This article was published on October 16, 2026.
Preface: The HPE NonStop ODBC/MX connectivity architecture operates through the NonStop SQL/MX Connectivity Service (MXCS), which implements the Open Database Connectivity (ODBC 3.5 / 3.0) standard to interface client applications with the ANSI-compliant NonStop SQL/MX database.
Background: The following instructions require the use of the accompanying infographics.
Missing Output Converters for SQL_LONGVARCHAR / CLOB
Attached infographic shown configured example decoding for SQL_CHAR and SQL_WCHAR, but you did not configure it for large text blocks.
• The Vulnerability: In HPE NonStop SQL/MX, large text fields (like chat histories, system prompts, or documents) are usually stored as CLOB or LONG VARCHAR types. These map to pyodbc.SQL_LONGVARCHAR.
• The Buffer Issue: Without an explicit decoding rule or an assigned output converter for SQL_LONGVARCHAR, pyodbc will either fallback to a default raw byte stream or miscalculate the buffer size required to hold the text. If an AI agent logs a massive 32k-token prompt, the NonStop ODBC driver may truncate the buffer midway, throwing a memory allocation fault or crashing the Python process.
The code explicitly forces pyodbc to encode/decode everything as iso-8859-1 (Latin-1). If an LLM pushes text into user_comments that contains modern characters like smart quotes (“ ”), em-dashes (—), emojis, or non-English characters, UTF-8 uses multi-byte sequences (3 to 4 bytes per character). iso-8859-1 expects exactly 1 byte per character.
• The Buffer Issue: When pyodbc attempts to read these multi-byte UTF-8 sequences out of a standard SQL/MX VARCHAR or CLOB using an iso-8859-1 decoder, it breaks the byte boundaries. This causes severe data corruption (mojibake) or abrupt truncation because the byte-length calculations between Python and the NonStop ODBC driver fall completely out of alignment.
Ref: ISO-8859-1 uses a fixed 1 byte per character for all 256 values, while UTF-8 is a variable-width encoding that uses 1 to 4 bytes depending on the character.
Remedy: Apply the HPE Security Patch: Because CVE-2026-79839 is an environmental security vulnerability, your infrastructure team must apply the official HPE emergency hotfixes for the NonStop ODBC/MX Server component to securely validate input boundaries at the OS level.
Vulnerability details: A buffer overflow vulnerability in HPE NonStop ODBC/MX Server may cause memory corruption, resulting in application crash.
Official announcement: Please refer to the link for details –