Add parser for LEB128 integers. (#31)

* Add parser for LEB128 integers.
* Add tests for LEB128 overflow. Fixed overflow check and added a max byte count check.

Author: willtemperley

Sources/BinaryParsing/Parsers/Integer.swift

@@ -696,6 +696,86 @@ extension FixedWidthInteger where Self: BitwiseCopyable {
     }
     self = try Self(_throwing: T(truncatingIfNeeded: result))
   }
+
+  /// Creates an integer by parsing a little-endian base 128 (LEB128) encoded value of this type's size
+  /// from the start of the given parser span.
+  ///
+  /// - Parameter input: The `ParserSpan` to parse from. If parsing succeeds,
+  ///   the start position of `input` is moved forward by the number of bytes consumed. This will
+  ///   usually be `ceil(N / 7)` where N is the minimum number of bits required to encode
+  ///   this integer. In rare cases an encoder may produce valid but unnecessary padding bytes,
+  ///   in which case the number of bytes consumed can be up to `ceil(bitWidth / 7)` where
+  ///   bitWidth is the  full width of this type.
+  /// - Throws: A `ParsingError` if `input` overflows the max value of  this integer type,
+  ///   or if the maximum byte count for this type's size has been consumed.
+  @inlinable
+  @_lifetime(&input)
+  public init(parsingLEB128 input: inout ParserSpan) throws(ParsingError) {
+    var result: Self = 0
+    var shift = 0
+    var byte: UInt8 = 0
+    while true {
+      byte = try UInt8(parsing: &input)
+      let lowBits = byte & 0x7F
+      let availableBits = Self.bitWidth - shift
+      let isFinalByte = (byte & 0x80) == 0
+      if availableBits <= 0 {
+        let maxBytes = (Self.bitWidth + 6) / 7
+        let byteCount = shift / 7 + 1
+        if byteCount > maxBytes {
+          throw ParsingError(
+            status: .invalidValue,
+            location: input.startPosition)
+        }
+        // Allow padding bytes that do not affect the value
+        let expectedBits: UInt8 = (result < 0) ? 0x7F : 0x00
+        guard lowBits == expectedBits else {
+          throw ParsingError(
+            status: .invalidValue,
+            location: input.startPosition)
+        }
+      } else if availableBits < 7 {
+        let allowedMask: UInt8 = (1 &<< availableBits) &- 1
+        let extraBits: UInt8 = lowBits & ~allowedMask
+        if Self.isSigned {
+          let signPadding: UInt8 = (~allowedMask) & 0x7F
+          guard extraBits == signPadding || extraBits == 0 else {
+            throw ParsingError(
+              status: .invalidValue, location: input.startPosition)
+          }
+        } else {
+          guard extraBits == 0 else {
+            throw ParsingError(
+              status: .invalidValue,
+              location: input.startPosition)
+          }
+        }
+        let part = Self(lowBits & allowedMask) << shift
+        result |= part
+        if Self.isSigned && isFinalByte {
+          let finalByteNegative = (byte & 0x40) != 0
+          let resultNegative = result & (1 << (Self.bitWidth - 1)) != 0
+          if finalByteNegative != resultNegative {
+            // The value's sign has flipped - it has wrapped around.
+            throw ParsingError(
+              status: .invalidValue,
+              location: input.startPosition)
+          }
+        }
+      } else {
+        result |= Self(lowBits) &<< shift
+      }
+      shift += 7
+      if isFinalByte { break }
+    }
+    if Self.isSigned {
+      // Sign-extend if needed
+      if shift < Self.bitWidth && (byte & 0x40) != 0 {
+        result |= (~0) << shift
+      }
+    }
+    self = result
+  }
 }
 
 extension RawRepresentable where RawValue: MultiByteInteger {

Tests/BinaryParsingTests/IntegerParsingTests.swift

@@ -171,6 +171,12 @@ struct IntegerParsingTests {
           }
         }
       }
+
+      do {
+        let lebEncoded = [UInt8](encodingLEB128: number)
+        let parsed = try lebEncoded.withParserSpan { try T(parsingLEB128: &$0) }
+        #expect(parsed == number)
+      }
     }
 
     try runTest(for: .zero)
@@ -258,6 +264,12 @@ struct IntegerParsingTests {
           }
         }
       }
+
+      do {
+        let lebEncoded = [UInt8](encodingLEB128: number)
+        let parsed = try lebEncoded.withParserSpan { try T(parsingLEB128: &$0) }
+        #expect(parsed == number)
+      }
     }
 
