[perl5.git] / ext / XS-APItest / t / utf8_warn_base.pl

#!perl -w

# This is a base file to be used by various .t's in its directory

use strict;
use Test::More;

BEGIN {
    use_ok('XS::APItest');
    require 'charset_tools.pl';
    require './t/utf8_setup.pl';
};

$|=1;

no warnings 'deprecated'; # Some of the below are above IV_MAX on 32 bit
                          # machines, and that is tested elsewhere

use XS::APItest;
use Data::Dumper;

my @warnings_gotten;

use warnings 'utf8';
local $SIG{__WARN__} = sub { my @copy = @_;
                             push @warnings_gotten, map { chomp; $_ } @copy;
                           };


# Now test the cases where a legal code point is generated, but may or may not
# be allowed/warned on.
my @tests = (
     # ($testname, $bytes, $controlling_warning_category,
     #  $allowed_uv, $needed_to_discern_len )
    [ "lowest surrogate",
        (isASCII) ? "\xed\xa0\x80" : I8_to_native("\xf1\xb6\xa0\xa0"),
        'surrogate', 0xD800,
    ],
    [ "a middle surrogate",
        (isASCII) ? "\xed\xa4\x8d" : I8_to_native("\xf1\xb6\xa8\xad"),
        'surrogate', 0xD90D,
    ],
    [ "highest surrogate",
        (isASCII) ? "\xed\xbf\xbf" : I8_to_native("\xf1\xb7\xbf\xbf"),
        'surrogate', 0xDFFF,
    ],
    [ "first non_unicode",
        (isASCII) ? "\xf4\x90\x80\x80" : I8_to_native("\xf9\xa2\xa0\xa0\xa0"),
        'non_unicode', 0x110000,
        2,
    ],
    [ "non_unicode whose first byte tells that",
        (isASCII) ? "\xf5\x80\x80\x80" : I8_to_native("\xfa\xa0\xa0\xa0\xa0"),
        'non_unicode',
        (isASCII) ? 0x140000 : 0x200000,
        1,
    ],
    [ "first of 32 consecutive non-character code points",
        (isASCII) ? "\xef\xb7\x90" : I8_to_native("\xf1\xbf\xae\xb0"),
        'nonchar', 0xFDD0,
    ],
    [ "a mid non-character code point of the 32 consecutive ones",
        (isASCII) ? "\xef\xb7\xa0" : I8_to_native("\xf1\xbf\xaf\xa0"),
        'nonchar', 0xFDE0,
    ],
    [ "final of 32 consecutive non-character code points",
        (isASCII) ? "\xef\xb7\xaf" : I8_to_native("\xf1\xbf\xaf\xaf"),
        'nonchar', 0xFDEF,
    ],
    [ "non-character code point U+FFFE",
        (isASCII) ? "\xef\xbf\xbe" : I8_to_native("\xf1\xbf\xbf\xbe"),
        'nonchar', 0xFFFE,
    ],
    [ "non-character code point U+FFFF",
        (isASCII) ? "\xef\xbf\xbf" : I8_to_native("\xf1\xbf\xbf\xbf"),
        'nonchar', 0xFFFF,
    ],
    [ "non-character code point U+1FFFE",
        (isASCII) ? "\xf0\x9f\xbf\xbe" : I8_to_native("\xf3\xbf\xbf\xbe"),
        'nonchar', 0x1FFFE,
    ],
    [ "non-character code point U+1FFFF",
        (isASCII) ? "\xf0\x9f\xbf\xbf" : I8_to_native("\xf3\xbf\xbf\xbf"),
        'nonchar', 0x1FFFF,
    ],
    [ "non-character code point U+2FFFE",
        (isASCII) ? "\xf0\xaf\xbf\xbe" : I8_to_native("\xf5\xbf\xbf\xbe"),
        'nonchar', 0x2FFFE,
    ],
    [ "non-character code point U+2FFFF",
        (isASCII) ? "\xf0\xaf\xbf\xbf" : I8_to_native("\xf5\xbf\xbf\xbf"),
        'nonchar', 0x2FFFF,
    ],
    [ "non-character code point U+3FFFE",
        (isASCII) ? "\xf0\xbf\xbf\xbe" : I8_to_native("\xf7\xbf\xbf\xbe"),
        'nonchar', 0x3FFFE,
    ],
    [ "non-character code point U+3FFFF",
        (isASCII) ? "\xf0\xbf\xbf\xbf" : I8_to_native("\xf7\xbf\xbf\xbf"),
        'nonchar', 0x3FFFF,
    ],
    [ "non-character code point U+4FFFE",
        (isASCII) ? "\xf1\x8f\xbf\xbe" : I8_to_native("\xf8\xa9\xbf\xbf\xbe"),
        'nonchar', 0x4FFFE,
    ],
    [ "non-character code point U+4FFFF",
        (isASCII) ? "\xf1\x8f\xbf\xbf" : I8_to_native("\xf8\xa9\xbf\xbf\xbf"),
        'nonchar', 0x4FFFF,
    ],
    [ "non-character code point U+5FFFE",
        (isASCII) ? "\xf1\x9f\xbf\xbe" : I8_to_native("\xf8\xab\xbf\xbf\xbe"),
        'nonchar', 0x5FFFE,
    ],
    [ "non-character code point U+5FFFF",
        (isASCII) ? "\xf1\x9f\xbf\xbf" : I8_to_native("\xf8\xab\xbf\xbf\xbf"),
        'nonchar', 0x5FFFF,
    ],
    [ "non-character code point U+6FFFE",
        (isASCII) ? "\xf1\xaf\xbf\xbe" : I8_to_native("\xf8\xad\xbf\xbf\xbe"),
        'nonchar', 0x6FFFE,
    ],
    [ "non-character code point U+6FFFF",
        (isASCII) ? "\xf1\xaf\xbf\xbf" : I8_to_native("\xf8\xad\xbf\xbf\xbf"),
        'nonchar', 0x6FFFF,
    ],
    [ "non-character code point U+7FFFE",
        (isASCII) ? "\xf1\xbf\xbf\xbe" : I8_to_native("\xf8\xaf\xbf\xbf\xbe"),
        'nonchar', 0x7FFFE,
    ],
    [ "non-character code point U+7FFFF",
        (isASCII) ? "\xf1\xbf\xbf\xbf" : I8_to_native("\xf8\xaf\xbf\xbf\xbf"),
        'nonchar', 0x7FFFF,
    ],
    [ "non-character code point U+8FFFE",
        (isASCII) ? "\xf2\x8f\xbf\xbe" : I8_to_native("\xf8\xb1\xbf\xbf\xbe"),
        'nonchar', 0x8FFFE,
