[typemap] Validate manifest aliases and DEX names

Resolve manifest placeholders before validation and alias rewriting, decode DEX and classfile modified UTF-8 exactly, and record the supplementary class-loader limitation.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Author: simonrozsival

Documentation/docs-mobile/messages/xa4258.md

@@ -22,8 +22,8 @@ with the .NET runtime used to build the app.
 
 .NET for Android requires Normalization Form C (NFC) names and rejects combining and format
 characters because normalizing file systems can change or merge generated Java source paths.
-Supplementary characters are also rejected because Android tooling does not retain those class
-names in DEX files.
+Supplementary characters are rejected because Android's class loader cannot resolve classes with
+supplementary characters in their simple name, even when the exact descriptor is present in DEX.
 
 This can originate from the `$(ApplicationId)` MSBuild property, the `package` attribute in `AndroidManifest.xml`, a managed type name, or an explicit Java name supplied by an attribute such as `[Register]` or `[JniTypeSignature]`.
 

external/Java.Interop/src/Xamarin.Android.Tools.Bytecode/ConstantPool.cs

@@ -384,31 +384,36 @@ public ConstantPoolUtf8Item (ConstantPool constantPool, Stream stream)
 			for (int i = 0; i < data.Length; ++i)
 				data [i] = stream.ReadNetworkByte ();
 
-			// The .class file specially encodes NUL so that it takes 2 bytes, not 1.
-			// http://en.wikipedia.org/wiki/UTF-8#Modified_UTF-8
-			var fixup   = new List<byte> (data.Length);
+			// Modified UTF-8 encodes UTF-16 code units, so supplementary characters remain
+			// a high-surrogate/low-surrogate pair after decoding.
+			var decoded = new StringBuilder (data.Length);
 			for (int i = 0; i < data.Length; ++i) {
-				if (data [i] == 0xc0 && (i + 1) < data.Length && data [i + 1] == 0x80) {
-					fixup.Add (0x00);
-					i++;
+				byte first = data [i];
+				if ((first & 0x80) == 0) {
+					decoded.Append ((char) first);
 					continue;
 				}
-				// ...and they couldn't be bothered with supporting 4-byte UTF-8 sequences,
-				// needed for Emoji and chars off the Basic Multilingual Plane; instead, they're
-				// encoded as a surrogate pair. (What is this I don't even...)
-				if (data [i] == 0xed && i+6 < data.Length && data [i+3] == 0xed) {
-					var surrogatePair = new char [] {
-						(char) (0xD800 + (((data [i+1] & 0x0F) << 6) | (data [i+2] & 0x3F))),
-						(char) (0xDC00 + (((data [i+4] & 0x0F) << 6) | (data [i+5] & 0x3F))),
-					};
-					fixup.AddRange (Encoding.UTF8.GetBytes (surrogatePair));
-					i += 5;
+				if ((first & 0xe0) == 0xc0 && i + 1 < data.Length) {
+					byte second = ReadContinuationByte (data [++i]);
+					decoded.Append ((char) (((first & 0x1f) << 6) | (second & 0x3f)));
+					continue;
+				}
+				if ((first & 0xf0) == 0xe0 && i + 2 < data.Length) {
+					byte second = ReadContinuationByte (data [++i]);
+					byte third = ReadContinuationByte (data [++i]);
+					decoded.Append ((char) (((first & 0x0f) << 12) | ((second & 0x3f) << 6) | (third & 0x3f)));
 					continue;
-
 				}
-				fixup.Add (data [i]);
+				throw new InvalidDataException ($"Invalid modified UTF-8 lead byte 0x{first:x2}.");
 			}
-			value   = Encoding.UTF8.GetString (fixup.Count == data.Length ? data : fixup.ToArray ());
+			value = decoded.ToString ();
+		}
+
+		static byte ReadContinuationByte (byte value)
+		{
+			if ((value & 0xc0) != 0x80)
+				throw new InvalidDataException ($"Invalid modified UTF-8 continuation byte 0x{value:x2}.");
+			return value;
 		}
 
 		public override ConstantPoolItemType Type {

external/Java.Interop/tests/Xamarin.Android.Tools.Bytecode-Tests/ClassFileTests.cs

