fix(mpack): boundary values for negative integer encoding #37255

Problem:
libmpack encodes boundary values -129 and -32769 with wrong integer
sizes:
- -129 as int8 instead of int16
- -32769 as int16 instead of int32
because the boundary checks compare against the wrong values (e.g., lo
< 0xffffff7f instead of lo < 0xffffff80). This caused data corruption:
-129 would decode as 127.

Solution:
Fix off-by-one errors in the two's complement boundary constants:
0xffffff80 (-128, min int8) and 0xffff8000 (-32768, min int16).

Fixes #37202

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
benarcher2691
2026-03-21 22:07:33 +01:00
committed by GitHub
parent 4d3a67cd62
commit 2069be281c
4 changed files with 59 additions and 5 deletions

View File

@@ -122,8 +122,8 @@ MPACK_API mpack_token_t mpack_pack_number(double v)
if (!tok.data.value.lo) tok.data.value.hi++;
if (tok.data.value.lo == 0 && tok.data.value.hi == 0) tok.length = 1;
else if (tok.data.value.lo < 0x80000000) tok.length = 8;
else if (tok.data.value.lo < 0xffff7fff) tok.length = 4;
else if (tok.data.value.lo < 0xffffff7f) tok.length = 2;
else if (tok.data.value.lo < 0xffff8000) tok.length = 4;
else if (tok.data.value.lo < 0xffffff80) tok.length = 2;
else tok.length = 1;
} else {
tok.type = MPACK_TOKEN_UINT;

View File

@@ -408,11 +408,11 @@ static int mpack_wnint(char **buf, size_t *buflen, mpack_value_t val)
return mpack_w1(buf, buflen, 0xd3) ||
mpack_w4(buf, buflen, hi) ||
mpack_w4(buf, buflen, lo);
} else if (lo < 0xffff7fff) {
} else if (lo < 0xffff8000) {
/* int 32 */
return mpack_w1(buf, buflen, 0xd2) ||
mpack_w4(buf, buflen, lo);
} else if (lo < 0xffffff7f) {
} else if (lo < 0xffffff80) {
/* int 16 */
return mpack_w1(buf, buflen, 0xd1) ||
mpack_w2(buf, buflen, lo);

View File

@@ -28,6 +28,31 @@ describe('lua vim.mpack', function()
)
end)
it('encodes negative integers at type boundaries correctly #37202', function()
-- Test boundary values between int8/int16/int32
-- int8 range: -128 to -33 (fixint handles -32 to -1)
-- int16 range: -32768 to -129
-- int32 range: -2147483648 to -32769
local result = exec_lua(function()
local tests = {
{ -128, -128 }, -- int8 boundary (minimum int8)
{ -129, -129 }, -- int16 boundary (one past int8)
{ -32768, -32768 }, -- int16 boundary (minimum int16)
{ -32769, -32769 }, -- int32 boundary (one past int16)
}
local results = {}
for _, test in ipairs(tests) do
local input, expected = test[1], test[2]
local decoded = vim.mpack.decode(vim.mpack.encode(input))
table.insert(results, { input = input, decoded = decoded, ok = decoded == expected })
end
return results
end)
for _, r in ipairs(result) do
eq(true, r.ok, string.format('encode/decode %d returned %d', r.input, r.decoded))
end
end)
it('encodes dict keys of length 20-31 as fixstr #32784', function()
-- MessagePack fixstr format: 0xa0 | length (for lengths 0-31)
-- Before #36737, strings 20-31 bytes were incorrectly encoded as str8 (0xd9, len)

View File

@@ -1,7 +1,8 @@
local t = require('test.unit.testutil')
local cimport = t.cimport
local itp = t.gen_itp(it)
local lib = cimport('./src/nvim/msgpack_rpc/unpacker.h', './src/nvim/memory.h')
local lib =
cimport('./src/nvim/msgpack_rpc/unpacker.h', './src/nvim/memory.h', './src/mpack/conv.h')
local ffi = t.ffi
local eq = t.eq
local to_cstr = t.to_cstr
@@ -77,3 +78,31 @@ describe('msgpack', function()
end)
end)
end)
describe('libmpack', function()
describe('mpack_pack_number', function()
itp('token length at negative integer boundaries #37202', function()
-- tok.length is the byte count for the integer value (not including the msgpack type prefix byte).
-- int8 min: -128 → 1 byte
local tok = lib.mpack_pack_number(-128)
eq(lib.MPACK_TOKEN_SINT, tok.type)
eq(1, tok.length)
-- one past int8: -129 → 2 bytes (int16)
tok = lib.mpack_pack_number(-129)
eq(lib.MPACK_TOKEN_SINT, tok.type)
eq(2, tok.length)
-- int16 min: -32768 → 2 bytes
tok = lib.mpack_pack_number(-32768)
eq(lib.MPACK_TOKEN_SINT, tok.type)
eq(2, tok.length)
-- one past int16: -32769 → 4 bytes (int32)
tok = lib.mpack_pack_number(-32769)
eq(lib.MPACK_TOKEN_SINT, tok.type)
eq(4, tok.length)
end)
end)
end)