From 35f6050ef76c2e1f60abd187d00a6afbad9c6b01 Mon Sep 17 00:00:00 2001
From: WuYongJie <549727955@qq.com>
Date: Fri, 14 Aug 2026 20:44:10 +0800
Subject: [PATCH] fix: sanitize legacy mirror text encoding

---
 .../Services/BusinessMirrorSynchronizer.php   | 16 ++++++++++++
 .../BusinessMirrorSynchronizerTest.php        | 25 +++++++++++++++++++
 2 files changed, 41 insertions(+)

diff --git a/modules/EcommerceFinance/Services/BusinessMirrorSynchronizer.php b/modules/EcommerceFinance/Services/BusinessMirrorSynchronizer.php
index 80802317..c46b391a 100644
--- a/modules/EcommerceFinance/Services/BusinessMirrorSynchronizer.php
+++ b/modules/EcommerceFinance/Services/BusinessMirrorSynchronizer.php
@@ -226,6 +226,8 @@ class BusinessMirrorSynchronizer
                     $row[$name] = substr($value, 0, 10);
                 } elseif (in_array($type, ['datetime', 'datetime2', 'smalldatetime'], true)) {
                     $row[$name] = preg_replace('/(\.\d{6})\d+/', '$1', str_replace('T', ' ', $value));
+                } elseif (in_array($type, ['char', 'nchar', 'varchar', 'nvarchar', 'text', 'ntext', 'xml'], true)) {
+                    $row[$name] = $this->normalizeText($value);
                 }
             }
 
@@ -233,6 +235,20 @@ class BusinessMirrorSynchronizer
         }, $rows);
     }
 
+    private function normalizeText(string $value): string
+    {
+        // A small number of legacy SQL Server rows contain embedded NUL bytes or
+        // malformed byte sequences. MariaDB correctly rejects those values as
+        // invalid utf8mb4. Do not guess a source encoding: preserve all valid
+        // UTF-8 and mark only irrecoverable bytes with the Unicode replacement
+        // character so one corrupt label cannot block the entire table mirror.
+        $value = str_replace("\0", '', $value);
+
+        return mb_check_encoding($value, 'UTF-8')
+            ? $value
+            : mb_scrub($value, 'UTF-8');
+    }
+
     private function quote(string $identifier): string
     {
         if (preg_match('/^[A-Za-z_][A-Za-z0-9_]*$/', $identifier) !== 1) {
diff --git a/tests/Unit/EcommerceFinance/BusinessMirrorSynchronizerTest.php b/tests/Unit/EcommerceFinance/BusinessMirrorSynchronizerTest.php
index 29925c62..a1205baa 100644
--- a/tests/Unit/EcommerceFinance/BusinessMirrorSynchronizerTest.php
+++ b/tests/Unit/EcommerceFinance/BusinessMirrorSynchronizerTest.php
@@ -143,6 +143,31 @@ class BusinessMirrorSynchronizerTest extends TestCase
         $this->assertSame([], $this->query->calls);
     }
 
+    public function test_full_sync_sanitizes_invalid_utf8_text_without_changing_valid_text(): void
+    {
+        $this->query->batches = [[
+            [
+                'Id' => 1,
+                'updateTime' => '2026-08-14 10:00:00',
+                'displayName' => hex2bin('e80044495900'),
+            ],
+            [
+                'Id' => 2,
+                'updateTime' => '2026-08-14 11:00:00',
+                'displayName' => '合法名称',
+            ],
+        ]];
+
+        $this->synchronizer->sync('test.dbo.records', true, 10, 1);
+
+        $rows = DB::connection('business_mirror')->table('test_records')->orderBy('Id')->pluck('displayName')->all();
+        $this->assertStringEndsWith('DIY', $rows[0]);
+        $this->assertNotSame(hex2bin('e80044495900'), $rows[0]);
+        $this->assertSame('合法名称', $rows[1]);
+        $this->assertTrue(mb_check_encoding($rows[0], 'UTF-8'));
+        $this->assertStringNotContainsString("\0", $rows[0]);
+    }
+
     public function test_empty_full_read_does_not_delete_previous_mirror_rows(): void
     {
         $this->query->batches = [[