MDEV-26713 Set activeCodePage=UTF8 for windows programs

- Use corresponding entry in the manifest, as described in
https://docs.microsoft.com/en-us/windows/apps/design/globalizing/use-utf8-code-page

- If if ANSI codepage is UTF8 (i.e for Windows 1903 and later)
  Use UTF8 as default client charset
  Set console codepage(s) to UTF8, in case process is using console

- Allow some previously disabled MTR tests, that used Unicode for in "exec",
  for the recent Windows versions

cmake/win_compatibility.manifest

@@ -19,4 +19,9 @@
 
     </application>
   </compatibility>
+  <application>
+    <windowsSettings>
+      <activeCodePage xmlns="http://schemas.microsoft.com/SMI/2019/WindowsSettings">UTF-8</activeCodePage>
+    </windowsSettings>
+  </application>
 </asmv1:assembly>

mysql-test/include/check_utf8_cli.inc

@@ -0,0 +1,3 @@
+# Check if utf8 can be used on the command line for  --exec
+# The real check is done in the suite.pm
+#

mysql-test/include/no_utf8_cli.inc

@@ -0,0 +1,3 @@
+# Check if utf8 can't be used on the command line for  --exec
+# The real check is done in the suite.pm
+#

mysql-test/main/charset_client_win.test

@@ -1,2 +1,3 @@
 --source include/windows.inc
+--source include/no_utf8_cli.inc
 --exec chcp 1257 > NUL && $MYSQL --default-character-set=auto -e "select @@character_set_client"

mysql-test/main/charset_client_win_utf8mb4.result

@@ -0,0 +1,2 @@
+@@character_set_client
+utf8mb4

mysql-test/main/charset_client_win_utf8mb4.test

@@ -0,0 +1,3 @@
+--source include/windows.inc
+--source include/check_utf8_cli.inc
+--exec $MYSQL --default-character-set=auto -e "select @@character_set_client"

mysql-test/main/grant_utf8_cli.test

@@ -1,6 +1,5 @@
- # UTF8 parameters to mysql client do not work on Windows
---source include/not_windows.inc
 --source include/not_embedded.inc
+--source include/check_utf8_cli.inc
 
 #
 # Bug#21432 Database/Table name limited to 64 bytes, not chars, problems with multi-byte

mysql-test/suite.pm

@@ -87,6 +87,28 @@ sub skip_combinations {
   $skip{'main/ssl_verify_ip.test'} = 'x509v3 support required'
     unless $openssl_ver ge "1.0.2";
 
+  sub utf8_command_line_ok() {
+   if (IS_WINDOWS) {
+     # Can use UTF8 on command line since Windows 10 1903 (10.0.18362)
+     # or if OS codepage is set to UTF8
+     my($os_name, $os_major, $os_minor, $os_build, $os_id) = Win32::GetOSVersion();
+     if($os_major lt 10){
+       return 0;
+     } elsif($os_major gt 10 or $os_minor gt 0 or $os_build ge 18362){
+       return 1;
+     } elsif(Win32::GetACP() eq 65001) {
+       return 1;
+     }
+     return 0;
+   }
+   return 1;
+  }
+
+  $skip{'include/check_utf8_cli.inc'} = 'No utf8 command line support'
+    unless utf8_command_line_ok();
+
+  $skip{'include/no_utf8_cli.inc'} = 'Not tested with utf8 command line support'
+    unless !utf8_command_line_ok();
 
   %skip;
 }

mysys/charset.c

@@ -1517,9 +1517,15 @@ const char* my_default_csname()
   const char* csname = NULL;
 #ifdef _WIN32
   char cpbuf[64];
-  int cp = GetConsoleCP();
-  if (cp == 0)
-    cp = GetACP();
+  UINT cp;
+  if (GetACP() == CP_UTF8)
+    cp= CP_UTF8;
+  else
+  {
+    cp= GetConsoleCP();
+    if (cp == 0)
+      cp= GetACP();
+  }
   snprintf(cpbuf, sizeof(cpbuf), "cp%d", (int)cp);
   csname = my_os_charset_to_mysql_charset(cpbuf);
 #elif defined(HAVE_SETLOCALE) && defined(HAVE_NL_LANGINFO)

mysys/get_password.c

@@ -100,7 +100,7 @@ char *get_tty_password(const char *opt_message)
   /*
     Allocate output string, and convert UTF16 password to output codepage.
   */
-  cp= GetConsoleCP();
+  cp= GetACP() == CP_UTF8 ? CP_UTF8 : GetConsoleCP();
 
   if (!(to_len= WideCharToMultiByte(cp, 0, wbuf, -1, NULL, 0, NULL, NULL)))
     DBUG_RETURN(NULL);

mysys/my_getopt.c

@@ -38,7 +38,7 @@ static double getopt_double(char *arg, const struct my_option *optp, int *err);
 static void init_variables(const struct my_option *, init_func_p);
 static void init_one_value(const struct my_option *, void *, longlong);
 static void fini_one_value(const struct my_option *, void *, longlong);
-static int setval(const struct my_option *, void *, char *, my_bool);
+static int setval(const struct my_option *, void *, char *, my_bool, const char *);
 static char *check_struct_option(char *cur_arg, char *key_name);
 
 /*
@@ -133,6 +133,50 @@ double getopt_ulonglong2double(ulonglong v)
   return u.dbl;
 }
 
