Tweak the filename regexp in parser.t so that ./TEST -utf16 op/parser.t passes.

[p5sagit/p5-mst-13.2.git] / t / comp / utf.t

diff --git a/t/comp/utf.t b/t/comp/utf.t
index 69ede95..d59ba2d 100644 (file)
--- a/t/comp/utf.t
+++ b/t/comp/utf.t
@@ -1,6 +1,6 @@
 #!./perl -w
 
-print "1..76\n";
+print "1..3980\n";
 my $test = 0;
 
 my %templates = (
@@ -13,7 +13,23 @@ sub bytes_to_utf {
     my ($enc, $content, $do_bom) = @_;
     my $template = $templates{$enc};
     die "Unsupported encoding $enc" unless $template;
-    return pack "$template*", ($do_bom ? 0xFEFF : ()), unpack "U*", $content;
+    my @chars = unpack "U*", $content;
+    if ($enc ne 'utf8') {
+       # Make surrogate pairs
+       my @remember_that_utf_16_is_variable_length;
+       foreach my $ord (@chars) {
+           if ($ord < 0x10000) {
+               push @remember_that_utf_16_is_variable_length,
+                   $ord;
+           } else {
+               $ord -= 0x10000;
+               push @remember_that_utf_16_is_variable_length,
+                   (0xD800 | ($ord >> 10)), (0xDC00 | ($ord & 0x3FF));
+           }
+       }
+       @chars = @remember_that_utf_16_is_variable_length;
+    }
+    return pack "$template*", ($do_bom ? 0xFEFF : ()), @chars;
 }
 
 sub test {
@@ -45,13 +61,31 @@ for my $bom (0, 1) {
            # right now, as here we're testing the input filter itself.
 
            for my $expect ("N", "\xFF", "\x{100}", "\x{010a}", "\x{0a23}",
+                           "\x{10000}", "\x{64321}", "\x{10FFFD}",
+                           "\x{1000a}", # 0xD800 0xDC0A
+                           "\x{12800}", # 0xD80A 0xDC00
                           ) {
-               # A space so that the UTF-16 heuristc triggers - " '" gives two
+               # A space so that the UTF-16 heuristic triggers - " '" gives two
                # characters of ASCII.
                my $write = " '$expect'";
                my $name = 'chrs ' . join ', ', map {ord $_} split '', $expect;
                test($enc, $write, $expect, $bom, $nl, $name);
            }
+
+           # This is designed to try to trip over the end of the buffer,
+           # with similar results to U-1000A and U-12800 above.
+           for my $pad (2 .. 162) {
+               for my $chr ("\x{10000}", "\x{1000a}", "\x{12800}") {
+                   my $padding = ' ' x $pad;
+                   # Need 4 octets that were from 2 ASCII characters to trigger
+                   # the heuristic that detects UTF-16 without a BOM. For
+                   # UTF-16BE, one space and the newline will do, as the
+                   # newline's high octet comes first. But for UTF-16LE, a
+                   # newline is "\n\0", so it doesn't trigger it.
+                   test($enc, "  \n$padding'$chr'", $chr, $bom, $nl,
+                        sprintf "'\\x{%x}' with $pad spaces before it", ord $chr);
+               }
+           }
        }
     }
 }