    ],
    [ "non-character code point U+8FFFF",
        (isASCII) ? "\xf2\x8f\xbf\xbf" : I8_to_native("\xf8\xb1\xbf\xbf\xbf"),
        'nonchar', 0x8FFFF,
    ],
    [ "non-character code point U+9FFFE",
        (isASCII) ? "\xf2\x9f\xbf\xbe" : I8_to_native("\xf8\xb3\xbf\xbf\xbe"),
        'nonchar', 0x9FFFE,
    ],
    [ "non-character code point U+9FFFF",
        (isASCII) ? "\xf2\x9f\xbf\xbf" : I8_to_native("\xf8\xb3\xbf\xbf\xbf"),
        'nonchar', 0x9FFFF,
    ],
    [ "non-character code point U+AFFFE",
        (isASCII) ? "\xf2\xaf\xbf\xbe" : I8_to_native("\xf8\xb5\xbf\xbf\xbe"),
        'nonchar', 0xAFFFE,
    ],
    [ "non-character code point U+AFFFF",
        (isASCII) ? "\xf2\xaf\xbf\xbf" : I8_to_native("\xf8\xb5\xbf\xbf\xbf"),
        'nonchar', 0xAFFFF,
    ],
    [ "non-character code point U+BFFFE",
        (isASCII) ? "\xf2\xbf\xbf\xbe" : I8_to_native("\xf8\xb7\xbf\xbf\xbe"),
        'nonchar', 0xBFFFE,
    ],
    [ "non-character code point U+BFFFF",
        (isASCII) ? "\xf2\xbf\xbf\xbf" : I8_to_native("\xf8\xb7\xbf\xbf\xbf"),
        'nonchar', 0xBFFFF,
    ],
    [ "non-character code point U+CFFFE",
        (isASCII) ? "\xf3\x8f\xbf\xbe" : I8_to_native("\xf8\xb9\xbf\xbf\xbe"),
        'nonchar', 0xCFFFE,
    ],
    [ "non-character code point U+CFFFF",
        (isASCII) ? "\xf3\x8f\xbf\xbf" : I8_to_native("\xf8\xb9\xbf\xbf\xbf"),
        'nonchar', 0xCFFFF,
    ],
    [ "non-character code point U+DFFFE",
        (isASCII) ? "\xf3\x9f\xbf\xbe" : I8_to_native("\xf8\xbb\xbf\xbf\xbe"),
        'nonchar', 0xDFFFE,
    ],
    [ "non-character code point U+DFFFF",
        (isASCII) ? "\xf3\x9f\xbf\xbf" : I8_to_native("\xf8\xbb\xbf\xbf\xbf"),
        'nonchar', 0xDFFFF,
    ],
    [ "non-character code point U+EFFFE",
        (isASCII) ? "\xf3\xaf\xbf\xbe" : I8_to_native("\xf8\xbd\xbf\xbf\xbe"),
        'nonchar', 0xEFFFE,
    ],
    [ "non-character code point U+EFFFF",
        (isASCII) ? "\xf3\xaf\xbf\xbf" : I8_to_native("\xf8\xbd\xbf\xbf\xbf"),
        'nonchar', 0xEFFFF,
    ],
    [ "non-character code point U+FFFFE",
        (isASCII) ? "\xf3\xbf\xbf\xbe" : I8_to_native("\xf8\xbf\xbf\xbf\xbe"),
        'nonchar', 0xFFFFE,
    ],
    [ "non-character code point U+FFFFF",
        (isASCII) ? "\xf3\xbf\xbf\xbf" : I8_to_native("\xf8\xbf\xbf\xbf\xbf"),
        'nonchar', 0xFFFFF,
    ],
    [ "non-character code point U+10FFFE",
        (isASCII) ? "\xf4\x8f\xbf\xbe" : I8_to_native("\xf9\xa1\xbf\xbf\xbe"),
        'nonchar', 0x10FFFE,
    ],
    [ "non-character code point U+10FFFF",
        (isASCII) ? "\xf4\x8f\xbf\xbf" : I8_to_native("\xf9\xa1\xbf\xbf\xbf"),
        'nonchar', 0x10FFFF,
    ],
    [ "requires at least 32 bits",
        (isASCII)
         ?  "\xfe\x82\x80\x80\x80\x80\x80"
         : I8_to_native(
            "\xff\xa0\xa0\xa0\xa0\xa0\xa0\xa2\xa0\xa0\xa0\xa0\xa0\xa0"),
        # This code point is chosen so that it is representable in a UV on
        # 32-bit machines
        'non_unicode', 0x80000000,
        (isASCII) ? 1 : 8,
    ],
    [ "highest 32 bit code point",
        (isASCII)
         ?  "\xfe\x83\xbf\xbf\xbf\xbf\xbf"
         : I8_to_native(
            "\xff\xa0\xa0\xa0\xa0\xa0\xa0\xa3\xbf\xbf\xbf\xbf\xbf\xbf"),
        'non_unicode', 0xFFFFFFFF,
        (isASCII) ? 1 : 8,
    ],
    [ "requires at least 32 bits, and use SUPER-type flags, instead of"
    . " ABOVE_31_BIT",
        (isASCII)
         ? "\xfe\x82\x80\x80\x80\x80\x80"
         : I8_to_native(
           "\xff\xa0\xa0\xa0\xa0\xa0\xa0\xa2\xa0\xa0\xa0\xa0\xa0\xa0"),
        'non_unicode', 0x80000000,
        1,
    ],
    [ "overflow with warnings/disallow for more than 31 bits",
        # This tests the interaction of WARN_ABOVE_31_BIT/DISALLOW_ABOVE_31_BIT
        # with overflow.  The overflow malformation is never allowed, so
        # preventing it takes precedence if the ABOVE_31_BIT options would
        # otherwise allow in an overflowing value.  The ASCII code points (1
        # for 32-bits; 1 for 64) were chosen because the old overflow
        # detection algorithm did not catch them; this means this test also
        # checks for that fix.  The EBCDIC are arbitrary overflowing ones
        # since we have no reports of failures with it.
       (($::is64bit)
        ? ((isASCII)
           ?    "\xff\x80\x90\x90\x90\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf"
           : I8_to_native(
                "\xff\xB0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"))
        : ((isASCII)
           ?    "\xfe\x86\x80\x80\x80\x80\x80"
           : I8_to_native(
                "\xff\xa0\xa0\xa0\xa0\xa0\xa0\xa4\xa0\xa0\xa0\xa0\xa0\xa0"))),
        'non_unicode', -1,
        (isASCII || $::is64bit) ? 2 : 8,
    ],
);