@@ -1,4 +1,5 @@
 using System;
+using System.IO;
 using System.Reflection;
 
 using Xamarin.Android.Tools.Bytecode;
@@ -15,6 +16,31 @@ public void Constructor_Exceptions ()
 		{
 			Assert.Throws<ArgumentNullException> (() => new ClassFile (null));
 		}
+
+		[Test]
+		public void ModifiedUtf8_DecodesUtf16CodeUnits ()
+		{
+			using var poolStream = new MemoryStream (new byte [] { 0, 1 });
+			var pool = new ConstantPool (poolStream);
+
+			Assert.AreEqual (
+				"\ud000\ud001",
+				Decode (pool, new byte [] { 0xed, 0x80, 0x80, 0xed, 0x80, 0x81 })
+			);
+			Assert.AreEqual (
+				"\U00010400",
+				Decode (pool, new byte [] { 0xed, 0xa0, 0x81, 0xed, 0xb0, 0x80 })
+			);
+
+			static string Decode (ConstantPool pool, byte [] bytes)
+			{
+				using var stream = new MemoryStream ();
+				stream.WriteByte ((byte) (bytes.Length >> 8));
+				stream.WriteByte ((byte) bytes.Length);
+				stream.Write (bytes, 0, bytes.Length);
+				stream.Position = 0;
+				return new ConstantPoolUtf8Item (pool, stream).Value;
+			}
+		}
 	}
 }
-

src/Microsoft.Android.Sdk.TrimmableTypeMap/Generator/ManifestConstants.cs

@@ -1,3 +1,5 @@
+using System;
+using System.Collections.Generic;
 using System.Xml.Linq;
 
 namespace Microsoft.Android.Sdk.TrimmableTypeMap;
@@ -6,4 +8,13 @@ static class ManifestConstants
 {
 	public static readonly XNamespace AndroidNs = "http://schemas.android.com/apk/res/android";
 	public static readonly XName AttName = AndroidNs + "name";
+	public static readonly HashSet<string> ComponentElementNames = new (StringComparer.Ordinal) {
+		"application",
+		"activity",
+		"activity-alias",
+		"instrumentation",
+		"service",
+		"receiver",
+		"provider",
+	};
 }

src/Microsoft.Android.Sdk.TrimmableTypeMap/Generator/ManifestGenerator.cs

@@ -16,14 +16,6 @@ class ManifestGenerator
 	static readonly XNamespace AndroidNs = ManifestConstants.AndroidNs;
 	static readonly XName AttName = ManifestConstants.AttName;
 	static readonly char [] PlaceholderSeparators = [';'];
-	static readonly HashSet<string> ComponentElementNames = new (StringComparer.Ordinal) {
-		"application",
-		"activity",
-		"instrumentation",
-		"service",
-		"receiver",
-		"provider",
-	};
 
 	/// <summary>Warning code for library-manifest merge failures (maps to XA4302).</summary>
 	internal const int LibraryManifestMergeWarningCode = 4302;
@@ -60,6 +52,7 @@ class ManifestGenerator
 	// own package when they are relative (start with '.'). Mirrors ManifestDocument.ManifestAttributeFixups.
 	static readonly Dictionary<string, string []> ManifestAttributeFixups = new (StringComparer.Ordinal) {
 		{ "activity", ["name"] },
+		{ "activity-alias", ["name", "targetActivity"] },
 		{ "application", ["backupAgent"] },
 		{ "instrumentation", ["name"] },
 		{ "provider", ["name"] },
@@ -83,6 +76,9 @@ class ManifestGenerator
 		}
 
 		EnsureManifestAttributes (manifest);
+		// Template component names must be resolved before compat-name rewriting. Apply again
+		// after library-manifest merging so placeholders introduced by libraries are also covered.
+		ApplyPlaceholders (doc, ManifestPlaceholders, PackageName);
 		var app = EnsureApplicationElement (manifest);
 		var targetSdkVersionValue = GetTargetSdkVersionValue (manifest);
 
@@ -320,20 +316,22 @@ void RewriteCompatNames (XElement manifest, IReadOnlyList<JavaPeerInfo> allPeers
 				continue;
 			}
 
-			var nameAttr = element.Attribute (AttName);
-			if (nameAttr is null) {
+			var classNameAttr = element.Name.LocalName == "activity-alias"
+				? element.Attribute (AndroidNs + "targetActivity")
+				: element.Attribute (AttName);
+			if (classNameAttr is null) {
 				continue;
 			}
-			var resolved = ManifestNameResolver.Resolve (nameAttr.Value, packageName);
+			var resolved = ManifestNameResolver.Resolve (classNameAttr.Value, packageName);
 			if (compatToCrc.TryGetValue (resolved, out var crcName)) {
-				nameAttr.Value = crcName;
+				classNameAttr.Value = crcName;
 			}
 		}
 	}
 
 	static bool IsComponentElement (XElement element)
 	{
-		return element.Name.NamespaceName.Length == 0 && ComponentElementNames.Contains (element.Name.LocalName);
+		return element.Name.NamespaceName.Length == 0 && ManifestConstants.ComponentElementNames.Contains (element.Name.LocalName);
 	}
 
 	void EnsureManifestAttributes (XElement manifest)