switch(code) {
        case PY_CODE_LOCATION_INFO_NONE:
            bounds->ar_line = *endline = -1;
            *column =  *endcolumn = -1;
            break;
        case PY_CODE_LOCATION_INFO_LONG:
        {
            bounds->opaque.computed_line += read_signed_varint(bounds);
            bounds->ar_line = bounds->opaque.computed_line;
            *endline = bounds->ar_line + read_varint(bounds);
            *column = read_varint(bounds)-1;
            *endcolumn = read_varint(bounds)-1;
            break;
        }
        case PY_CODE_LOCATION_INFO_NO_COLUMNS:
        {
            /* No column */
            bounds->opaque.computed_line += read_signed_varint(bounds);
            *endline = bounds->ar_line = bounds->opaque.computed_line;
            *column = *endcolumn = -1;
            break;
        }
        case PY_CODE_LOCATION_INFO_ONE_LINE0:
        case PY_CODE_LOCATION_INFO_ONE_LINE1:
        case PY_CODE_LOCATION_INFO_ONE_LINE2:
        {
            /* one line form */
            int line_delta = code - 10;
            bounds->opaque.computed_line += line_delta;
            *endline = bounds->ar_line = bounds->opaque.computed_line;
            *column = read_byte(bounds);
            *endcolumn = read_byte(bounds);
            break;
        }
        default:
        {
            /* Short forms */
            int second_byte = read_byte(bounds);
            assert((second_byte & 128) == 0);
            *endline = bounds->ar_line = bounds->opaque.computed_line;
            *column = code << 3 | (second_byte >> 4);
            *endcolumn = *column + (second_byte & 15);
        }
    }
    ASSERT_VALID_BOUNDS(bounds);
}
int
PyCode_Addr2Location(PyCodeObject *co, int addrq,
                     int *start_line, int *start_column,
                     int *end_line, int *end_column)
{
    if (addrq < 0) {
        *start_line = *end_line = co->co_firstlineno;
        *start_column = *end_column = 0;
        return 1;
    }
    assert(addrq >= 0 && addrq < _PyCode_NBYTES(co));
    PyCodeAddressRange bounds;
    _PyCode_InitAddressRange(co, &bounds);
    _PyCode_CheckLineNumber(addrq, &bounds);
    retreat(&bounds);
    advance_with_locations(&bounds, end_line, start_column, end_column);
    *start_line = bounds.ar_line;
    return 1;
}


static inline int
at_end(PyCodeAddressRange *bounds) {
    return bounds->opaque.lo_next >= bounds->opaque.limit;
}

int
_PyLineTable_PreviousAddressRange(PyCodeAddressRange *range)
{
    if (range->ar_start <= 0) {
        return 0;
    }
    retreat(range);
    assert(range->ar_end > range->ar_start);
    return 1;
}