diff options
Diffstat (limited to 'Objects')
| -rw-r--r-- | Objects/codeobject.c | 168 | ||||
| -rw-r--r-- | Objects/dictobject.c | 17 | ||||
| -rw-r--r-- | Objects/funcobject.c | 54 | ||||
| -rw-r--r-- | Objects/longobject.c | 4 | ||||
| -rw-r--r-- | Objects/unicodeobject.c | 33 |
5 files changed, 222 insertions, 54 deletions
diff --git a/Objects/codeobject.c b/Objects/codeobject.c index 92eaf5e00bc..4f06a36a130 100644 --- a/Objects/codeobject.c +++ b/Objects/codeobject.c @@ -1955,12 +1955,130 @@ finally: } +int +_PyCode_CheckNoInternalState(PyCodeObject *co, const char **p_errmsg) +{ + const char *errmsg = NULL; + // We don't worry about co_executors, co_instrumentation, + // or co_monitoring. They are essentially ephemeral. + if (co->co_extra != NULL) { + errmsg = "only basic code objects are supported"; + } + + if (errmsg != NULL) { + if (p_errmsg != NULL) { + *p_errmsg = errmsg; + } + return 0; + } + return 1; +} + +int +_PyCode_CheckNoExternalState(PyCodeObject *co, _PyCode_var_counts_t *counts, + const char **p_errmsg) +{ + const char *errmsg = NULL; + assert(counts->locals.hidden.total == 0); + if (counts->numfree > 0) { // It's a closure. + errmsg = "closures not supported"; + } + else if (counts->unbound.globals.numglobal > 0) { + errmsg = "globals not supported"; + } + else if (counts->unbound.globals.numbuiltin > 0 + && counts->unbound.globals.numunknown > 0) + { + errmsg = "globals not supported"; + } + // Otherwise we don't check counts.unbound.globals.numunknown since we can't + // distinguish beween globals and builtins here. + + if (errmsg != NULL) { + if (p_errmsg != NULL) { + *p_errmsg = errmsg; + } + return 0; + } + return 1; +} + +int +_PyCode_VerifyStateless(PyThreadState *tstate, + PyCodeObject *co, PyObject *globalnames, + PyObject *globalsns, PyObject *builtinsns) +{ + const char *errmsg; + _PyCode_var_counts_t counts = {0}; + _PyCode_GetVarCounts(co, &counts); + if (_PyCode_SetUnboundVarCounts( + tstate, co, &counts, globalnames, NULL, + globalsns, builtinsns) < 0) + { + return -1; + } + // We may consider relaxing the internal state constraints + // if it becomes a problem. + if (!_PyCode_CheckNoInternalState(co, &errmsg)) { + _PyErr_SetString(tstate, PyExc_ValueError, errmsg); + return -1; + } + if (builtinsns != NULL) { + // Make sure the next check will fail for globals, + // even if there aren't any builtins. + counts.unbound.globals.numbuiltin += 1; + } + if (!_PyCode_CheckNoExternalState(co, &counts, &errmsg)) { + _PyErr_SetString(tstate, PyExc_ValueError, errmsg); + return -1; + } + // Note that we don't check co->co_flags & CO_NESTED for anything here. + return 0; +} + + +int +_PyCode_CheckPureFunction(PyCodeObject *co, const char **p_errmsg) +{ + const char *errmsg = NULL; + if (co->co_flags & CO_GENERATOR) { + errmsg = "generators not supported"; + } + else if (co->co_flags & CO_COROUTINE) { + errmsg = "coroutines not supported"; + } + else if (co->co_flags & CO_ITERABLE_COROUTINE) { + errmsg = "coroutines not supported"; + } + else if (co->co_flags & CO_ASYNC_GENERATOR) { + errmsg = "generators not supported"; + } + + if (errmsg != NULL) { + if (p_errmsg != NULL) { + *p_errmsg = errmsg; + } + return 0; + } + return 1; +} + /* Here "value" means a non-None value, since a bare return is identical * to returning None explicitly. Likewise a missing return statement * at the end of the function is turned into "return None". */ static int code_returns_only_none(PyCodeObject *co) { + if (!