if (! $::is64bit) {
    if (isASCII) {
        no warnings qw{portable overflow};
        push @tests,
            [ "Lowest 33 bit code point: overflow",
                "\xFE\x84\x80\x80\x80\x80\x80",
                'non_unicode', -1,
                1,
            ];
    }
}
else {
    no warnings qw{portable overflow};
    push @tests,
        [ "More than 32 bits",
            (isASCII)
            ?       "\xff\x80\x80\x80\x80\x80\x81\x80\x80\x80\x80\x80\x80"
            : I8_to_native(
                    "\xff\xa0\xa0\xa0\xa0\xa0\xa2\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
            'non_unicode', 0x1000000000,
            (isASCII) ? 1 : 7,
        ];
    if (! isASCII) {
        push @tests,   # These could falsely show wrongly in a naive
                       # implementation
            [ "requires at least 32 bits",
                I8_to_native(
                    "\xff\xa0\xa0\xa0\xa0\xa0\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
                'non_unicode', 0x800000000,
                7,
            ],
            [ "requires at least 32 bits",
                I8_to_native(
                    "\xff\xa0\xa0\xa0\xa0\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
                'non_unicode', 0x10000000000,
                6,
            ],
            [ "requires at least 32 bits",
                I8_to_native(
                    "\xff\xa0\xa0\xa0\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
                'non_unicode', 0x200000000000,
                5,
            ],
            [ "requires at least 32 bits",
                I8_to_native(
                    "\xff\xa0\xa0\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
                'non_unicode', 0x4000000000000,
                4,
            ],
            [ "requires at least 32 bits",
                I8_to_native(
                    "\xff\xa0\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
                'non_unicode', 0x80000000000000,
                3,
            ],
            [ "requires at least 32 bits",
                I8_to_native(
                    "\xff\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
                'non_unicode', 0x1000000000000000,
                2,
            ];
    }
}

sub flags_to_text($$)
{
    my ($flags, $flags_to_text_ref) = @_;

    # Returns a string containing a mnemonic representation of the bits that
    # are set in the $flags.  These are assumed to be flag bits.  The return
    # looks like "FOO|BAR|BAZ".  The second parameter is a reference to an
    # array that gives the textual representation of all the possible flags.
    # Element 0 is the text for the bit 0 flag; element 1 for bit 1; ....  If
    # no bits at all are set the string "0" is returned;

    my @flag_text;
    my $shift = 0;

    return "0" if $flags == 0;

    while ($flags) {
        #diag sprintf "%x", $flags;
        if ($flags & 1) {
            push @flag_text, $flags_to_text_ref->[$shift];
        }
        $shift++;
        $flags >>= 1;
    }

    return join "|", @flag_text;
}

# Possible flag returns from utf8n_to_uvchr_error().  These should have G_,
# instead of A_, D_, but the prefixes will be used in a a later commit, so
# minimize churn by having them here.
my @utf8n_flags_to_text =  ( qw(
        A_EMPTY
        A_CONTINUATION
        A_NON_CONTINUATION
        A_SHORT
        A_LONG
        A_LONG_AND_ITS_VALUE
        PLACEHOLDER
        A_OVERFLOW
        D_SURROGATE
        W_SURROGATE
        D_NONCHAR
        W_NONCHAR
        D_SUPER
        W_SUPER
        D_ABOVE_31_BIT
        W_ABOVE_31_BIT
        CHECK_ONLY
        NO_CONFIDENCE_IN_CURLEN_
    ) );

sub utf8n_display_call($)
{
    # Converts an eval string that calls test_utf8n_to_uvchr into a more human
    # readable form, and returns it.  Doesn't work if the byte string contains
    # an apostrophe.  The return will look something like:
    #   test_utf8n_to_uvchr_error('$bytes', $length, $flags)
    #diag $_[0];

