Use ANSI encoding (GetACP) for tool stdout/stderr instead of OEM

documentation/wiki/ChangeWaves.md

@@ -32,6 +32,7 @@ Change wave checks around features will be removed in the release that accompani
 ### 18.8
 - [RAR task: across multiple input properties, resolve relative paths against the project directory (not the process current directory)](https://github.com/dotnet/msbuild/pull/13319)
 - [Console, parallel console, and terminal loggers print the paths of log files written by registered loggers (e.g. file logger and binary logger) as part of the end-of-build summary.](https://github.com/dotnet/msbuild/pull/13577)
+- [ToolTask and Exec now read tool stdout/stderr using the system ANSI code page (GetACP) instead of OEM (GetOEMCP), fixing garbled output from native tools on Western locales (e.g., MSVC link.exe on French Windows). Some older console-oriented tools still emit OEM-encoded output and may now be misdecoded; if this affects you, you can temporarily opt out with `MSBUILDDISABLEFEATURESFROMVERSION=18.8`, or for `Exec` specifically set `StdOutEncoding`/`StdErrEncoding` explicitly.](https://github.com/dotnet/msbuild/issues/12290)
 
 ### 18.7
 - [Copy task retries on ERROR_ACCESS_DENIED on non-Windows platforms to handle transient lock conflicts (e.g. macOS CoW filesystems)](https://github.com/dotnet/msbuild/issues/13463)

src/Framework/EncodingUtilities.cs

@@ -28,6 +28,7 @@ internal static class EncodingUtilities
         internal static readonly Encoding Utf8WithoutBom = new UTF8Encoding(encoderShouldEmitUTF8Identifier: false);
 
         private static Encoding s_currentOemEncoding;
+        private static Encoding s_currentAnsiEncoding;
 
         internal const string UseUtf8Always = "ALWAYS";
         internal const string UseUtf8Never = "NEVER";
@@ -86,6 +87,59 @@ internal static Encoding CurrentSystemOemEncoding
             }
         }
 
+        /// <summary>
+        /// Get the current system locale code page, ANSI version (GetACP). ANSI code pages are used
+        /// by most native Windows tools for string resources compiled into their binaries.
+        /// This differs from OEM code pages which are used for console I/O historically.
+        /// For example, on a French Windows system: ANSI = CP1252, OEM = CP850.
+        ///
+        /// Many native build tools (e.g., MSVC v141 link.exe, cl.exe) write output using the ANSI
+        /// code page rather than the OEM code page, so MSBuild must read with ANSI to decode correctly.
+        /// See: https://github.com/dotnet/msbuild/issues/12290
+        /// </summary>
+        internal static Encoding CurrentSystemAnsiEncoding
+        {
+            get
+            {
+                if (s_currentAnsiEncoding != null)
+                {
+                    return s_currentAnsiEncoding;
+                }
+
+#if FEATURE_ENCODING_DEFAULT
+                // On .NET Framework, Encoding.Default returns the system ANSI code page (GetACP()).
+                // e.g., CP1252 for Western European Windows. This is what most native Windows tools
+                // use when writing string resources to stdout.
+                s_currentAnsiEncoding = Encoding.Default;
+#else
+                // On .NET Core/5+, Encoding.Default is always UTF-8, so we must explicitly call GetACP().
+                s_currentAnsiEncoding = Encoding.UTF8; // fallback if GetACP() fails or on non-Windows
+
+                try
+                {
+#if FEATURE_WINDOWSINTEROP
+                    if (NativeMethods.IsWindows)
+                    {
+                        Encoding.RegisterProvider(CodePagesEncodingProvider.Instance);
+                        // GetACP() returns the system ANSI code page (e.g., 1252 for French Windows).
+                        s_currentAnsiEncoding = Encoding.GetEncoding((int)PInvoke.GetACP());
+                    }
+#endif
+                }
+                catch (ArgumentException ex)
+                {
+                    Debug.Assert(false, "GetEncoding(system ANSI code page) threw an ArgumentException in EncodingUtilities.CurrentSystemAnsiEncoding! Please log a bug against MSBuild.", ex.Message);
+                }
+                catch (NotSupportedException ex)
+                {
+                    Debug.Assert(false, "GetEncoding(system ANSI code page) threw a NotSupportedException in EncodingUtilities.CurrentSystemAnsiEncoding! Please log a bug against MSBuild.", ex.Message);
+                }
+#endif
+
+                return s_currentAnsiEncoding;
+            }
+        }
+
         /// <summary>
         /// Checks two encoding types to determine if they are similar to each other (equal or if
         /// the Encoding Name is the same).

