[IMPALA-3732] Avro scanner can break on strings >= 2GB

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: Impala 2.7.0
• Fix Version/s: Impala 2.7.0, Impala 2.6.2
• Component/s: Backend
• Labels: correctness crash
• Target Version: Impala 2.7.0, Impala 2.5.2, Impala 2.6.2
• Assignee: Tim Armstrong
• Reporter: Tim Armstrong
• Created: 13/Jun/16 19:25
• Updated: 06/Sep/16 17:51
• Resolved: 15/Jun/16 15:24

Description

The avro scanner casts the string length to a 32-bit integer when reading strings. The avro spec states that strings can have 64-bit integer lengths. The truncation can overflow the 32-bit integer to a negative number.