    $_[0] =~ / ^ ( [^(]* \( ) ' ( [^']*? ) ' ( .+ , \D* ) ( \d+ ) \) $ /x;
    my $text1 = $1;     # Everything before the byte string
    my $bytes = $2;
    my $text2 = $3;     # Includes the length
    my $flags = $4;

    return $text1
         . display_bytes($bytes)
         . $text2
         . flags_to_text($flags, \@utf8n_flags_to_text)
         . ')';
}

sub uvchr_display_call($)
{
    # Converts an eval string that calls test_uvchr_to_utf8 into a more human
    # readable form, and returns it.  The return will look something like:
    #   test_uvchr_to_utf8n_flags($uv, $flags)
    #diag $_[0];

    my @flags_to_text =  ( qw(
            W_SURROGATE
            W_NONCHAR
            W_SUPER
            W_ABOVE_31_BIT
            D_SURROGATE
            D_NONCHAR
            D_SUPER
            D_ABOVE_31_BIT
       ) );

    $_[0] =~ / ^ ( [^(]* \( ) ( \d+ ) , \s* ( \d+ ) \) $ /x;
    my $text = $1;
    my $cp = sprintf "%X", $2;
    my $flags = $3;

    return "${text}0x$cp, " . flags_to_text($flags, \@flags_to_text) . ')';
}

sub do_warnings_test(@)
{
    my @expected_warnings = @_;

    # Compares the input expected warnings array with @warnings_gotten,
    # generating a pass for each found, removing it from @warnings_gotten.
    # Any discrepancies generate test failures.  Returns TRUE if no
    # discrepcancies; otherwise FALSE.

    my $succeeded = 1;

    if (@expected_warnings == 0) {
        if (! is(@warnings_gotten, 0, "    Expected and got no warnings")) {
            output_warnings(@warnings_gotten);
            $succeeded = 0;
        }
        return $succeeded;
    }

    # Check that we got all the expected warnings,
    # removing each one found
  WARNING:
    foreach my $expected (@expected_warnings) {
        foreach (my $i = 0; $i < @warnings_gotten; $i++) {
            if ($warnings_gotten[$i] =~ $expected) {
                pass("    Expected and got warning: "
                    . " $warnings_gotten[$i]");
                splice @warnings_gotten, $i, 1;
                next WARNING;
            }
        }
        fail("    Expected a warning that matches "
            . $expected . " but didn't get it");
        $succeeded = 0;
    }

    if (! is(@warnings_gotten, 0, "    Got no unexpected warnings")) {
        output_warnings(@warnings_gotten);
        $succeeded = 0;
    }

    return $succeeded;
}

# This test is split into this number of files.
my $num_test_files = $ENV{TEST_JOBS} || 1;
$num_test_files = 10 if $num_test_files > 10;

my $test_count = -1;
foreach my $test (@tests) {
    $test_count++;
    next if $test_count % $num_test_files != $::TEST_CHUNK;

    my ($testname, $bytes,
        $controlling_warning_category, $allowed_uv, $needed_to_discern_len
       ) = @$test;

    my $length = length $bytes;
    my $will_overflow = $allowed_uv < 0;

    my $uv_string = sprintf(($allowed_uv < 0x100) ? "%02X" : "%04X", $allowed_uv);

    my $utf8n_flag_to_warn;
    my $utf8n_flag_to_disallow;
    my $uvchr_flag_to_warn;
    my $uvchr_flag_to_disallow;

    # Many of the code points being tested are middling in that if code point
    # edge cases work, these are very likely to as well.  Because this test
    # file takes a while to execute, we skip testing the edge effects of code
    # points deemed middling, while testing their basics and continuing to
    # fully test the non-middling code points.
    my $skip_most_tests = 0;

    my $cp_message_qr;      # Pattern that matches the message raised when
                            # that message contains the problematic code
                            # point.  The message is the same (currently) both
                            # when going from/to utf8.
    my $non_cp_trailing_text;   # The suffix text when the message doesn't
                                # contain a code point.  (This is a result of
                                # some sort of malformation that means we
                                # can't get an exact code poin

    if ($will_overflow || $allowed_uv > 0x10FFFF) {

        $utf8n_flag_to_warn     = $::UTF8_WARN_SUPER;
        $utf8n_flag_to_disallow = $::UTF8_DISALLOW_SUPER;
        $uvchr_flag_to_warn     = $::UNICODE_WARN_SUPER;
        $uvchr_flag_to_disallow = $::UNICODE_DISALLOW_SUPER;;

        if ($will_overflow) {
            $non_cp_trailing_text = "if you see this, there is an error";
            $cp_message_qr = qr/\Q$non_cp_trailing_text\E/;
        }
        elsif ($allowed_uv > 0x7FFFFFFF) {
            $cp_message_qr = qr/\QCode point 0x$uv_string is not Unicode,\E
                                \Q and not portable\E/x;
            $non_cp_trailing_text = "is for a non-Unicode code point, and is not portable";
        }
        else {
            $cp_message_qr = qr/\QCode point 0x$uv_string is not Unicode,\E
                                \Q may not be portable\E/x;
            $non_cp_trailing_text = "is for a non-Unicode code point, may not"
                                . " be portable";
        }
    }
    elsif ($allowed_uv >= 0xD800 && $allowed_uv <= 0xDFFF) {
        $cp_message_qr = qr/UTF-16 surrogate U\+$uv_string/;
        $non_cp_trailing_text = "is for a surrogate";
        $needed_to_discern_len = 2 unless defined $needed_to_discern_len;
        $skip_most_tests = 1 if $allowed_uv > 0xD800 && $allowed_uv < 0xDFFF;

        $utf8n_flag_to_warn     = $::UTF8_WARN_SURROGATE;
        $utf8n_flag_to_disallow = $::UTF8_DISALLOW_SURROGATE;
        $uvchr_flag_to_warn     = $::UNICODE_WARN_SURROGATE;
        $uvchr_flag_to_disallow = $::UNICODE_DISALLOW_SURROGATE;;
    }
    elsif (   ($allowed_uv >= 0xFDD0 && $allowed_uv <= 0xFDEF)
           || ($allowed_uv & 0xFFFE) == 0xFFFE)
    {
        $cp_message_qr = qr/\QUnicode non-character U+$uv_string\E
                            \Q is not recommended for open interchange\E/x;
        $non_cp_trailing_text = "if you see this, there is an error";
        $needed_to_discern_len = $length unless defined $needed_to_discern_len;
        if (   ($allowed_uv > 0xFDD0 && $allowed_uv < 0xFDEF)
            || ($allowed_uv > 0xFFFF && $allowed_uv < 0x10FFFE))
        {
            $skip_most_tests = 1;
        }