     try runTest(for: .zero)
@@ -352,6 +364,12 @@ struct IntegerParsingTests {
           }
         }
       }
+
+      do {
+        let lebEncoded = [UInt8](encodingLEB128: number)
+        let parsed = try lebEncoded.withParserSpan { try T(parsingLEB128: &$0) }
+        #expect(parsed == number)
+      }
     }
 
     try runTest(for: .zero)
@@ -621,4 +639,20 @@ struct IntegerParsingTests {
     try fuzzIntegerCasting(
       UInt.self, loadingFrom: UInt64.self, using: &rng)
   }
+
+  // Some LEB128 encoders output padding bytes which are considered
+  // valid if the number of bytes does not exceed `ceil(bitWidth / 7)`.
+  @Test(arguments: [
+    ([0x80, 0x81, 0x80, 0x00], 0x80),
+    ([0xFF, 0x00], 0x7F),
+    ([0xFF, 0x80, 0x00], 0x7F),
+    ([0x80, 0x81, 0x00], 0x80),
+    ([0x80, 0x81, 0x80, 0x00], 0x80),
+    ([0xFE, 0xFF, 0x7F], -0x02),
+  ])
+  func validPaddingLEB128(input: [Int], expected: Int) throws {
+    let lebEncoded = input.map(UInt8.init)
+    let result = try lebEncoded.withParserSpan { try Int(parsingLEB128: &$0) }
+    #expect(result == expected)
+  }
 }

Tests/BinaryParsingTests/TestingSupport.swift

@@ -108,6 +108,35 @@ extension Array where Element == UInt8 {
       Swift.withUnsafeBytes(of: value.littleEndian, Array.init)
       + Array(repeating: paddingByte, count: paddingCount)
   }
+
+  init<T: FixedWidthInteger>(encodingLEB128 value: T) {
+    var out: [UInt8] = []
+    if T.isSigned {
+      var v = value
+      while true {
+        var byte = UInt8(truncatingIfNeeded: v)
+        v >>= 6  // Keep the sign bit
+        let done = v == 0 || v == -1
+        if done {
+          byte &= 0x7F
+        } else {
+          v >>= 1
+          byte |= 0x80
+        }
+        out.append(byte)
+        if done { break }
+      }
+    } else {
+      var v = value
+      repeat {
+        var byte = UInt8(truncatingIfNeeded: v)
+        v >>= 7
+        if v != 0 { byte |= 0x80 }
+        out.append(byte)
+      } while v != 0
+    }
+    self = out
+  }
 }
 
 /// A seeded random number generator type.

Tests/BinaryParsingTests/ThrowingOperationsTests.swift

@@ -172,4 +172,37 @@ struct ThrowingOperationsTests {
       }
     }
   }
+
+  @Test(arguments: [[0xFE, 0xFF, 0xFF, 0x7F]])
+  func tooManyPaddingBytesLEB128(_ input: [Int]) throws {
+    let lebEncoded = input.map(UInt8.init)
+    #expect(throws: ParsingError.self) {
+      try lebEncoded.withParserSpan { try Int16(parsingLEB128: &$0) }
+    }
+  }
+
+  @Test func overflowLEB128() async throws {
+    func overflowTest<
+      T: FixedWidthInteger & BitwiseCopyable, U: MultiByteInteger
+    >(
+      _ type: T.Type,
+      value: U,
+    ) throws {
+      let lebEncoded: [UInt8] = .init(encodingLEB128: value)
+      #expect(throws: ParsingError.self) {
+        try lebEncoded.withParserSpan { try T(parsingLEB128: &$0) }
+      }
+    }
+    for i in 1...100 {
+      try overflowTest(Int8.self, value: Int16(Int8.min) - Int16(i))
+      try overflowTest(Int8.self, value: Int16(Int8.max) + Int16(i))
+      try overflowTest(UInt8.self, value: UInt16(UInt8.max) + UInt16(i))
+      try overflowTest(Int16.self, value: Int32(Int16.min) - Int32(i))
+      try overflowTest(Int16.self, value: Int32(Int16.max) + Int32(i))
+      try overflowTest(UInt16.self, value: UInt32(UInt16.max) + UInt32(i))
+      try overflowTest(Int32.self, value: Int64(Int32.min) - Int64(i))
+      try overflowTest(Int32.self, value: Int64(Int32.max) + Int64(i))
+      try overflowTest(UInt32.self, value: UInt64(UInt32.max) + UInt64(i))
+    }
+  }
 }