+#ifdef _WIN32
+/**
+
+ On Windows, if program is running in UTF8 mode, but some arguments are not UTF8.
+
+ This will mostly likely be a sign of old "ANSI" my.ini, and it is likely that
+ something will go wrong, e.g file access error.
+*/
+static void validate_value(const char *key, const char *value,
+                           const char *filename)
+{
+  MY_STRCOPY_STATUS status;
+  const struct charset_info_st *cs= &my_charset_utf8mb4_bin;
+  size_t len;
+  if (GetACP() != CP_UTF8)
+    return;
+  if (!(len= strlen(value)))
+    return;
+  cs->cset->well_formed_char_length(cs, value, value + len, len, &status);
+  if (!status.m_well_formed_error_pos)
+    return;
+  if (filename && *filename)
+  {
+    my_getopt_error_reporter(WARNING_LEVEL,
+      "%s: invalid (non-UTF8) characters found for option '%s'"
+      " in file '%s'",
+      my_progname, key, filename);
+  }
+  else
+  {
+    /*
+     Should never happen, non-UTF8 can be read from option's
+     file only.
+    */
+    DBUG_ASSERT(0);
+    my_getopt_error_reporter(
+    WARNING_LEVEL, "%s: invalid (non-UTF8) characters for option %s",
+      my_progname, key);
+  }
+}
+#else
+#define validate_value(key, value, filename) (void)filename
+#endif
+
 /**
   Handle command line options.
   Sort options.
@@ -564,7 +608,7 @@ int handle_options(int *argc, char ***argv, const struct my_option *longopts,
 		}
 	      }
 	      if ((error= setval(optp, optp->value, argument,
-				 set_maximum_value)))
+                             set_maximum_value,filename)))
 		DBUG_RETURN(error);
               if (get_one_option(optp, argument, filename))
                 DBUG_RETURN(EXIT_UNSPECIFIED_ERROR);
@@ -610,7 +654,7 @@ int handle_options(int *argc, char ***argv, const struct my_option *longopts,
 	continue;
       }
       if ((!option_is_autoset) &&
-	  ((error= setval(optp, value, argument, set_maximum_value))) &&
+        ((error= setval(optp, value, argument, set_maximum_value,filename))) &&
           !option_is_loose)
 	DBUG_RETURN(error);
       if (get_one_option(optp, argument, filename))
@@ -711,7 +755,7 @@ static my_bool get_bool_argument(const struct my_option *opts,
 */
 
 static int setval(const struct my_option *opts, void *value, char *argument,
-		  my_bool set_maximum_value)
+                  my_bool set_maximum_value, const char *option_file)
 {
   int err= 0, res= 0;
   DBUG_ENTER("setval");
@@ -858,6 +902,7 @@ static int setval(const struct my_option *opts, void *value, char *argument,
       goto ret;
     };
   }
+  validate_value(opts->name, argument, option_file);
   DBUG_RETURN(0);
 
 ret:

mysys/my_init.c

@@ -34,6 +34,7 @@
 #endif
 static void my_win_init(void);
 static my_bool win32_init_tcp_ip();
+static void setup_codepages();
 #else
 #define my_win_init()
 #endif
@@ -67,6 +68,69 @@ static ulong atoi_octal(const char *str)
 MYSQL_FILE *mysql_stdin= NULL;
 static MYSQL_FILE instrumented_stdin;
 
+#ifdef _WIN32
+static UINT orig_console_cp, orig_console_output_cp;
+
+static void reset_console_cp(void)
+{
+  /*
+    We try not to call SetConsoleCP unnecessarily, to workaround a bug on
+    older Windows 10 (1803), which could switch truetype console fonts to
+    raster, eventhough SetConsoleCP would be a no-op (switch from UTF8 to UTF8).
+  */
+  if (GetConsoleCP() != orig_console_cp)
+    SetConsoleCP(orig_console_cp);
+  if (GetConsoleOutputCP() != orig_console_output_cp)
+    SetConsoleOutputCP(orig_console_output_cp);
+}
+
+/*
+  The below fixes discrepancies in console output and
+  command line parameter encoding. command line is in
+  ANSI codepage, output to console by default is in OEM, but
+  we like them to be in the same encoding.
+
+  We do this only if current codepage is UTF8, i.e when we
+  know we're on Windows that can handle UTF8 well.
+*/
+static void setup_codepages()
+{
+  UINT acp;
+  BOOL is_a_tty= fileno(stdout) >= 0 && isatty(fileno(stdout));
+
+  if (is_a_tty)
+  {
+    /*
+      Save console codepages, in case we change them,
+      to restore them on exit.
+    */
+    orig_console_cp= GetConsoleCP();
+    orig_console_output_cp= GetConsoleOutputCP();
+    if (orig_console_cp && orig_console_output_cp)
+      atexit(reset_console_cp);
+  }
+
+  if ((acp= GetACP()) != CP_UTF8)
+    return;
+
+  /*
+    Use setlocale to make mbstowcs/mkdir/getcwd behave, see
+    https://docs.microsoft.com/en-us/cpp/c-runtime-library/reference/setlocale-wsetlocale
+  */
+  setlocale(LC_ALL, "en_US.UTF8");
+
+  if (is_a_tty && (orig_console_cp != acp || orig_console_output_cp != acp))
+  {
+    /*
+      If ANSI codepage is UTF8, we actually want to switch console
+      to it as well.
+    */
+    SetConsoleCP(acp);
+    SetConsoleOutputCP(acp);
+  }
+}
+#endif
+
 /**
   Initialize my_sys functions, resources and variables
 
@@ -337,6 +401,17 @@ static void my_win_init(void)
 
   _tzset();
 
+  /*
+   We do not want text translation (LF->CRLF)
+   when stdout is console/terminal, it is buggy
+  */
+  if (fileno(stdout) >= 0 && isatty(fileno(stdout)))
+    (void)setmode(fileno(stdout), O_BINARY);
+
+  if (fileno(stderr) >= 0 && isatty(fileno(stderr)))
+    (void) setmode(fileno(stderr), O_BINARY);
+
+  setup_codepages();
   DBUG_VOID_RETURN;
 }