        $utf8n_flag_to_warn     = $::UTF8_WARN_NONCHAR;
        $utf8n_flag_to_disallow = $::UTF8_DISALLOW_NONCHAR;
        $uvchr_flag_to_warn     = $::UNICODE_WARN_NONCHAR;
        $uvchr_flag_to_disallow = $::UNICODE_DISALLOW_NONCHAR;;
    }
    else {
        die "Can't figure out what type of warning to test for $testname"
    }

    die 'Didn\'t set $needed_to_discern_len for ' . $testname
                                        unless defined $needed_to_discern_len;
    {
        use warnings;
        undef @warnings_gotten;
        my $ret = test_isUTF8_CHAR($bytes, $length);
        my $ret_flags = test_isUTF8_CHAR_flags($bytes, $length, 0);
        if ($will_overflow) {
            is($ret, 0, "For $testname: isUTF8_CHAR() returns 0");
            is($ret_flags, 0, "    And isUTF8_CHAR_flags() returns 0");
        }
        else {
            is($ret, $length,
               "For $testname: isUTF8_CHAR() returns expected length: $length");
            is($ret_flags, $length, "    And isUTF8_CHAR_flags(...,0)"
                                  . " returns expected length: $length");
        }
        is(scalar @warnings_gotten, 0,
                "    And neither isUTF8_CHAR() nor isUTF8_CHAR()_flags generated"
              . " any warnings")
          or output_warnings(@warnings_gotten);

        undef @warnings_gotten;
        $ret = test_isSTRICT_UTF8_CHAR($bytes, $length);
        if ($will_overflow) {
            is($ret, 0, "    And isSTRICT_UTF8_CHAR() returns 0");
        }
        else {
            my $expected_ret = (   $testname =~ /surrogate|non-character/
                                || $allowed_uv > 0x10FFFF)
                               ? 0
                               : $length;
            is($ret, $expected_ret, "    And isSTRICT_UTF8_CHAR() returns"
                                  . " expected length: $expected_ret");
            $ret = test_isUTF8_CHAR_flags($bytes, $length,
                                          $::UTF8_DISALLOW_ILLEGAL_INTERCHANGE);
            is($ret, $expected_ret,
                    "    And isUTF8_CHAR_flags('DISALLOW_ILLEGAL_INTERCHANGE')"
                    . " acts like isSTRICT_UTF8_CHAR");
        }
        is(scalar @warnings_gotten, 0,
                "    And neither isSTRICT_UTF8_CHAR() nor isUTF8_CHAR_flags"
              . " generated any warnings")
          or output_warnings(@warnings_gotten);

        undef @warnings_gotten;
        $ret = test_isC9_STRICT_UTF8_CHAR($bytes, $length);
        if ($will_overflow) {
            is($ret, 0, "    And isC9_STRICT_UTF8_CHAR() returns 0");
        }
        else {
            my $expected_ret = (   $testname =~ /surrogate/
                                || $allowed_uv > 0x10FFFF)
                               ? 0
                               : $length;
            is($ret, $expected_ret, "    And isC9_STRICT_UTF8_CHAR()"
                                   ." returns expected length: $expected_ret");
            $ret = test_isUTF8_CHAR_flags($bytes, $length,
                                          $::UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE);
            is($ret, $expected_ret,
                  "    And isUTF8_CHAR_flags('DISALLOW_ILLEGAL_C9_INTERCHANGE')"
                . " acts like isC9_STRICT_UTF8_CHAR");
        }
        is(scalar @warnings_gotten, 0,
                "    And neither isC9_STRICT_UTF8_CHAR() nor isUTF8_CHAR_flags"
              . " generated any warnings")
          or output_warnings(@warnings_gotten);

        # Test partial character handling, for each byte not a full character
        for my $j (1.. $length - 1) {

            # Skip the test for the interaction between overflow and above-31
            # bit.  It is really testing other things than the partial
            # character tests, for which other tests in this file are
            # sufficient
            last if $will_overflow;

            foreach my $disallow_flag (0, $utf8n_flag_to_disallow) {
                my $partial = substr($bytes, 0, $j);
                my $ret_should_be;
                my $comment;
                if ($disallow_flag) {
                    $ret_should_be = 0;
                    $comment = "disallowed";
                    if ($j < $needed_to_discern_len) {
                        $ret_should_be = 1;
                        $comment .= ", but need $needed_to_discern_len bytes"
                                 .  " to discern:";
                    }
                }
                else {
                    $ret_should_be = 1;
                    $comment = "allowed";
                }

                undef @warnings_gotten;

                $ret = test_is_utf8_valid_partial_char_flags($partial, $j,
                                                             $disallow_flag);
                is($ret, $ret_should_be,
                                "    And is_utf8_valid_partial_char_flags("
                              . display_bytes($partial)
                              . "), $comment: returns $ret_should_be");
                is(scalar @warnings_gotten, 0,
                        "    And is_utf8_valid_partial_char_flags()"
                      . " generated no warnings")
                  or output_warnings(@warnings_gotten);
            }
        }
    }

    # This is more complicated than the malformations tested in other files in
    # this directory, as there are several orthogonal variables involved.  We
    # test most possible combinations
    foreach my $trial_warning_category ('surrogate', 'nonchar', 'non_unicode') {
      next if $skip_most_tests && $trial_warning_category ne $controlling_warning_category;
      foreach my $do_disallow (0, 1) {
        next if $skip_most_tests && ! $do_disallow;