src/Framework/NativeMethods.txt

@@ -52,6 +52,7 @@ GetLogicalProcessorInformationEx
 GetLongPathName
 GetModuleFileName
 GetNativeSystemInfo
+GetACP
 GetOEMCP
 GetProcAddress
 GetRunningObjectTable

src/Tasks.UnitTests/Exec_Tests.cs

@@ -494,12 +494,16 @@ public void ExecTaskUnicodeCharacterInCommand()
         }
 
         /// <summary>
-        /// Tests that Exec task will choose the default code page when UTF8 is not needed.
+        /// Tests that Exec task uses ANSI (GetACP) encoding by default when the command contains
+        /// only ASCII characters (no UTF-8 batch file override needed).
+        /// See: https://github.com/dotnet/msbuild/issues/12290
         /// </summary>
         [Fact]
         public void ExecTaskWithoutUnicodeCharacterInCommand()
         {
-            RunExec(false, EncodingUtilities.CurrentSystemOemEncoding.EncodingName);
+            // Default stdout encoding is ANSI (GetACP) so that native tool output (e.g., MSVC v141
+            // link.exe on French Windows) is decoded correctly instead of garbled as OEM-850.
+            RunExec(false, EncodingUtilities.CurrentSystemAnsiEncoding.EncodingName);
         }
 
         /// <summary>
@@ -521,26 +525,29 @@ public void ExecTaskUtf8AlwaysWithAnsi()
         }
 
         /// <summary>
-        /// Exec task will NOT use UTF8 when UTF8 Never is specified and non-ANSI characters are in the Command
-        /// <remarks>Exec task will fail as the cmd processor will not be able to run the command.</remarks>
+        /// Exec task will NOT use UTF8 when UTF8 Never is specified and non-ANSI characters are in the Command.
+        /// UseUtf8=Never only controls the batch file encoding (stays OEM, no chcp); the stdout reading
+        /// encoding remains ANSI (GetACP) independently of the batch file encoding.
+        /// <remarks>Exec task will fail as the cmd processor cannot run the garbled batch file.</remarks>
         /// </summary>
         [WindowsOnlyTheory]
         [InlineData("Never")]
         [InlineData("System")]
         public void ExecTaskUtf8NeverWithNonAnsi(string useUtf8)
         {
-            RunExec(true, EncodingUtilities.CurrentSystemOemEncoding.EncodingName, useUtf8, false);
+            RunExec(true, EncodingUtilities.CurrentSystemAnsiEncoding.EncodingName, useUtf8, false);
         }
 
         /// <summary>
-        /// Exec task will NOT use UTF8 when UTF8 Never is specified and only ANSI characters are in the Command
+        /// Exec task will NOT use UTF8 when UTF8 Never is specified and only ANSI characters are in the Command.
+        /// UseUtf8=Never controls the batch file encoding only; stdout reading encoding remains ANSI (GetACP).
         /// </summary>
         [Theory]
         [InlineData("Never")]
         [InlineData("System")]
         public void ExecTaskUtf8NeverWithAnsi(string useUtf8)
         {
-            RunExec(false, EncodingUtilities.CurrentSystemOemEncoding.EncodingName, useUtf8);
+            RunExec(false, EncodingUtilities.CurrentSystemAnsiEncoding.EncodingName, useUtf8);
         }
 
