respond to copilot -- further fixes and undo copilot's proposal

Author: tballison

tika-core/src/main/resources/org/apache/tika/mime/tika-mimetypes.xml

@@ -356,15 +356,15 @@
     <acronym>AXML</acronym>
     <_comment>Android Binary XML</_comment>
     <tika:link>https://developer.android.com/guide/topics/manifest/manifest-intro</tika:link>
-    <!-- Compiled binary form of AndroidManifest.xml and res/*.xml inside an APK.
-         These carry a .xml extension and live inside a zip, so the *.xml glob would
-         otherwise route them to application/xml and the XML parser, which fails on the
-         binary header. Detect by the leading ResChunk_header: type=RES_XML_TYPE(0x0003)
-         + headerSize(0x0008) -> 0x00080003 little-endian. The following 4 bytes are the
-         per-file chunk size, so the signature is only the first 4 bytes. Deliberately
-         NOT a sub-class-of application/xml: it is not XML and must not reach an XML parser. -->
-    <magic priority="50">
-      <match value="0x03000800" type="string" offset="0"/>
+    <!-- Compiled AndroidManifest.xml / res/*.xml inside an APK. The .xml extension would
+         otherwise route it to the XML parser, which fails on the binary header. Signature:
+         RES_XML_TYPE(0x0003)+headerSize(0x0008)=0x00080003 LE, plus RES_STRING_POOL_TYPE
+         (0x0001) at offset 8 (the variable per-file size at offset 4 is skipped). Not a
+         sub-class-of application/xml: must not reach an XML parser. -->
+    <magic priority="50">
+      <match value="0x03000800" type="string" offset="0">
+        <match value="0x0001" type="little16" offset="8"/>
+      </match>
     </magic>
   </mime-type>
   <mime-type type="application/x-tika-java-enterprise-archive">

tika-parsers/tika-parsers-standard/tika-parsers-standard-integration-tests/src/test/java/org/apache/tika/parser/AndroidBinaryXMLTest.java

@@ -52,16 +52,18 @@ public class AndroidBinaryXMLTest extends TikaTest {
     private static final String AXML = "application/vnd.android.axml";
 
     /**
-     * Minimal but structurally-plausible Android Binary XML header:
-     * ResChunk_header {type=RES_XML_TYPE(0x0003), headerSize=0x0008, size=<total>}
-     * followed by a zeroed ResStringPool_header. Only the leading 4 bytes (0x00080003 LE)
-     * are the detection signature; the following 4 bytes are the per-file size.
+     * Minimal compiled-AXML header: a RES_XML_TYPE ResChunk_header plus the ResStringPool
+     * chunk real AXML always carries. The magic matches 0x00080003 (LE) at offset 0 and the
+     * string-pool type 0x0001 at offset 8, so both must be present.
      */
     private static byte[] axmlBytes() {
         ByteBuffer bb = ByteBuffer.allocate(64).order(ByteOrder.LITTLE_ENDIAN);
         bb.putShort((short) 0x0003);   // RES_XML_TYPE
         bb.putShort((short) 0x0008);   // headerSize
-        bb.putInt(0x00000038);         // string pool chunk size
+        bb.putInt(64);                 // total chunk size == file length (skipped by magic)
+        bb.putShort((short) 0x0001);   // RES_STRING_POOL_TYPE (checked at offset 8)
+        bb.putShort((short) 0x001C);   // string-pool headerSize
+        bb.putInt(0x0000003C);         // string-pool chunk size
         // remaining bytes (string/style counts, flags, offsets) left zero
         return bb.array();
     }