            # We try each of the above with various combinations of
            # malformations that can occur on the same input sequence.
            foreach my $short ("", "short") {
              next if $skip_most_tests && $short;
              foreach my $unexpected_noncont ("",
                                              "unexpected non-continuation")
              {
                next if $skip_most_tests && $unexpected_noncont;
                foreach my $overlong ("", "overlong") {
                    next if $overlong && $skip_most_tests;

                    # If we're creating an overlong, it can't be longer than
                    # the maximum length, so skip if we're already at that
                    # length.
                    next if $overlong && $length >= $::max_bytes;

              # We classify the warnings into certain "interesting" types,
              # described later
              foreach my $warning_type (0..4) {
                next if $skip_most_tests && $warning_type != 1;
                foreach my $use_warn_flag (0, 1) {
                    next if $skip_most_tests && ! $use_warn_flag;

                    my $this_utf8n_flag_to_warn = $utf8n_flag_to_warn;
                    my $this_utf8n_flag_to_disallow = $utf8n_flag_to_disallow;
                    my $this_uvchr_flag_to_warn = $uvchr_flag_to_warn;
                    my $this_uvchr_flag_to_disallow = $uvchr_flag_to_disallow;

                    my $eval_warn;
                    my $expect_regular_warnings;
                    my $expect_warnings_for_malformed;
                    my $expect_warnings_for_overflow;

                    if ($warning_type == 0) {
                        $eval_warn = "use warnings; no warnings 'deprecated'";
                        $expect_regular_warnings = $use_warn_flag;
                        $expect_warnings_for_overflow = 1;
                        $expect_warnings_for_malformed = 1;
                    }
                    elsif ($warning_type == 1) {
                        $eval_warn = "no warnings";
                        $expect_regular_warnings = 0;
                        $expect_warnings_for_overflow = 0;
                        $expect_warnings_for_malformed = 0;
                    }
                    elsif ($warning_type == 2) {
                        $eval_warn = "no warnings; use warnings 'utf8'";
                        $expect_regular_warnings = $use_warn_flag;
                        $expect_warnings_for_overflow = 1;
                        $expect_warnings_for_malformed = 1;
                    }
                    elsif ($warning_type == 3) {
                        $eval_warn = "no warnings; use warnings"
                                   . " '$trial_warning_category'";
                        $expect_regular_warnings
                            = (   $trial_warning_category eq $controlling_warning_category
                               && $use_warn_flag);
                        $expect_warnings_for_overflow
                                    = $trial_warning_category eq 'non_unicode';
                        $expect_warnings_for_malformed = 0;
                    }
                    elsif ($warning_type == 4) {  # Like type 3, but uses the
                                                  # above-31-bit flags
                        # The complement flags were set up so that the
                        # above-31-bit flags have been tested that they don't
                        # trigger wrongly for too small code points.  And the
                        # flags have been set up so that those small code
                        # points are tested for being above Unicode.  What's
                        # left to test is that the large code points do
                        # trigger the above-31-bit flags.
                        next if ! $will_overflow && $allowed_uv < 0x80000000;
                        next if $trial_warning_category ne 'non_unicode';
                        $eval_warn = "no warnings; use warnings 'non_unicode'";
                        $expect_regular_warnings = 1;
                        $expect_warnings_for_overflow = 1;
                        $expect_warnings_for_malformed = 0;
                        $this_utf8n_flag_to_warn   = $::UTF8_WARN_ABOVE_31_BIT;
                        $this_utf8n_flag_to_disallow
                                                = $::UTF8_DISALLOW_ABOVE_31_BIT;
                        $this_uvchr_flag_to_warn = $::UNICODE_WARN_ABOVE_31_BIT;
                        $this_uvchr_flag_to_disallow
                                             = $::UNICODE_DISALLOW_ABOVE_31_BIT;
                    }
                    else {
                       die "Unexpected warning type '$warning_type'";
                    }

                    # We only need to test the case where all warnings are
                    # enabled (type 0) to see if turning off the warning flag
                    # causes things to not be output.  If those pass, then
                    # turning on some sub-category of warnings, or turning off
                    # warnings altogether are extremely likely to not output
                    # warnings either, given how the warnings subsystem is
                    # supposed to work, and this file assumes it does work.
                    next if $warning_type != 0 && ! $use_warn_flag;

                    # The convention is that the 'got' flag is the same value
                    # as the disallow one.  If this were violated, the tests
                    # here should start failing.
                    my $return_flag = $this_utf8n_flag_to_disallow;

                    my $this_warning_flags = ($use_warn_flag)
                                             ? $this_utf8n_flag_to_warn
                                             : 0;
                    my $this_disallow_flags = ($do_disallow)
                                             ? $this_utf8n_flag_to_disallow
                                             : 0;
                    my $this_bytes = $bytes;
                    my $this_length = $length;
                    my $expected_uv = $allowed_uv;
                    my $this_expected_len = $length;
                    my $this_needed_to_discern_len = $needed_to_discern_len;

                    my @malformation_names;
                    my @expected_warnings;
                    my @expected_return_flags;

                    # Now go through the possible malformations wanted,  and
                    # change the input accordingly.  We also can set up
                    # certain other things now, like whether we expect a
                    # return flag from this malformation and which flag.
                    if ($overlong) {
                            # For an overlong, we convert the original
                            # start byte into a continuation byte with
                            # the same data bits as originally. ...
                            substr($this_bytes, 0, 1)
                                = start_byte_to_cont(substr($this_bytes,
                                                            0, 1));

