From 2c56ddb5d0025ed481d962c0f5d62d19dec7476d Mon Sep 17 00:00:00 2001 From: Inada Naoki Date: Fri, 19 Jun 2026 00:13:13 +0900 Subject: [PATCH] Merge commit from fork * fix Unpacker crash after unpack failure. * fixup --- msgpack/_unpacker.pyx | 8 ++++++-- msgpack/unpack_template.h | 22 +++++++++++++++------- requirements.txt | 3 ++- test/test_except.py | 26 ++++++++++++++++++++++++++ 4 files changed, 49 insertions(+), 10 deletions(-) diff --git a/msgpack/_unpacker.pyx b/msgpack/_unpacker.pyx index 4bfbe06..e0cdde4 100644 --- a/msgpack/_unpacker.pyx +++ b/msgpack/_unpacker.pyx @@ -51,6 +51,7 @@ cdef extern from "unpack.h": execute_fn unpack_skip execute_fn read_array_header execute_fn read_map_header + void unpack_init(unpack_context* ctx) object unpack_data(unpack_context* ctx) void unpack_clear(unpack_context* ctx) @@ -197,6 +198,7 @@ def unpackb(object packed, *, object object_hook=None, object list_hook=None, if off < buf_len: raise ExtraData(obj, PyBytes_FromStringAndSize(buf+off, buf_len-off)) return obj + unpack_clear(&ctx) if ret == 0: raise ValueError("Unpack failed: incomplete input") @@ -475,7 +477,7 @@ cdef class Unpacker: obj = unpack_data(&self.ctx) unpack_init(&self.ctx) return obj - elif ret == 0: + if ret == 0: if self.file_like is not None: self.read_from_file() continue @@ -483,7 +485,9 @@ cdef class Unpacker: raise StopIteration("No more data to unpack.") else: raise OutOfData("No more data to unpack.") - elif ret == -2: + + unpack_clear(&self.ctx) + if ret == -2: raise FormatError elif ret == -3: raise StackError diff --git a/msgpack/unpack_template.h b/msgpack/unpack_template.h index 4230661..ab5887a 100644 --- a/msgpack/unpack_template.h +++ b/msgpack/unpack_template.h @@ -72,15 +72,14 @@ static inline PyObject* unpack_data(unpack_context* ctx) static inline void unpack_clear(unpack_context *ctx) { - unsigned int i; - for (i = 1; i < ctx->top; i++) { - Py_CLEAR(ctx->stack[i].obj); + for (unsigned int i = 0; i < ctx->top; i++) { /* map_key holds a live reference only while waiting for the value */ if (ctx->stack[i].ct == CT_MAP_VALUE) { Py_CLEAR(ctx->stack[i].map_key); } + Py_CLEAR(ctx->stack[i].obj); } - Py_CLEAR(ctx->stack[0].obj); + unpack_init(ctx); } static inline int unpack_execute(bool construct, unpack_context* ctx, const char* data, Py_ssize_t len, Py_ssize_t* off) @@ -200,7 +199,7 @@ static inline int unpack_execute(bool construct, unpack_context* ctx, const char case 0xd5: // fixext 2 case 0xd6: // fixext 4 case 0xd7: // fixext 8 - again_fixed_trail_if_zero(ACS_EXT_VALUE, + again_fixed_trail_if_zero(ACS_EXT_VALUE, (1 << (((unsigned int)*p) & 0x03))+1, _ext_zero); case 0xd8: // fixext 16 @@ -344,6 +343,7 @@ _push: goto _header_again; case CT_MAP_VALUE: if(construct_cb(_map_item)(user, c->count, &c->obj, c->map_key, obj) < 0) { goto _failed; } + c->map_key = NULL; if(++c->count == c->size) { obj = c->obj; if (construct_cb(_map_end)(user, &obj) < 0) { goto _failed; } @@ -406,10 +406,18 @@ _end: #undef start_container static int unpack_construct(unpack_context *ctx, const char *data, Py_ssize_t len, Py_ssize_t *off) { - return unpack_execute(1, ctx, data, len, off); + int ret = unpack_execute(1, ctx, data, len, off); + if (ret == -1) { + unpack_clear(ctx); + } + return ret; } static int unpack_skip(unpack_context *ctx, const char *data, Py_ssize_t len, Py_ssize_t *off) { - return unpack_execute(0, ctx, data, len, off); + int ret = unpack_execute(0, ctx, data, len, off); + if (ret == -1) { + unpack_clear(ctx); + } + return ret; } #define unpack_container_header read_array_header diff --git a/requirements.txt b/requirements.txt index 9e4643b..7991e3f 100644 --- a/requirements.txt +++ b/requirements.txt @@ -1,3 +1,4 @@ -Cython==3.2.1 +cython==3.2.5 setuptools==78.1.1 +pytest build diff --git a/test/test_except.py b/test/test_except.py index c04b110..a3bf467 100644 --- a/test/test_except.py +++ b/test/test_except.py @@ -33,6 +33,22 @@ def test_raise_from_object_hook(): object_pairs_hook=hook, ) + up = Unpacker(object_hook=hook) + + def up_unpack(x): + up.feed(x) + return up.unpack() + + raises(DummyException, up_unpack, packb({})) + raises(DummyException, up_unpack, packb({"fizz": "buzz"})) + raises(DummyException, up_unpack, packb({"fizz": "buzz"})) + raises(DummyException, up_unpack, packb({"fizz": {"buzz": "spam"}})) + raises( + DummyException, + up_unpack, + packb({"fizz": {"buzz": "spam"}}), + ) + def test_raise_from_list_hook(): def hook(lst: list) -> list: @@ -138,3 +154,13 @@ def test_strict_map_key_with_object_pairs_hook(): packed = packb(valid, use_bin_type=True) result = unpackb(packed, raw=False, strict_map_key=True, object_pairs_hook=list) assert result == [("key", "value")] + + +def test_unpacker_should_not_crash_after_exception(): + up = Unpacker() # default: strict_map_key=True + up.feed(b"\x83\x73\xc4\x00") # fixmap(3): int key (rejected) + empty bin8 + try: + up.unpack() # ValueError: int is not allowed for map key ... + except Exception: + pass + up.skip() # SIGSEGV (resumes from a corrupt parser context)