_PyCode_CheckPureFunction(co, NULL)) { + return 0; + } + int len = (int)Py_SIZE(co); + assert(len > 0); + + // The last instruction either returns or raises. We can take advantage + // of that for a quick exit. + _Py_CODEUNIT final = _Py_GetBaseCodeUnit(co, len-1); + // Look up None in co_consts. Py_ssize_t nconsts = PyTuple_Size(co->co_consts); int none_index = 0; @@ -1971,26 +2089,42 @@ code_returns_only_none(PyCodeObject *co) } if (none_index == nconsts) { // None wasn't there, which means there was no implicit return, - // "return", or "return None". That means there must be - // an explicit return (non-None). - return 0; - } + // "return", or "return None". - // Walk the bytecode, looking for RETURN_VALUE. - Py_ssize_t len = Py_SIZE(co); - for (int i = 0; i < len; i += _PyInstruction_GetLength(co, i)) { - _Py_CODEUNIT inst = _Py_GetBaseCodeUnit(co, i); - if (IS_RETURN_OPCODE(inst.op.code)) { - assert(i != 0); - // Ignore it if it returns None. - _Py_CODEUNIT prev = _Py_GetBaseCodeUnit(co, i-1); - if (prev.op.code == LOAD_CONST) { - // We don't worry about EXTENDED_ARG for now. - if (prev.op.arg == none_index) { - continue; + // That means there must be + // an explicit return (non-None), or it only raises. + if (IS_RETURN_OPCODE(final.op.code)) { + // It was an explicit return (non-None). + return 0; + } + // It must end with a raise then. We still have to walk the + // bytecode to see if there's any explicit return (non-None). + assert(IS_RAISE_OPCODE(final.op.code)); + for (int i = 0; i < len; i += _PyInstruction_GetLength(co, i)) { + _Py_CODEUNIT inst = _Py_GetBaseCodeUnit(co, i); + if (IS_RETURN_OPCODE(inst.op.code)) { + // We alraedy know it isn't returning None. + return 0; + } + } + // It must only raise. + } + else { + // Walk the bytecode, looking for RETURN_VALUE. + for (int i = 0; i < len; i += _PyInstruction_GetLength(co, i)) { + _Py_CODEUNIT inst = _Py_GetBaseCodeUnit(co, i); + if (IS_RETURN_OPCODE(inst.op.code)) { + assert(i != 0); + // Ignore it if it returns None. + _Py_CODEUNIT prev = _Py_GetBaseCodeUnit(co, i-1); + if (prev.op.code == LOAD_CONST) { + // We don't worry about EXTENDED_ARG for now. + if (prev.op.arg == none_index) { + continue; + } } + return 0; } - return 0; } } return 1; diff --git a/Objects/dictobject.c b/Objects/dictobject.c index 59b0cf1ce7d..ce27e47dabf 100644 --- a/Objects/dictobject.c +++ b/Objects/dictobject.c @@ -547,13 +547,13 @@ static inline uint8_t calculate_log2_keysize(Py_ssize_t minsize) { #if SIZEOF_LONG == SIZEOF_SIZE_T - minsize = (minsize | PyDict_MINSIZE) - 1; - return _Py_bit_length(minsize | (PyDict_MINSIZE-1)); + minsize = Py_MAX(minsize, PyDict_MINSIZE); + return _Py_bit_length(minsize - 1); #elif defined(_MSC_VER) - // On 