         [Theory]
@@ -1178,6 +1185,58 @@ public void Exec_RelativeWorkingDirectory_ResolvedAgainstProjectDirectory()
             exec.ValidateParametersAccessor().ShouldBeTrue();
             exec.GetWorkingDirectoryAccessor().ShouldBe(Path.Combine(projectDir.Path, "builddir"));
         }
+
+        #region stdout/stderr encoding tests
+
+        /// <summary>
+        /// Exec task constructor initialises stdout/stderr encoding to ANSI (GetACP) — not OEM (GetOEMCP).
+        /// This fixes garbled output from native tools (e.g., MSVC v141 link.exe on French Windows)
+        /// that write string resources using the ANSI code page, not OEM.
+        /// See: https://github.com/dotnet/msbuild/issues/12290
+        /// </summary>
+        [WindowsOnlyFact]
+        public void ExecTask_DefaultStdEncodingIsAnsi()
+        {
+            ChangeWaves.ResetStateForTests();
+            Exec exec = PrepareExec("echo test");
+            exec.StdOutEncoding.ShouldBe(EncodingUtilities.CurrentSystemAnsiEncoding.EncodingName);
+            exec.StdErrEncoding.ShouldBe(EncodingUtilities.CurrentSystemAnsiEncoding.EncodingName);
+        }
+
+        /// <summary>
+        /// When Wave18_8 is opted out, Exec falls back to OEM encoding — preserving pre-Wave18_8 behaviour.
+        /// </summary>
+        [WindowsOnlyFact]
+        public void ExecTask_Wave18_8_OptedOut_DefaultEncodingIsOem()
+        {
+            try
+            {
+                using TestEnvironment env = TestEnvironment.Create(_output);
+                ChangeWaves.ResetStateForTests();
+                env.SetEnvironmentVariable("MSBUILDDISABLEFEATURESFROMVERSION", ChangeWaves.Wave18_8.ToString());
+
+                Exec exec = PrepareExec("echo test");
+                exec.StdOutEncoding.ShouldBe(EncodingUtilities.CurrentSystemOemEncoding.EncodingName);
+                exec.StdErrEncoding.ShouldBe(EncodingUtilities.CurrentSystemOemEncoding.EncodingName);
+            }
+            finally
+            {
+                ChangeWaves.ResetStateForTests();
+            }
+        }
+
+        /// <summary>
+        /// UseUtf8Encoding=Always overrides the ANSI default: when the batch file needs UTF-8
+        /// (chcp 65001 is injected), the stdout encoding is also switched to UTF-8.
+        /// UseUtf8 controls the batch-file encoding; the chcp injection then also changes stdout reading.
+        /// </summary>
+        [WindowsOnlyFact]
+        public void ExecTask_UseUtf8AlwaysOverridesAnsiDefault()
+        {
+            RunExec(false, new UTF8Encoding(false).EncodingName, "Always");
+        }
+
+        #endregion
     }
 
     internal sealed class ExecWrapper : Exec

src/Tasks/Exec.cs

@@ -33,12 +33,21 @@ public Exec()
         {
             Command = string.Empty;
 
-            // Console-based output uses the current system OEM code page by default. Note that we should not use Console.OutputEncoding
-            // here since processes we run don't really have much to do with our console window (and also Console.OutputEncoding
-            // doesn't return the OEM code page if the running application that hosts MSBuild is not a console application).
-            // If the cmd file contains non-ANSI characters encoding may change.
-            _standardOutputEncoding = EncodingUtilities.CurrentSystemOemEncoding;
-            _standardErrorEncoding = EncodingUtilities.CurrentSystemOemEncoding;
+            // Wave18_8: use ANSI code page (GetACP) instead of OEM (GetOEMCP). Most native Windows tools
+            // write output using the ANSI code page (e.g., MSVC link.exe on French Windows: ANSI=CP1252,
+            // OEM=CP850). Reading with OEM garbles non-ASCII characters (e.g., 'é' → 'Ú').
+            // See: https://github.com/dotnet/msbuild/issues/12290
+            // If the cmd file contains non-ANSI characters the encoding may change later (see CreateTemporaryBatchFile).
+            if (ChangeWaves.AreFeaturesEnabled(ChangeWaves.Wave18_8))
+            {
+                _standardOutputEncoding = EncodingUtilities.CurrentSystemAnsiEncoding;
+                _standardErrorEncoding = EncodingUtilities.CurrentSystemAnsiEncoding;
+            }
+            else
+            {
+                _standardOutputEncoding = EncodingUtilities.CurrentSystemOemEncoding;
+                _standardErrorEncoding = EncodingUtilities.CurrentSystemOemEncoding;
+            }
         }
 
