tar: Use UTF-8 as a default encoding

Use UTF-8 by default, as this has become common for cross-platform
archive distribution, and our previous default was platform-specific.

Also revert commit f590010167 (libarchive: Restore tolerance of archives
with non-OEM paths on Windows, 2026-01-23, v3.31.11~1^2).  Its primary
purpose was to tolerate UTF-8 paths when extracting archives, which we
now handle by default.

Issue: #26903

Help/cpack_gen/archive.rst

@@ -224,8 +224,11 @@ CPack generators which are essentially archives at their core. These include:
 
   Specify the pathname character encoding used in package archives.
 
-  :Default: ``OEM``
+  :Default: ``UTF-8``
 
   See the :option:`cmake -E tar <cmake-E tar>` tool's
   :option:`--cmake-tar-encoding <cmake-E_tar --cmake-tar-encoding>` flag
   for supported encoding names.
+
+  In CMake 4.3 and below, the ``OEM`` encoding (current locale)
+  was always used.

Help/manual/cmake.1.rst

@@ -1705,12 +1705,15 @@ Available commands are:
     ``UTF-8``
       Archive pathnames are encoded as UTF-8.
 
+      This is the default since CMake 4.4.
+
     ``OEM``
       On Windows platforms, pathnames are encoded as using the original
       equipment manufacturer (OEM) code page.  On non-Windows platforms,
       pathnames are encoded according to the current locale.
 
-      This is the default value.
+      In CMake 4.3 and below, the ``OEM`` encoding (current locale)
+      was always used.
 
     ``UTF-16LE``, ``UTF-16BE``
       Archive pathnames are encoded as UTF-16 little-endian or big-endian.

Source/CPack/cmCPackArchiveGenerator.cxx

@@ -614,7 +614,7 @@ int cmCPackArchiveGenerator::GetCompressionLevel() const
 
 std::string cmCPackArchiveGenerator::GetEncoding() const
 {
-  std::string encoding = "OEM";
+  std::string encoding = "UTF-8";
 
   if (cmValue v = this->GetOptionIfSet("CPACK_ARCHIVE_ENCODING")) {
     encoding = *v;

Source/CPack/cmCPackDebGenerator.cxx

@@ -213,7 +213,7 @@ bool DebGenerator::generateDataTar() const
   }
   cmArchiveWrite data_tar(
     fileStream_data_tar, this->TarCompressionType, this->DebianArchiveType,
-    "OEM", this->CompressionLevel, static_cast<int>(this->NumThreads));
+    "UTF-8", this->CompressionLevel, static_cast<int>(this->NumThreads));
   if (!data_tar.Open()) {
     cmCPackLogger(cmCPackLog::LOG_ERROR,
                   "Error opening the archive \""
@@ -344,7 +344,7 @@ bool DebGenerator::generateControlTar(std::string const& md5Filename) const
     return false;
   }
   cmArchiveWrite control_tar(fileStream_control_tar, this->TarCompressionType,
-                             this->DebianArchiveType, "OEM",
+                             this->DebianArchiveType, "UTF-8",
                              this->CompressionLevel);
   if (!control_tar.Open()) {
     cmCPackLogger(cmCPackLog::LOG_ERROR,

Source/cmArchiveWrite.h

@@ -56,8 +56,8 @@ public:
   /** Construct with output stream to which to write archive.  */
   cmArchiveWrite(std::ostream& os, Compress c = CompressNone,
                  std::string const& format = "paxr",
-                 std::string const& encoding = "OEM", int compressionLevel = 0,
-                 int numThreads = 1);
+                 std::string const& encoding = "UTF-8",
+                 int compressionLevel = 0, int numThreads = 1);
 
   ~cmArchiveWrite();
 

Source/cmCTest.cxx

@@ -1252,7 +1252,7 @@ std::string cmCTest::Base64GzipEncodeFile(std::string const& file)
   files.push_back(file);
 
   if (!cmSystemTools::CreateTar(
-        tarFile, files, {}, cmSystemTools::TarCompressGZip, "OEM", false)) {
+        tarFile, files, {}, cmSystemTools::TarCompressGZip, "UTF-8", false)) {
     cmCTestLog(this, ERROR_MESSAGE,
                "Error creating tar while "
                "encoding file: "

Source/cmcmd.cxx

@@ -1920,7 +1920,7 @@ int cmcmd::ExecuteCMakeCommand(std::vector<std::string> const& args,
         cmSystemTools::TarCompressAuto;
       int nCompress = 0;
       bool doing_options = true;
-      std::string encoding = "OEM";
+      std::string encoding = "UTF-8";
       for (auto const& arg : cmMakeRange(args).advance(4)) {
         if (doing_options && cmHasLiteralPrefix(arg, "--")) {
           if (arg == "--") {

Tests/Fuzzing/cmArchiveExtractFuzzer.cxx

@@ -87,7 +87,7 @@ extern "C" int LLVMFuzzerTestOneInput(uint8_t const* data, size_t size)
 
     // Extract without verbose, with timestamps
     bool result1 = cmSystemTools::ExtractTar(
-      archiveFile, files, cmSystemTools::cmTarExtractTimestamps::Yes, "OEM",
+      archiveFile, files, cmSystemTools::cmTarExtractTimestamps::Yes, "UTF-8",
       false);
     (void)result1;
 
@@ -103,7 +103,7 @@ extern "C" int LLVMFuzzerTestOneInput(uint8_t const* data, size_t size)
       // Extract with verbose, without timestamps
       files.clear();
       bool result2 = cmSystemTools::ExtractTar(
-        archiveFile, files, cmSystemTools::cmTarExtractTimestamps::No, "OEM",
+        archiveFile, files, cmSystemTools::cmTarExtractTimestamps::No, "UTF-8",
         true);
       (void)result2;
 

Utilities/cmlibarchive/libarchive/archive_string.c

@@ -575,12 +575,7 @@ archive_wstring_append_from_mbs_in_codepage(struct archive_wstring *dest,
 		} else
 			mbflag = MB_PRECOMPOSED;
 
-		/* FIXME(CMake#26903): Offer control over encoding conversion.
-		   For now, we instead tolerate invalid characters as
-		   libarchive 3.7.2 / CMake 3.30 and below did.  */
-		#if 0
 		mbflag |= MB_ERR_INVALID_CHARS;
-		#endif
 
 		buffsize = dest->length + length + 1;
 		do {