64bit Windows, sizeof(long) == 4. - minsize = (minsize | PyDict_MINSIZE) - 1; + // On 64bit Windows, sizeof(long) == 4. We cannot use _Py_bit_length. + minsize = Py_MAX(minsize, PyDict_MINSIZE); unsigned long msb; - _BitScanReverse64(&msb, (uint64_t)minsize); + _BitScanReverse64(&msb, (uint64_t)minsize - 1); return (uint8_t)(msb + 1); #else uint8_t log2_size; @@ -3178,9 +3178,10 @@ dict_set_fromkeys(PyInterpreterState *interp, PyDictObject *mp, Py_ssize_t pos = 0; PyObject *key; Py_hash_t hash; - - if (dictresize(interp, mp, - estimate_log2_keysize(PySet_GET_SIZE(iterable)), 0)) { + uint8_t new_size = Py_MAX( + estimate_log2_keysize(PySet_GET_SIZE(iterable)), + DK_LOG_SIZE(mp->ma_keys)); + if (dictresize(interp, mp, new_size, 0)) { Py_DECREF(mp); return NULL; } diff --git a/Objects/funcobject.c b/Objects/funcobject.c index 56df5730db0..27214a129c2 100644 --- a/Objects/funcobject.c +++ b/Objects/funcobject.c @@ -1,12 +1,14 @@ /* Function object implementation */ #include "Python.h" +#include "pycore_code.h" // _PyCode_VerifyStateless() #include "pycore_dict.h" // _Py_INCREF_DICT() #include "pycore_function.h" // _PyFunction_Vectorcall #include "pycore_long.h" // _PyLong_GetOne() #include "pycore_modsupport.h" // _PyArg_NoKeywords() #include "pycore_object.h" // _PyObject_GC_UNTRACK() #include "pycore_pyerrors.h" // _PyErr_Occurred() +#include "pycore_setobject.h" // _PySet_NextEntry() #include "pycore_stats.h" @@ -1240,6 +1242,58 @@ PyTypeObject PyFunction_Type = { }; +int +_PyFunction_VerifyStateless(PyThreadState *tstate, PyObject *func) +{ + assert(!PyErr_Occurred()); + assert(PyFunction_Check(func)); + + // Check the globals. + PyObject *globalsns = PyFunction_GET_GLOBALS(func); + if (globalsns != NULL && !PyDict_Check(globalsns)) { + _PyErr_Format(tstate, PyExc_TypeError, + "unsupported globals %R", globalsns); + return -1; + } + // Check the builtins. + PyObject *builtinsns = PyFunction_GET_BUILTINS(func); + if (builtinsns != NULL && !PyDict_Check(builtinsns)) { + _PyErr_Format(tstate, PyExc_TypeError, + "unsupported builtins %R", builtinsns); + return -1; + } + // Disallow __defaults__. + PyObject *defaults = PyFunction_GET_DEFAULTS(func); + if (defaults != NULL && defaults != Py_None && PyDict_Size(defaults) > 0) + { + _PyErr_SetString(tstate, PyExc_ValueError, "defaults not supported"); + return -1; + } + // Disallow __kwdefaults__. + PyObject *kwdefaults = PyFunction_GET_KW_DEFAULTS(func); + if (kwdefaults != NULL && kwdefaults != Py_None + && PyDict_Size(kwdefaults) > 0) + { + _PyErr_SetString(tstate, PyExc_ValueError, + "keyword defaults not supported"); + return -1; + } + // Disallow __closure__. + PyObject *closure = PyFunction_GET_CLOSURE(func); + if (closure != NULL && closure != Py_None && PyTuple_GET_SIZE(closure) > 0) + { + _PyErr_SetString(tstate, PyExc_ValueError, "closures not supported"); + return -1; + } + // Check the code. + PyCodeObject *co = (PyCodeObject *)PyFunction_GET_CODE(func); + if (_PyCode_VerifyStateless(tstate, co, NULL, globalsns, builtinsns) < 0) { + return -1; + } + return 0; +} + + static int