[varint] Add signed LEB128 encoding.

This commit is contained in:
Jeroen van Rijn
2022-03-08 18:07:16 +01:00
parent 6d7217f37a
commit e76a5d8e12
2 changed files with 119 additions and 24 deletions

View File

@@ -6,21 +6,31 @@
Jeroen van Rijn: Initial implementation.
*/
// package varint implements variable length integer encoding and decoding
// using the LEB128 format as used by DWARF debug and other file formats
// package varint implements variable length integer encoding and decoding using
// the LEB128 format as used by DWARF debug info, Android .dex and other file formats.
package varint
// Decode a slice of bytes encoding an unsigned LEB128 integer into value and number of bytes used.
// Returns `size` == 0 for an invalid value, empty slice, or a varint > 16 bytes.
// In theory we should use the bigint package. In practice, varints bigger than this indicate a corrupted file.
decode_uleb128 :: proc(buf: []u8) -> (val: u128, size: int) {
// Instead we'll set limits on the values we'll encode/decode
// 18 * 7 bits = 126, which means that a possible 19th byte may at most be `0b0000_0011`.
LEB128_MAX_BYTES :: 19
Error :: enum {
None = 0,
Buffer_Too_Small = 1,
Value_Too_Large = 2,
}
// Decode a slice of bytes encoding an unsigned LEB128 integer into value and number of bytes used.
// Returns `size` == 0 for an invalid value, empty slice, or a varint > 18 bytes.
decode_uleb128 :: proc(buf: []u8) -> (val: u128, size: int, err: Error) {
more := true
for v, i in buf {
size = i + 1
if size > size_of(u128) {
return
if size == LEB128_MAX_BYTES && v > 0b0000_0011 {
return 0, 0, .Value_Too_Large
}
val |= u128(v & 0x7f) << uint(i * 7)
@@ -33,25 +43,26 @@ decode_uleb128 :: proc(buf: []u8) -> (val: u128, size: int) {
// If the buffer runs out before the number ends, return an error.
if more {
return 0, 0
return 0, 0, .Buffer_Too_Small
}
return
}
// Decode a slice of bytes encoding a signed LEB128 integer into value and number of bytes used.
// Returns `size` == 0 for an invalid value, empty slice, or a varint > 16 bytes.
// In theory we should use the bigint package. In practice, varints bigger than this indicate a corrupted file.
decode_ileb128 :: proc(buf: []u8) -> (val: i128, size: int) {
// Returns `size` == 0 for an invalid value, empty slice, or a varint > 18 bytes.
decode_ileb128 :: proc(buf: []u8) -> (val: i128, size: int, err: Error) {
shift: uint
if len(buf) == 0 {
return
return 0, 0, .Buffer_Too_Small
}
for v in buf {
size += 1
if size > size_of(i128) {
return
// 18 * 7 bits = 126, which means that a possible 19th byte may at most be 0b0000_0011.
if size == LEB128_MAX_BYTES && v > 0b0000_0011 {
return 0, 0, .Value_Too_Large
}
val |= i128(v & 0x7f) << shift
@@ -64,4 +75,61 @@ decode_ileb128 :: proc(buf: []u8) -> (val: i128, size: int) {
val |= max(i128) << shift
}
return
}
// Encode `val` into `buf` as an unsigned LEB128 encoded series of bytes.
// `buf` must be appropriately sized.
encode_uleb128 :: proc(buf: []u8, val: u128) -> (size: int, err: Error) {
val := val
for {
size += 1
if size > len(buf) {
return 0, .Buffer_Too_Small
}
low := val & 0x7f
val >>= 7
if val > 0 {
low |= 0x80 // more bytes to follow
}
buf[size - 1] = u8(low)
if val == 0 { break }
}
return
}
@(private)
SIGN_MASK :: (i128(1) << 121) // sign extend mask
// Encode `val` into `buf` as a signed LEB128 encoded series of bytes.
// `buf` must be appropriately sized.
encode_ileb128 :: proc(buf: []u8, val: i128) -> (size: int, err: Error) {
val := val
more := true
for more {
size += 1
if size > len(buf) {
return 0, .Buffer_Too_Small
}
low := val & 0x7f
val >>= 7
low = (low ~ SIGN_MASK) - SIGN_MASK
if (val == 0 && low & 0x40 != 0x40) || (val == -1 && low & 0x40 == 0x40) {
more = false
} else {
low |= 0x80
}
buf[size - 1] = u8(low)
}
return
}