                            # ... Then we prepend it with a known
                            # overlong sequence.  This should evaluate
                            # to the exact same code point as the
                            # original.
                            $this_bytes
                            = I8_to_native("\xff")
                            . (I8_to_native(chr $::lowest_continuation)
                            x ( $::max_bytes - 1 - length($this_bytes)))
                            . $this_bytes;
                            $this_length = length($this_bytes);
                            $this_needed_to_discern_len
                                 = $::max_bytes - ($this_expected_len
                                               - $this_needed_to_discern_len);
                            $this_expected_len = $::max_bytes;
                            push @expected_return_flags, $::UTF8_GOT_LONG;
                        push @malformation_names, $overlong;
                        if ($expect_warnings_for_malformed) {
                            if (   ! $short
                                && ! $unexpected_noncont
                                && ! $will_overflow)
                            {
                                my $overlong_bytes
                                        = display_bytes_no_quotes($this_bytes);
                                my $correct_bytes
                                             = display_bytes_no_quotes($bytes);
                                push @expected_warnings,
                                     qr/\QMalformed UTF-8 character:\E
                                        \Q $overlong_bytes (overlong;\E
                                        \Q instead use $correct_bytes to\E
                                        \Q represent U+$uv_string)/x;
                            }
                            else {
                                push @expected_warnings, qr/overlong/;
                            }
                        }
                    }

                    if ($short) {
                        push @malformation_names, $short;
                        push @expected_warnings, qr/short/
                                            if $expect_warnings_for_malformed;

                        # To force this malformation, just tell the test to
                        # not look as far as it should into the input.
                        $this_length--;
                        $this_expected_len--;
                        push @expected_return_flags, $::UTF8_GOT_SHORT;
                    }

                    if ($unexpected_noncont) {
                        push @malformation_names, $unexpected_noncont;
                        push @expected_warnings, qr/$unexpected_noncont/
                                            if $expect_warnings_for_malformed;

                        # To force this malformation, change the final
                        # continuation byte into a non continuation.
                        my $pos = ($short) ? -2 : -1;
                        substr($this_bytes, $pos, 1) = '?';
                        $this_expected_len--;
                        push @expected_return_flags,
                                        $::UTF8_GOT_NON_CONTINUATION;
                    }

                    # The overflow malformation is done differently than other
                    # malformations.  It comes from manually typed tests in
                    # the test array, but it also is above Unicode and uses
                    # Perl extended UTF-8, so affects some of the flags being
                    # tested.  We now make it be treated like one of the other
                    # generated malformations.
                    if ($will_overflow) {

                        # An overflow is (way) above Unicode, and overrides
                        # everything else.
                        $expect_regular_warnings = 0;

                        push @malformation_names, 'overflow';
                        if ($expect_warnings_for_overflow) {
                            my $qr = display_bytes_no_quotes(
                                   substr($this_bytes, 0, $this_expected_len));
                            $qr = qr/\QMalformed UTF-8 character: \E
                                     \Q$qr (overflows)\E/x;
                            push @expected_warnings, $qr;
                        }
                        push @expected_return_flags, $::UTF8_GOT_OVERFLOW;
                    }

                    # Here, we've set things up based on the malformations.
                    # Now generate the text for them for the test name.
                    my $malformations_name = "";
                    if (@malformation_names) {
                        $malformations_name .= "malformation";
                        $malformations_name .= "s" if @malformation_names > 1;
                        $malformations_name .= ": ";
                        $malformations_name .=  join "/", @malformation_names;
                        $malformations_name =  " ($malformations_name)";
                    }

                    # It may be that the malformations have shortened the
                    # amount of input we look at so much that we can't tell
                    # what the category the code point was in.  Otherwise, set
                    # up the expected return flags based on the warnings and
                    # disallowments.
                    if ($this_expected_len < $this_needed_to_discern_len) {
                        $expect_regular_warnings = 0;
                    }
                    elsif (   ($this_warning_flags & $this_utf8n_flag_to_warn)
                           || (  $this_disallow_flags
                               & $this_utf8n_flag_to_disallow))
                    {
                        push @expected_return_flags, $return_flag;
                    }

                    # Finish setting up the expected warning.
                    if ($expect_regular_warnings) {

                        # So far the array contains warnings generated by
                        # malformations.  Add the expected regular one.
                        unshift @expected_warnings, $cp_message_qr;

                        # But it may need to be modified, because either of
                        # these malformations means we can't determine the
                        # expected code point.
                        if ($short || $unexpected_noncont) {
                            my $first_byte = substr($this_bytes, 0, 1);
                            $expected_warnings[0] = display_bytes(
                                    substr($this_bytes, 0, $this_expected_len));
                            $expected_warnings[0]
                                = qr/[Aa]\Qny UTF-8 sequence that starts with\E
                                     \Q $expected_warnings[0]\E
                                     \Q $non_cp_trailing_text\E/x;
                        }
                    }

                    # Is effectively disallowed if we've set up a
                    # malformation, even if the flag indicates it is
                    # allowed.  Fix up test name to indicate this as
                    # well
                    my $disallowed = (  $this_disallow_flags
                                      & $this_utf8n_flag_to_disallow)
                                  || $malformations_name;
                    my $this_name = "utf8n_to_uvchr_error() $testname: "
                                                . (($disallowed)
                                                   ? 'disallowed'
                                                   : 'allowed');
                    $this_name .= ", $eval_warn";
                    $this_name .= ", " . ((  $this_warning_flags
                                            & $this_utf8n_flag_to_warn)
                                          ? 'with flag for raising warnings'
                                          : 'no flag for raising warnings');
                    $this_name .= $malformations_name;