         #endregion

src/Utilities/ToolTask.cs

@@ -238,49 +238,54 @@ public virtual string ToolExe
         /// Overridable property specifying the encoding of the captured task standard output stream
         /// </summary>
         /// <remarks>
-        /// Console-based output uses the current system OEM code page by default. Note that we should not use Console.OutputEncoding
-        /// here since processes we run don't really have much to do with our console window (and also Console.OutputEncoding
-        /// doesn't return the OEM code page if the running application that hosts MSBuild is not a console application).
+        /// Most native Windows tools (e.g., MSVC v141 link.exe, cl.exe) write their string resources
+        /// using the system ANSI code page (GetACP), not the OEM code page (GetOEMCP).
+        /// On a French Windows system for example: ANSI = CP1252, OEM = CP850.
+        /// Reading tool output with OEM-850 when the tool outputs CP1252 causes garbled non-ASCII
+        /// characters (e.g., 'é' → 'Ú', 'à' → 'Ó'). Using ANSI encoding fixes this mismatch.
+        /// See: https://github.com/dotnet/msbuild/issues/12290
         /// </remarks>
-        protected virtual Encoding StandardOutputEncoding
+        protected virtual Encoding StandardOutputEncoding => GetDefaultToolEncoding();
+
+        /// <summary>
+        /// Returns the encoding to use for reading standard output and error streams from child tool processes.
+        /// Priority: UTF-8 encoding set by UseUtf8Encoding (Wave17_10) > ANSI (Wave18_8) > OEM legacy.
+        /// </summary>
+        private Encoding GetDefaultToolEncoding()
         {
-            get
+            if (ChangeWaves.AreFeaturesEnabled(ChangeWaves.Wave17_10))
             {
-                if (ChangeWaves.AreFeaturesEnabled(ChangeWaves.Wave17_10))
+                if (_encoding != null)
                 {
-                    if (_encoding != null)
-                    {
-                        // Keep the encoding of standard output & error consistent with the console code page.
-                        return _encoding;
-                    }
+                    // Keep the encoding of standard output & error consistent with the console code page.
+                    return _encoding;
                 }
-                return EncodingUtilities.CurrentSystemOemEncoding;
             }
+
+            // Wave18_8: use ANSI code page (GetACP) rather than OEM (GetOEMCP). Most native Windows tools
+            // compile their string resources with the ANSI code page, so reading with OEM garbles
+            // non-ASCII characters (e.g., 'é' → 'Ú' on French Windows: ANSI=CP1252, OEM=CP850).
+            // See: https://github.com/dotnet/msbuild/issues/12290
+            if (ChangeWaves.AreFeaturesEnabled(ChangeWaves.Wave18_8))
+            {
+                return EncodingUtilities.CurrentSystemAnsiEncoding;
+            }
+
+            return EncodingUtilities.CurrentSystemOemEncoding;
         }
 
         /// <summary>
         /// Overridable property specifying the encoding of the captured task standard error stream
         /// </summary>
         /// <remarks>
-        /// Console-based output uses the current system OEM code page by default. Note that we should not use Console.OutputEncoding
-        /// here since processes we run don't really have much to do with our console window (and also Console.OutputEncoding
-        /// doesn't return the OEM code page if the running application that hosts MSBuild is not a console application).
+        /// Most native Windows tools (e.g., MSVC v141 link.exe, cl.exe) write their string resources
+        /// using the system ANSI code page (GetACP), not the OEM code page (GetOEMCP).
+        /// On a French Windows system for example: ANSI = CP1252, OEM = CP850.
+        /// Reading tool output with OEM-850 when the tool outputs CP1252 causes garbled non-ASCII
+        /// characters (e.g., 'é' → 'Ú', 'à' → 'Ó'). Using ANSI encoding fixes this mismatch.
+        /// See: https://github.com/dotnet/msbuild/issues/12290
         /// </remarks>
-        protected virtual Encoding StandardErrorEncoding
-        {
-            get
-            {
-                if (ChangeWaves.AreFeaturesEnabled(ChangeWaves.Wave17_10))
-                {
-                    if (_encoding != null)
-                    {
-                        // Keep the encoding of standard output & error consistent with the console code page.
-                        return _encoding;
-                    }
-                }
-                return EncodingUtilities.CurrentSystemOemEncoding;
-            }
-        }
+        protected virtual Encoding StandardErrorEncoding => GetDefaultToolEncoding();
 
         /// <summary>
         /// Gets the Path override value.