[IMPALA-4237] Incorrect results when materializing a decimal in a data source scan node.

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: Impala 2.3.0, Impala 2.5.0, Impala 2.4.0, Impala 2.6.0, Impala 2.7.0
• Fix Version/s: Impala 2.6.3, Impala 2.8.0
• Component/s: Backend
• Labels: correctness
• Target Version: Impala 2.6.3, Impala 2.8.0
• Assignee: Alexander Behm
• Reporter: Alexander Behm
• Created: 01/Oct/16 04:42
• Updated: 15/Nov/16 13:05
• Resolved: 07/Oct/16 03:59

Description

When scanning a DECIMAL-typed column with precision <= 9 a data source scan node may return incorrect results.

The reason is that 8 bytes are accidentally written to a 4 byte slot because of a missing break in a switch. See data-source-scan-node.cc:

...
inline Status SetDecimalVal(const ColumnType& type, char* bytes, int len,
    void* slot) {
  uint8_t* buffer = reinterpret_cast<uint8_t*>(bytes);
  switch (type.GetByteSize()) {
    case 4: {
      Decimal4Value* val = reinterpret_cast<Decimal4Value*>(slot);
      if (UNLIKELY(len > sizeof(Decimal4Value) ||
          ParquetPlainEncoder::Decode(buffer, buffer + len, len, val) < 0)) {
        return Status(ERROR_INVALID_DECIMAL);
      } <-- missing break, falling through to case 8
    }
    case 8: {
      Decimal8Value* val = reinterpret_cast<Decimal8Value*>(slot);
      if (UNLIKELY(len > sizeof(Decimal8Value) ||
          ParquetPlainEncoder::Decode(buffer, buffer + len, len, val) < 0)) {
        return Status(ERROR_INVALID_DECIMAL);
      }
      break;
    }
...