fix: execute replication CREATE TABLE SQL from SHOW CREATE output

[somethingwithproof]

Summary

• execute the CREATE TABLE statement text ($create_sql) when a replicated rules table is missing
• initialize and guard create_sql before execution in syslog_replace_data()
• if CREATE SQL cannot be derived, log a replication warning and return safely
• keep the existing truncate + bulk upsert replication flow unchanged
• add regression coverage for syslog_replace_data() table bootstrap path
• harden regression test isolation to avoid accidental real DB execution
• run regression scripts from CI when tests/regression/*.php exist
• add changelog entry for issue bug: syslog_replace_data executes wrong variable when creating missing table #258

Validation

• php -l setup.php
• php -l tests/regression/issue258_replication_create_sql_test.php
• php tests/regression/issue258_replication_create_sql_test.php

Fixes #258

[Copilot]

Pull request overview

Fixes syslog rule replication bootstrap when the destination rules table is missing by executing the actual CREATE TABLE statement obtained from SHOW CREATE TABLE, and adds a regression test + CI coverage to prevent regressions.

Changes:

• Fix syslog_replace_data() to execute $create_sql (from SHOW CREATE TABLE) when creating a missing replicated rules table.
• Add a standalone regression script validating the “missing table => CREATE + TRUNCATE + bulk upsert” path.
• Run regression scripts in CI when tests/regression exists, and add a changelog entry for issue #258.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
setup.php	Fixes CREATE TABLE execution during replication bootstrap for missing rules tables.
tests/regression/issue258_replication_create_sql_test.php	Adds regression coverage for the missing-table bootstrap path.
.github/workflows/plugin-ci-workflow.yml	Runs regression scripts in CI when present.
CHANGELOG.md	Documents the fix for issue #258.

[Copilot]

$create_sql is only set when the SHOW CREATE TABLE result contains expected keys. If db_fetch_row() returns an empty/false value (e.g., permissions issue, table missing, or unexpected column label), this will call syslog_db_execute($create_sql) with an undefined variable and fail to create the destination table. Initialize $create_sql (e.g., to an empty string) and add a guard that logs/returns early when the CREATE statement cannot be derived before executing it.

[Copilot]

This regression script conditionally defines DB wrapper stubs using function_exists(). If someone runs it in an environment where Cacti/syslog functions are already loaded (so the stubs are skipped), the script will call the real syslog_replace_data() against a real DB and may create/modify syslog rule tables before failing its assertions. Consider adding an explicit safety check up front (e.g., abort when core functions like db_fetch_row/syslog_db_execute already exist) to ensure the test only runs in an isolated/stubbed context.

[somethingwithproof]

Addressed review feedback in the latest commit:

• initialized and guarded create_sql before execution in syslog_replace_data()
• when SHOW CREATE TABLE output cannot be derived, log warning and return early instead of executing undefined SQL
• hardened regression test isolation to abort if DB wrapper functions already exist
• extended regression coverage to verify safe no-op behavior when CREATE SQL is unavailable