ext/snmp: Promote 4 E_WARNING to ValueError for invalid input

[somethingwithproof]

Description

Four validation checks in ext/snmp still emit E_WARNING and return false for invalid input arguments, rather than throwing ValueError as is standard practice since PHP 8.0.

Affected locations

1. Malformed IPv6 address (snmp_session_init) — hostname like [dead:beef:: (missing closing ]) emits E_WARNING instead of throwing ValueError.

2. Type array too short (php_snmp_parse_oid) — when snmpset() receives fewer types than OIDs, emits E_WARNING("no type set") instead of ValueError.

3. Value array too short (php_snmp_parse_oid) — same pattern as JUNIT support #2 for values.

4. Invalid context engine ID (snmp_session_set_contextEngineID) — non-hex string emits E_WARNING. The existing code already has a // TODO Promote to Error? comment requesting this change.

Notes

• The IPv6 validation is called from both the procedural API (hostname = arg 1) and the OO SNMP::__construct (hostname = arg 2). A plain zend_value_error() is appropriate rather than zend_argument_value_error() to avoid showing a wrong argument position.
• The generate_Ku() warnings (authentication/privacy key generation, lines ~1089/1105) are intentionally left as E_WARNING since those can fail for runtime reasons (library configuration, unsupported algorithm), not purely bad input.
• This is a backward compatibility break: code relying on set_error_handler() + checking for false returns will now get uncaught ValueError exceptions.

PHP Version

master (targeting next minor/major)

[prateekbhujel]

Opened PR #21319 to address this:\n\n- Promotes malformed IPv6 host validation from warning to ValueError\n- Promotes missing type/value array entries in snmpset/snmp2_set from warning to ValueError\n- Promotes invalid contextEngineID from warning to ValueError\n\nUpdated tests in ext/snmp/tests accordingly (ipv6, snmpset, snmp2_set, and SNMP::setSecurity error case).

[prateekbhujel]

Fixed by #21319, which was merged into master on April 6, 2026. I believe this can be closed now.

[Girgias]

Closing as the linke PR did convert these to ValueErrors