functools_copy_attr(PyObject *wrapper, PyObject *wrapped, PyObject *name) { diff --git a/Objects/longobject.c b/Objects/longobject.c index 40d90ecf4fa..0b2dfa003fa 100644 --- a/Objects/longobject.c +++ b/Objects/longobject.c @@ -1760,6 +1760,10 @@ UNSIGNED_INT_CONVERTER(UnsignedInt, unsigned int) UNSIGNED_INT_CONVERTER(UnsignedLong, unsigned long) UNSIGNED_INT_CONVERTER(UnsignedLongLong, unsigned long long) UNSIGNED_INT_CONVERTER(Size_t, size_t) +UNSIGNED_INT_CONVERTER(UInt8, uint8_t) +UNSIGNED_INT_CONVERTER(UInt16, uint16_t) +UNSIGNED_INT_CONVERTER(UInt32, uint32_t) +UNSIGNED_INT_CONVERTER(UInt64, uint64_t) #define CHECK_BINOP(v,w) \ diff --git a/Objects/unicodeobject.c b/Objects/unicodeobject.c index eb3e1c48fd4..f3f0c9646a6 100644 --- a/Objects/unicodeobject.c +++ b/Objects/unicodeobject.c @@ -3730,7 +3730,7 @@ PyUnicode_Decode(const char *s, return NULL; } -PyObject * +PyAPI_FUNC(PyObject *) PyUnicode_AsDecodedObject(PyObject *unicode, const char *encoding, const char *errors) @@ -3740,12 +3740,6 @@ PyUnicode_AsDecodedObject(PyObject *unicode, return NULL; } - if (PyErr_WarnEx(PyExc_DeprecationWarning, - "PyUnicode_AsDecodedObject() is deprecated " - "and will be removed in 3.15; " - "use PyCodec_Decode() to decode from str", 1) < 0) - return NULL; - if (encoding == NULL) encoding = PyUnicode_GetDefaultEncoding(); @@ -3753,7 +3747,7 @@ PyUnicode_AsDecodedObject(PyObject *unicode, return PyCodec_Decode(unicode, encoding, errors); } -PyObject * +PyAPI_FUNC(PyObject *) PyUnicode_AsDecodedUnicode(PyObject *unicode, const char *encoding, const char *errors) @@ -3765,12 +3759,6 @@ PyUnicode_AsDecodedUnicode(PyObject *unicode, goto onError; } - if (PyErr_WarnEx(PyExc_DeprecationWarning, - "PyUnicode_AsDecodedUnicode() is deprecated " - "and will be removed in 3.15; " - "use PyCodec_Decode() to decode from str to str", 1) < 0) - return NULL; - if (encoding == NULL) encoding = PyUnicode_GetDefaultEncoding(); @@ -3793,7 +3781,7 @@ PyUnicode_AsDecodedUnicode(PyObject *unicode, return NULL; } -PyObject * +PyAPI_FUNC(PyObject *) PyUnicode_AsEncodedObject(PyObject *unicode, const char *encoding, const char *errors) @@ -3805,13 +3793,6 @@ PyUnicode_AsEncodedObject(PyObject *unicode, goto onError; } - if (PyErr_WarnEx(PyExc_DeprecationWarning, - "PyUnicode_AsEncodedObject() is deprecated " - "and will be removed in 3.15; " - "use PyUnicode_AsEncodedString() to encode from str to bytes " - "or PyCodec_Encode() for generic encoding", 1) < 0) - return NULL; - if (encoding == NULL) encoding = PyUnicode_GetDefaultEncoding(); @@ -4017,7 +3998,7 @@ PyUnicode_AsEncodedString(PyObject *unicode, return NULL; } -PyObject * +PyAPI_FUNC(PyObject *) PyUnicode_AsEncodedUnicode(PyObject *unicode, const char *encoding, const char *errors) @@ -4029,12 +4010,6 @@ PyUnicode_AsEncodedUnicode(PyObject *unicode, goto onError; } - if (PyErr_WarnEx(PyExc_DeprecationWarning, - "PyUnicode_AsEncodedUnicode() is deprecated " - "and will be removed in 3.15; " - "use PyCodec_Encode() to encode from str to str", 1) < 0) - return NULL; - if (encoding == NULL) encoding = PyUnicode_GetDefaultEncoding(); |