View File

@@ -4,6 +4,7 @@ import "core:encoding/varint"
import "core:testing"
import "core:fmt"
import "core:os"
import "core:slice"
TEST_count := 0
TEST_fail := 0
@@ -39,18 +40,40 @@ main :: proc() {
@(test)
test_dwarf :: proc(t: ^testing.T) {
buf: [varint.LEB128_MAX_BYTES]u8
for vector in ULEB_Vectors {
val, size := varint.decode_uleb128(vector.encoded)
val, size, err := varint.decode_uleb128(vector.encoded)
msg := fmt.tprintf("Expected %02x to decode to %v consuming %v bytes, got %v and %v", vector.encoded, vector.value, vector.size, val, size)
expect(t, size == vector.size && val == vector.value, msg)
msg = fmt.tprintf("Expected decoder to return error %v, got %v", vector.error, err)
expect(t, err == vector.error, msg)
if err == .None { // Try to roundtrip
size, err = varint.encode_uleb128(buf[:], vector.value)
msg = fmt.tprintf("Expected %v to encode to %02x, got %02x", vector.value, vector.encoded, buf[:size])
expect(t, size == vector.size && slice.simple_equal(vector.encoded, buf[:size]), msg)
}
}
for vector in ILEB_Vectors {
val, size := varint.decode_ileb128(vector.encoded)
val, size, err := varint.decode_ileb128(vector.encoded)
msg := fmt.tprintf("Expected %02x to decode to %v consuming %v bytes, got %v and %v", vector.encoded, vector.value, vector.size, val, size)
expect(t, size == vector.size && val == vector.value, msg)
msg = fmt.tprintf("Expected decoder to return error %v, got %v", vector.error, err)
expect(t, err == vector.error, msg)
if err == .None { // Try to roundtrip
size, err = varint.encode_ileb128(buf[:], vector.value)
msg = fmt.tprintf("Expected %v to encode to %02x, got %02x", vector.value, vector.encoded, buf[:size])
expect(t, size == vector.size && slice.simple_equal(vector.encoded, buf[:size]), msg)
}
}
}
@@ -58,24 +81,28 @@ ULEB_Test_Vector :: struct {
encoded: []u8,
value: u128,
size: int,
error: varint.Error,
}
ULEB_Vectors :: []ULEB_Test_Vector{
{ []u8{0x00}, 0, 1 },
{ []u8{0x7f}, 127, 1 },
{ []u8{0xE5, 0x8E, 0x26}, 624485, 3 },
{ []u8{0x80}, 0, 0 },
{ []u8{}, 0, 0 },
{ []u8{0x00}, 0, 1, .None },
{ []u8{0x7f}, 127, 1, .None },
{ []u8{0xE5, 0x8E, 0x26}, 624485, 3, .None },
{ []u8{0x80}, 0, 0, .Buffer_Too_Small },
{ []u8{}, 0, 0, .Buffer_Too_Small },
}
ILEB_Test_Vector :: struct {
encoded: []u8,
value: i128,
size: int,
error: varint.Error,
}
ILEB_Vectors :: []ILEB_Test_Vector{
{ []u8{0x00}, 0, 1 },
{ []u8{0xC0, 0xBB, 0x78}, -123456, 3 },
{ []u8{}, 0, 0 },
{ []u8{0x00}, 0, 1, .None },
{ []u8{0x3f}, 63, 1, .None },
{ []u8{0x40}, -64, 1, .None },
{ []u8{0xC0, 0xBB, 0x78}, -123456, 3, .None },
{ []u8{}, 0, 0, .Buffer_Too_Small },
}