fix(xps): repair scanner XPS packages that MuPDF fails to parse

src/tests/xps-repair.test.ts

@@ -0,0 +1,57 @@
+// @vitest-environment node
+import { describe, it, expect } from 'vitest';
+import JSZip from 'jszip';
+import { repairXml, repairXpsPackage } from '@/js/utils/xps-repair';
+
+describe('repairXml', () => {
+  it('turns stray characters between attributes into plain spaces', () => {
+    expect(repairXml('<A B="1"　C="2"\0/>')).toBe('<A B="1" C="2"/>');
+  });
+
+  it('drops attributes MuPDF cannot read and fixes "/ >"', () => {
+    expect(repairXml('<A Ïd="x" Width="1" / >')).toBe('<A Width="1"/>');
+  });
+
+  it('keeps comments, CDATA, values with > and closing tags intact', () => {
+    const xml =
+      '<?xml version="1.0" encoding="windows-1252"?><!-- a > b --><A N="a>b"><![CDATA[<x y>]]></A >';
+    expect(repairXml(xml)).toBe(
+      '<?xml version="1.0" encoding="utf-8"?><!-- a > b --><A N="a>b"><![CDATA[<x y>]]></A>'
+    );
+  });
+});
+
+describe('repairXpsPackage', () => {
+  it('decodes BOM-less UTF-16 and joins interleaved pieces', async () => {
+    const page = '<FixedPage Width="816"/>';
+    const zip = new JSZip();
+    zip.file(
+      '_rels/.rels',
+      new Uint8Array(
+        Buffer.from('<?xml version="1.0" encoding="utf-16"?><R/>', 'utf16le')
+      )
+    );
+    zip.file('1.fpage/[1].last.piece', page.slice(10));
+    zip.file('1.fpage/[0].piece', page.slice(0, 10));
+    zip.file('a.png', new Uint8Array([0, 0xff, 1]));
+
+    const out = await JSZip.loadAsync(
+      await (
+        await repairXpsPackage(
+          new Blob([await zip.generateAsync({ type: 'arraybuffer' })])
+        )
+      ).arrayBuffer()
+    );
+
+    expect(await out.file('_rels/.rels')?.async('string')).toBe(
+      '<?xml version="1.0" encoding="utf-8"?><R/>'
+    );
+    expect(await out.file('1.fpage')?.async('string')).toBe(
+      '<FixedPage Width="816"/>'
+    );
+    expect(Array.from(await out.file('a.png')!.async('uint8array'))).toEqual([
+      0, 0xff, 1,
+    ]);
+    expect(out.file(/piece/)).toHaveLength(0);
+  });
+});