                    undef @warnings_gotten;
                    my $ret_ref;
                    my $this_flags = $this_warning_flags|$this_disallow_flags;
                    my $eval_text =      "$eval_warn; \$ret_ref"
                            . " = test_utf8n_to_uvchr_error("
                            . "'$this_bytes', $this_length, $this_flags)";
                    eval "$eval_text";
                    if (! ok ("$@ eq ''", "$this_name: eval succeeded"))
                    {
                        diag "\$@='$@'; call was: "
                           . utf8n_display_call($eval_text);
                        next;
                    }
                    if ($disallowed) {
                        is($ret_ref->[0], 0, "    And returns 0")
                          or diag "Call was: " . utf8n_display_call($eval_text);
                    }
                    else {
                        is($ret_ref->[0], $expected_uv,
                                "    And returns expected uv: "
                              . $uv_string)
                          or diag "Call was: " . utf8n_display_call($eval_text);
                    }
                    is($ret_ref->[1], $this_expected_len,
                                        "    And returns expected length:"
                                      . " $this_expected_len")
                      or diag "Call was: " . utf8n_display_call($eval_text);

                    my $returned_flags = $ret_ref->[2];

                    for (my $i = @expected_return_flags - 1; $i >= 0; $i--) {
                        if ($expected_return_flags[$i] & $returned_flags) {
                            if ($expected_return_flags[$i]
                                            == $::UTF8_DISALLOW_ABOVE_31_BIT)
                            {
                                pass("    Expected and got return flag for"
                                   . " above_31_bit");
                            }
                                   # The first entries in this are
                                   # malformations
                            elsif ($i > @malformation_names - 1)  {
                                pass("    Expected and got return flag"
                                   . " for " . $trial_warning_category);
                            }
                            else {
                                pass("    Expected and got return flag for "
                                   . $malformation_names[$i]
                                   . " malformation");
                            }
                            $returned_flags &= ~$expected_return_flags[$i];
                            splice @expected_return_flags, $i, 1;
                        }
                    }

                    is($returned_flags, 0,
                       "    Got no unexpected return flags")
                      or diag "The unexpected flags gotten were: "
                           . (flags_to_text($returned_flags,
                                            \@utf8n_flags_to_text)
                                # We strip off any prefixes from the flag
                                # names
                             =~ s/ \b [A-Z] _ //xgr);
                    is (scalar @expected_return_flags, 0,
                        "    Got all expected return flags")
                        or diag "The expected flags not gotten were: "
                           . (flags_to_text(eval join("|",
                                                        @expected_return_flags),
                                            \@utf8n_flags_to_text)
                                # We strip off any prefixes from the flag
                                # names
                             =~ s/ \b [A-Z] _ //xgr);

                    do_warnings_test(@expected_warnings)
                      or diag "Call was: " . utf8n_display_call($eval_text);
                    undef @warnings_gotten;

                    # Check CHECK_ONLY results when the input is
                    # disallowed.  Do this when actually disallowed,
                    # not just when the $this_disallow_flags is set
                    if ($disallowed) {
                        my $this_flags = $this_disallow_flags|$::UTF8_CHECK_ONLY;
                        my $eval_text = "use warnings; no warnings 'deprecated'; \$ret_ref ="
                                      . " test_utf8n_to_uvchr_error('"
                                      . "$this_bytes', $this_length,"
                                      . " $this_flags)";
                        eval $eval_text;
                        if (! ok ("$@ eq ''",
                            "    And eval succeeded with CHECK_ONLY"))
                        {
                            diag "\$@='$@'; Call was: "
                               . utf8n_display_call($eval_text);
                            next;
                        }
                        is($ret_ref->[0], 0, "    CHECK_ONLY: Returns 0")
                          or diag "Call was: " . utf8n_display_call($eval_text);
                        is($ret_ref->[1], -1,
                                       "    CHECK_ONLY: returns -1 for length")
                          or diag "Call was: " . utf8n_display_call($eval_text);
                        if (! is(scalar @warnings_gotten, 0,
                                      "    CHECK_ONLY: no warnings generated"))
                        {
                            diag "Call was: " . utf8n_display_call($eval_text);
                            output_warnings(@warnings_gotten);
                        }
                    }

                    # Now repeat some of the above, but for
                    # uvchr_to_utf8_flags().  Since this comes from an
                    # existing code point, it hasn't overflowed, and
                    # isn't malformed.
                    next if @malformation_names;

                    $this_warning_flags = ($use_warn_flag)
                                          ? $this_uvchr_flag_to_warn
                                          : 0;
                    $this_disallow_flags = ($do_disallow)
                                           ? $this_uvchr_flag_to_disallow
                                           : 0;

                    $disallowed = $this_disallow_flags
                                & $this_uvchr_flag_to_disallow;
                    $this_name .= ", " . ((  $this_warning_flags
                                           & $this_utf8n_flag_to_warn)
                                          ? 'with flag for raising warnings'
                                          : 'no flag for raising warnings');

                    $this_name = "uvchr_to_utf8_flags() $testname: "
                                            . (($disallowed)
                                                ? 'disallowed'
                                                : 'allowed');
                    $this_name .= ", $eval_warn";
                    $this_name .= ", " . ((  $this_warning_flags
                                           & $this_uvchr_flag_to_warn)
                                        ? 'with warning flag'
                                        : 'no warning flag');

                    undef @warnings_gotten;
                    my $ret;
                    $this_flags = $this_warning_flags|$this_disallow_flags;
                    $eval_text = "$eval_warn; \$ret ="
                            . " test_uvchr_to_utf8_flags("
                            . "$allowed_uv, $this_flags)";
                    eval "$eval_text";
                    if (! ok ("$@ eq ''", "$this_name: eval succeeded"))
                    {
                        diag "\$@='$@'; call was: "
                           . uvchr_display_call($eval_text);
                        next;
                    }
                    if ($disallowed) {
                        is($ret, undef, "    And returns undef")
                          or diag "Call was: " . uvchr_display_call($eval_text);
                    }
                    else {
                        is($ret, $this_bytes, "    And returns expected string")
                          or diag "Call was: " . uvchr_display_call($eval_text);
                    }

                    do_warnings_test(@expected_warnings)
                      or diag "Call was: " . uvchr_display_call($eval_text);
                }
              }
            }
          }
        }
      }
    }
}

done_testing;