bugfix: scope partition maintenance to target table and lock create path

[somethingwithproof]

Summary

• scope partition create/check/remove queries to the requested table instead of hardcoded syslog
• add a table allowlist guard for partition maintenance helpers
• serialize partition creation using GET_LOCK/RELEASE_LOCK to avoid concurrent duplicate partition creation
• add regression test for table scoping and lock usage
• update changelog with issue bug: partition existence check hardcodes table name 'syslog', breaks syslog_removed and allows race condition #254

Fixes #254

Validation

• php -l functions.php
• php -l tests/regression/issue254_partition_table_locking_test.php
• php tests/regression/issue254_partition_table_locking_test.php

[Copilot]

Pull request overview

Fixes syslog partition maintenance so it correctly targets the requested table (e.g., syslog_removed), adds a guardrail for which tables are eligible for partition operations, and serializes partition creation to avoid concurrent “duplicate partition” errors.

Changes:

• Scope information_schema.partitions checks/removals to the $table argument instead of hardcoding syslog.
• Add syslog_partition_table_allowed() allowlist and use MySQL GET_LOCK/RELEASE_LOCK to serialize partition creation.
• Add a regression test that verifies table scoping, lock usage, and the presence of the allowlist helper; update changelog for #254.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

File	Description
functions.php	Scopes partition queries to the target table, adds allowlist guard, and adds locking around partition creation.
tests/regression/issue254_partition_table_locking_test.php	Adds a regression check ensuring table scoping and lock acquisition/release are present.
CHANGELOG.md	Records the fix for issue #254 under the develop section.

[Copilot]

The partition existence check query is built via string concatenation for table_schema, partition_name, and table_name. Since database.php provides prepared variants, use syslog_db_fetch_row_prepared() here to avoid quoting edge cases and reduce SQL injection risk (even with the allowlist).

Suggested change

	$exists = syslog_db_fetch_row("SELECT *
	FROM `information_schema`.`partitions`
	WHERE table_schema='" . $syslogdb_default . "'
	AND partition_name='" . $cformat . "'
	AND table_name='" . $table . "'
	ORDER BY partition_ordinal_position");
	$exists = syslog_db_fetch_row_prepared(
	"SELECT *
	FROM `information_schema`.`partitions`
	WHERE table_schema = ?
	AND partition_name = ?
	AND table_name = ?
	ORDER BY partition_ordinal_position",
	array($syslogdb_default, $cformat, $table)
	);

[somethingwithproof]

Fixed — using syslog_db_fetch_row_prepared() now.

[Copilot]

This information_schema.partitions query interpolates $syslogdb_default and $table directly into SQL. Consider switching to syslog_db_fetch_assoc_prepared() with parameters for table_schema / table_name to avoid quoting issues and keep partition-maintenance queries consistently parameterized.

[somethingwithproof]

Fixed.

[Copilot]

$syslogdb_default and $table are interpolated into this information_schema query. Using syslog_db_fetch_cell_prepared() with placeholders for table_schema / table_name would be safer and more robust if the configured schema name ever contains characters that require escaping.

[somethingwithproof]

Fixed.

[Copilot]

This regression test relies on exact source-code string matching (including specific quoting/whitespace), which is fragile and can fail on harmless refactors/formatting changes. Consider using a more tolerant pattern (e.g., regex allowing whitespace) or validating behavior via function calls against a test DB stub so the test asserts semantics rather than formatting.

Suggested change

	if (substr_count($functions, "table_name='\" . \$table . \"'") < 3) {
	$scopedTableCount = 0;
	foreach (preg_split('/\R/', $functions) as $line) {
	if (strpos($line, 'table_name') !== false && strpos($line, '$table') !== false) {
	$scopedTableCount++;
	}
	}
	if ($scopedTableCount < 3) {

[somethingwithproof]

The test verifies the exact code construct that implements the behavior; a content change would indicate a regression.