Exclude RtoL characters from paired string delimiters

Fixes #22228

Some scripts in the world are written right-to-left, such as Arabic and
Hebrew.  This can result in confusion for quote-like string delimitters
that we have chosen based on left-to_right.  Therefore exclude all such.
Currently, the only pair that falls into this category that we don't
exclude for other reasons are SYRIAC COLON SKEWED LEFT/RIGHT.

pod/perlop.pod

@@ -3850,7 +3850,6 @@ The complete list of accepted paired delimiters as of Unicode 14.0 is:
  {  }    U+007B, U+007D   LEFT/RIGHT CURLY BRACKET
  «  »    U+00AB, U+00BB   LEFT/RIGHT-POINTING DOUBLE ANGLE QUOTATION MARK
  »  «    U+00BB, U+00AB   RIGHT/LEFT-POINTING DOUBLE ANGLE QUOTATION MARK
- ܆  ܇    U+0706, U+0707   SYRIAC COLON SKEWED LEFT/RIGHT
  ༺  ༻    U+0F3A, U+0F3B   TIBETAN MARK GUG RTAGS GYON,  TIBETAN MARK GUG
                           RTAGS GYAS
  ༼  ༽    U+0F3C, U+0F3D   TIBETAN MARK ANG KHANG GYON,  TIBETAN MARK ANG
@@ -4231,5 +4230,4 @@ The complete list of accepted paired delimiters as of Unicode 14.0 is:
  🢩  🢨    U+1F8A9, U+1F8A8 RIGHT/LEFTWARDS BACK-TILTED SHADOWED WHITE ARROW
  🢫  🢪    U+1F8AB, U+1F8AA RIGHT/LEFTWARDS FRONT-TILTED SHADOWED WHITE
                           ARROW
-
 =cut

regen/unicode_constants.pl

@@ -378,6 +378,7 @@ my $unpaired = "Didn't find a mirror";
 my $illegal = "Mirror illegal";
 my $no_encoded_mate = "Mirrored, but Unicode has no encoded mirror";
 my $bidirectional = "Bidirectional";
+my $r2l = "Is in a Right to Left script";
 
 my %unused_bidi_pairs;
 my %inverted_unused_bidi_pairs;
@@ -634,6 +635,15 @@ foreach my $list (qw(Punctuation Symbol)) {
             next;
         }
 
+        # Exclude characters that are R to L ordering, as this can cause
+        # confusion.  See GH #22228
+        if ($chr =~ / (?[ \p{Bidi_Class:R} + \p{Bidi_Class:AL} ]) /x) {
+            $discards{$code_point} = { reason => $r2l,
+                                       mirror => $mirror_code_point
+                                     };
+            next;
+        }
+
         # We enter the pair with the original code point on the left; if it
         # should instead be on the R, swap.  Most Symbols that contain the
         # word REVERSE go on the rhs, except those whose names explicitly

t/lib/croak/toke

@@ -133,6 +133,24 @@ EXPECT
 Use of '«' is deprecated as a string delimiter at - line 3.
 Can't find string terminator "«" anywhere before EOF at - line 5.
 ########
+# NAME mirrored delimiters in R-to-L scripts are invalid
+BEGIN { binmode STDERR, ":utf8" }
+use utf8;
+use feature 'extra_paired_delimiters';
+my $good = q܈this string is delimitted by a symbol in a R-to-L script܈;
+$good = q܇this string is delimitted by a symbol in a R-to-L script܇;
+my $bad = q܈Can't use mirrored R-to-L script delimiters܇;
+EXPECT
+Can't find string terminator "܈" anywhere before EOF at - line 6.
+########
+# NAME mirrored delimiters in R-to-L scripts are invalid in the other order too
+BEGIN { binmode STDERR, ":utf8" }
+use utf8;
+use feature 'extra_paired_delimiters';
+my $bad = q܇Can't use mirrored R-to-L script delimiters܈;
+EXPECT
+Can't find string terminator "܇" anywhere before EOF at - line 4.
+########
 # NAME paired above Latin1 delimiters need feature enabled
 BEGIN { binmode STDERR, ":utf8" }
 use utf8;