4 Commits (feature/pgsql-native-backend-protocol)

Author	SHA1	Message	Date
Rene Cannao	51c44d10b2	TAP: parameterize session_track_variables helpers; add time_zone and sql_mode coverage ... Context ------- The four session_track_variables TAP tests previously hard-coded a single variable, 'innodb_lock_wait_timeout', for their round-trip assertion. That only exercised the integer path in 'MySQL_Session::handler_rc0_Process_Variables', leaving the string-value path (and the character-set special case elsewhere in the same function) untested. The stored-procedure helper was also named generically ('set_session_variable') while its body was pinned to InnoDB's lock wait timeout -- confusing for readers. Changes ------- * test/tap/tests/session_track_variables.h: - Introduce 'struct tracked_var_spec { name, set_expr; }' so the helpers can exercise arbitrary session variables from a single spec-driven code path. - 'set_session_variable()' now takes a spec, builds a per-variable stored procedure name ('test.set_session_track_<name>') so multiple specs can coexist without stepping on each other, and guarantees that the CALL is the last statement on the connection before the tracker is read (any query in between would reset the session tracker state). - Add 'drop_session_variable_proc()' for optional teardown. Kept separate from the set path so tests that need to preserve the CALL as the last statement can skip cleanup. - Switch the whole value surface from 'int' to 'std::string': 'extract_sess_var_mysql_pkt', 'select_sess_var_value', and 'extract_sess_var_proxy_internal' all return strings, which is how session tracking delivers values on the wire and matches both integer and string variables without encoding assumptions. - Provide a default set, 'session_track_default_vars', with three representative cases: * 'innodb_lock_wait_timeout' (integer; original coverage) * 'time_zone' (string, portable small value-space) * 'sql_mode' (string, representative of production usage) All three 'set_expr' values are pure server-side SQL (no client-side randomization), so a CALL alone produces a fresh value per run. * test/tap/tests/mysql-session_track_variables_optional-t.cpp, test/tap/tests/mysql-session_track_variables_enforced-t.cpp, test/tap/tests/mysql-session_track_variables_ff-t.cpp, test/tap/tests/mysql-session_track_variables_ff_mysql56-t.cpp: - Iterate over 'session_track_default_vars' and emit one TAP 'ok' per (spec[, mode]) combination. - Adjust plans accordingly. The ENFORCED-on-5.6 case still emits a single 'ok' because the backend is rejected before any per-variable tracking is attempted, so multiplying by the spec count there would be misleading. - Compare values as strings using the new string-returning helpers. The broadened coverage does not change what 'handler_rc0_Process_Variables' or 'handle_session_track_capabilities' do; it only exercises more of their existing behavior, notably the non-integer value path and the 'proxysql_find_charset_*' branches that activate for character-set variables (not exercised by these three specs but now reachable through the same harness for follow-up tests).	2 months ago
Wazir Ahmed	8d94919b22	fix: preserve EOF capability in enforced mode ... - Keep `CLIENT_DEPRECATE_EOF` enabled during frontend handshake when `session_track_variables=ENFORCED`. - Avoid unsafe mysql_error() calls in session-track TAP tests. Signed-off-by: Wazir Ahmed <wazir@proxysql.com>	2 months ago
Wazir Ahmed	0f5241f9f6	TAP: Update mysql_user for session variable tracking tests ... Signed-off-by: Wazir Ahmed <wazir@proxysql.com>	5 months ago
Wazir Ahmed	fffba7c90f	TAP: Restructure tests for session variable tracking ... Signed-off-by: Wazir Ahmed <wazir@proxysql.com>	5 months ago