fix(smarty): drop mb_regex_encoding()/mb_split() from smarty_mb_str_replace()

docs/changelog.270.txt

@@ -10,6 +10,7 @@ Bug fixes:
 - xoops_aw_encode() and xoops_aw_decode() take their element by reference, so the array_walk() calls that encode and decode XOBJ_DTYPE_UNICODE_ARRAY values act on the array; write.php and XoopsObject::setVar() serialize the walked array instead of array_walk()'s boolean, and XoopsObject::getVar() returns the decoded array instead of true (include/functions.encoding.php, class/model/write.php, kernel/object.php)
 - a XOBJ_DTYPE_UNICODE_ARRAY value reloads as the array that was stored: assignVar() keeps the serialized string as is instead of url-decoding it, which broke the string lengths of elements such as café or a+b and made getVar() return [], and XoopsModelWrite::cleanVars() no longer addslashes() the elements before url-encoding them, which stored the backslashes (kernel/object.php, class/model/write.php, tests/unit/htdocs/kernel/XoopsObjectUnicodeArrayTest.php)
 - xoops_getUrlDomain() checks the host key only when parse_url() returned an array (include/functions.php)
+- the |replace and |escape modifiers no longer emit PHP 8.6 deprecations: Smarty's smarty_mb_str_replace() uses str_replace() instead of mb_regex_encoding()/mb_split(), which are deprecated since PHP 8.6 and scheduled for removal in PHP 9.0; non-UTF-8 charsets are still round-tripped through UTF-8 (xoops_lib/vendor/smarty/smarty/libs/plugins/shared.mb_str_replace.php)
 - templates can use |sprintf without a Smarty 4 deprecation notice: XoopsTpl registers sprintf as a modifier, as it already does for strstr, trim and file_exists (class/template.php) (#264)
 
 Libraries and assets:

htdocs/xoops_lib/vendor/smarty/smarty/libs/plugins/shared.mb_str_replace.php

@@ -44,42 +44,18 @@ function smarty_mb_str_replace($search, $replace, $subject, &$count = 0)
                 }
             }
         } else {
-            $mb_reg_charset = mb_regex_encoding();
-            // Check if mbstring regex is using UTF-8
-            $reg_is_unicode = !strcasecmp($mb_reg_charset, "UTF-8");
-            if(!$reg_is_unicode) {
-                // ...and set to UTF-8 if not
-                mb_regex_encoding("UTF-8");
+            // mb_regex_encoding()/mb_split() are deprecated since PHP 8.6 (bundled oniguruma is unmaintained).
+            // str_replace() is byte-exact and safe on UTF-8, which is self-synchronizing; other multibyte
+            // charsets (EUC-JP, GB2312, ...) are round-tripped through UTF-8 so no match spans a character boundary.
+            $convert = (bool)strcasecmp(Smarty::$_CHARSET, 'UTF-8');
+            if ($convert) {
+                $subject = mb_convert_encoding($subject ?? '', 'UTF-8', Smarty::$_CHARSET);
+                $search = mb_convert_encoding($search, 'UTF-8', Smarty::$_CHARSET);
+                $replace = mb_convert_encoding($replace, 'UTF-8', Smarty::$_CHARSET);
             }
-
-            // See if charset used by Smarty is matching one used by regex...
-            $current_charset = mb_regex_encoding();
-            $convert_result = (bool)strcasecmp(Smarty::$_CHARSET, $current_charset);
-            if($convert_result) {
-                // ...convert to it if not.
-                $subject = mb_convert_encoding($subject, $current_charset, Smarty::$_CHARSET);
-                $search = mb_convert_encoding($search, $current_charset, Smarty::$_CHARSET);
-                $replace = mb_convert_encoding($replace, $current_charset, Smarty::$_CHARSET);
-            }
-
-            $parts = mb_split(preg_quote($search), $subject ?? "") ?: array();
-            // If original regex encoding was not unicode...
-            if(!$reg_is_unicode) {
-                // ...restore original regex encoding to avoid breaking the system.
-                mb_regex_encoding($mb_reg_charset);
-            }
-            if($parts === false) {
-                // This exception is thrown if call to mb_split failed.
-                // Usually it happens, when $search or $replace are not valid for given mb_regex_encoding().
-                // There may be other cases for it to fail, please file an issue if you find a reproducible one.
-                throw new SmartyException("Source string is not a valid $current_charset sequence (probably)");
-            }
-
-            $count = count($parts) - 1;
-            $subject = implode($replace, $parts);
-            // Convert results back to charset used by Smarty, if needed.
-            if($convert_result) {
-                $subject = mb_convert_encoding($subject, Smarty::$_CHARSET, $current_charset);
+            $subject = str_replace($search, $replace, $subject ?? '', $count);
+            if ($convert) {
+                $subject = mb_convert_encoding($subject, Smarty::$_CHARSET, 'UTF-8');
             }
         }
         return $subject;