sub: allow specifying a fallback codepage if input is not UTF-8

Normally, --subcp always forces conversion. This really always forces
conversion, even if the UTF-8 check on the input succeeds.

Extend the --subcp to allow codepages as fallback if UTF-8 doesn't
work. So, for example --subcp=utf8:cp1250 will use UTF-8 if the input
looks like UTF-8, and will fall back to use cp1250 if the UTF-8 check
fails.

I think this should actually be the default, but on the other hand,
this changes the semantics of the option, and a user would actually
expect --subcp to force conversion, rather than silently using UTF-8
if that happens to work.

DOCS/man/en/options.rst

@@ -2035,10 +2035,14 @@
     If your system supports ``iconv(3)``, you can use this option to specify
     the subtitle codepage.
 
+    Warning: if you force the charset, even subtitles that are known to be
+    UTF-8 will be recoded, which is perhaps not what you expect.
+
     .. admonition:: Examples
 
-        - ``--subcp=latin2``
+        - ``--subcp=utf8:latin2`` Use Latin 2 if input is not UTF-8.
-        - ``--subcp=cp1250``
+        - ``--subcp=utf8:cp1250`` Use CP1250 if input is not UTF-8.
+        - ``--subcp=cp1250`` Always force recoding to cp1250.
 
     If the player was compiled with ENCA support, you can use special syntax
     to use that::
@@ -2049,6 +2053,8 @@
     ENCA detect the codepage automatically. If unsure, enter anything (if the
     language is invalid, mpv will complain and list valid languages).
     Fallback codepage specifies the codepage to use if autodetection fails.
+    If no fallback is specified, the subtitle will be interpreted as UTF-8,
+    but with "Latin 1" as fallback for bytes that are not valid UTF-8 sequences.
 
     .. admonition:: Examples
 

mpvcore/charset_conv.c

@@ -70,9 +70,13 @@ static int split_colon(const char *user_cp, int max, bstr *out_arr)
 bool mp_charset_requires_guess(const char *user_cp)
 {
     bstr res[2] = {{0}};
-    split_colon(user_cp, 2, res);
+    int r = split_colon(user_cp, 2, res);
+    // Note that "utf8" is the UTF-8 codepage, while "utf8:..." specifies UTF-8
+    // by default, plus a codepage that is used if the input is not UTF-8.
     return bstrcasecmp0(res[0], "enca") == 0 ||
-           bstrcasecmp0(res[0], "guess") == 0;
+           bstrcasecmp0(res[0], "guess") == 0 ||
+           (r > 1 && bstrcasecmp0(res[0], "utf-8") == 0) ||
+           (r > 1 && bstrcasecmp0(res[0], "utf8") == 0);
 }
 
 #ifdef CONFIG_ENCA
@@ -155,6 +159,10 @@ const char *mp_charset_guess(bstr buf, const char *user_cp, int flags)
     if (bstrcasecmp0(type, "guess") == 0)
         res = libguess_guess(buf, lang);
 #endif
+    if (bstrcasecmp0(type, "utf8") == 0 || bstrcasecmp0(type, "utf-8") == 0) {
+        if (!fallback)
+            fallback = params[1].start; // must be already 0-terminated
+    }
 
     if (res) {
         mp_msg(MSGT_SUBREADER, MSGL_DBG2, "%.*s detected charset: '%s'\n",