xref: /openbsd/gnu/usr.bin/perl/t/re/fold_grind.pl (revision 5486feef)
1# Grind out a lot of combinatoric tests for folding.
2# It uses various charset modifiers, passed in via $::TEST_CHUNK.  The caller
3# will also have set the locale to use if /l is the modifier.
4#   L is a pseudo-modifier that indicates to use the modifier /l instead, and
5#     the locale set by the caller is known to be UTF-8,
6#   T is a pseudo-modifier that indicates to use the pseudo modifier /L
7#     instead, and the locale set by the caller is known to be Turkic UTF-8,
8
9binmode STDOUT, ":utf8";
10
11BEGIN {
12    chdir 't' if -d 't';
13    require './test.pl';
14    set_up_inc('../lib');
15    require Config; Config->import;
16    skip_all_if_miniperl("no dynamic loading on miniperl, no Encode nor POSIX");
17    if ($^O eq 'dec_osf') {
18      skip_all("$^O cannot handle this test");
19    }
20
21    watchdog(5 * 60);
22    require './loc_tools.pl';
23}
24
25use charnames ":full";
26
27my $DEBUG = 0;  # Outputs extra information for debugging this .t
28
29use strict;
30use warnings;
31no warnings 'locale';   # Plenty of these would otherwise get generated
32use Encode;
33use POSIX;
34
35my $charset = $::TEST_CHUNK;
36my $use_turkic_rules = 0;
37
38if ($charset eq 'T') {
39    $charset = 'L';
40    $use_turkic_rules = 1;
41}
42
43my $has_LC_CTYPE = is_category_valid('LC_CTYPE');
44
45# Special-cased characters in the .c's that we want to make sure get tested.
46my %be_sure_to_test = (
47        chr utf8::unicode_to_native(0xDF) => 1, # LATIN_SMALL_LETTER_SHARP_S
48
49        # This is included because the uppercase occupies more bytes, but the
50        # first two bytes of their representations differ only in one bit,
51        # that could lead the code looking for shortcuts astray; you can't do
52        # certain shortcuts if the lengths differ
53        "\x{29E}" => 1, # LATIN SMALL LETTER TURNED K
54
55        "\x{390}" => 1, # GREEK_SMALL_LETTER_IOTA_WITH_DIALYTIKA_AND_TONOS
56        "\x{3B0}" => 1, # GREEK_SMALL_LETTER_UPSILON_WITH_DIALYTIKA_AND_TONOS
57
58        # This is included because the uppercase and lowercase differ by only
59        # a single bit and it is in the first of the two byte representations.
60        # This showed that a previous way was erroneous of calculating if
61        # initial substrings were closely-related bit-wise.
62        "\x{3CC}" => 1, # GREEK SMALL LETTER OMICRON WITH TONOS
63
64        "\x{1E9E}" => 1, # LATIN_CAPITAL_LETTER_SHARP_S
65        "\x{1FD3}" => 1, # GREEK SMALL LETTER IOTA WITH DIALYTIKA AND OXIA
66        "\x{1FE3}" => 1, # GREEK SMALL LETTER UPSILON WITH DIALYTIKA AND OXIA
67
68        # These are included because they are adjacent and fold to the same
69        # result, U+01C6.  This has tripped up the code in the past that
70        # wrongly thought that sequential code points must fold to sequential
71        # code points
72        "\x{01C4}" => 1, # LATIN CAPITAL LETTER DZ WITH CARON
73        "\x{01C5}" => 1, # LATIN CAPITAL LETTER D WITH SMALL LETTER Z WITH CARON
74
75        "I" => 1,
76);
77
78# Tests both unicode and not, so make sure not implicitly testing unicode
79no feature 'unicode_strings';
80
81# Case-insensitive matching is a large and complicated issue.  Perl does not
82# implement it fully, properly.  For example, it doesn't include normalization
83# as part of the equation.  To test every conceivable combination is clearly
84# impossible; these tests are mostly drawn from visual inspection of the code
85# and experience, trying to exercise all areas.
86
87# There are three basic ranges of characters that Perl may treat differently:
88# 1) Invariants under utf8 which on ASCII-ish machines are ASCII, and are
89#    referred to here as ASCII.  On EBCDIC machines, the non-ASCII invariants
90#    are all controls that fold to themselves.
91my $ASCII = 1;
92
93# 2) Other characters that fit into a byte but are different in utf8 than not;
94#    here referred to, taking some liberties, as Latin1.
95my $Latin1 = 2;
96
97# 3) Characters that won't fit in a byte; here referred to as Unicode
98my $Unicode = 3;
99
100# Within these basic groups are equivalence classes that testing any character
101# in is likely to lead to the same results as any other character.  This is
102# used to cut down the number of tests needed, unless PERL_RUN_SLOW_TESTS is
103# set.
104my $skip_apparently_redundant = ! $ENV{PERL_RUN_SLOW_TESTS};
105
106# Additionally parts of this test run a lot of subtests, outputting the
107# resulting TAP can be expensive so the tests are summarised internally. The
108# PERL_DEBUG_FULL_TEST environment variable can be set to produce the full
109# output for debugging purposes.
110
111sub range_type {
112    my $ord = ord shift;
113
114    return $ASCII if utf8::native_to_unicode($ord) < 128;
115    return $Latin1 if $ord < 256;
116    return $Unicode;
117}
118
119sub numerically {
120    return $a <=> $b
121}
122
123my $list_all_tests = $ENV{PERL_DEBUG_FULL_TEST} || $DEBUG;
124$| = 1 if $list_all_tests;
125
126# Significant time is saved by not outputting each test but grouping the
127# output into subtests
128my $okays;          # Number of ok's in current subtest
129my $this_iteration; # Number of possible tests in current subtest
130my $count = 0;      # Number of subtests = number of total tests
131
132sub run_test($$$$) {
133    my ($test, $todo, $do_we_output_locale_name, $debug) = @_;
134
135    $debug = "" unless $DEBUG;
136    my $res = eval $test;
137
138    if ($do_we_output_locale_name) {
139        $do_we_output_locale_name = 'setlocale(LC_CTYPE, "'
140                         .  POSIX::setlocale(&POSIX::LC_CTYPE)
141                         . '"); ';
142    }
143    if (!$res || $list_all_tests) {
144      # Failed or debug; output the result
145      $count++;
146      ok($res, "$do_we_output_locale_name$test; $debug");
147    } else {
148      # Just count the test as passed
149      $okays++;
150    }
151    $this_iteration++;
152}
153
154my %has_test_by_participants;   # Makes sure has tests for each range and each
155                                # number of characters that fold to the same
156                                # thing
157my %has_test_by_byte_count; # Makes sure has tests for each combination of
158                            # n bytes folds to m bytes
159
160my %tests; # The set of tests we expect to pass.
161# Each key is a code point that folds to something else.
162# Each value is a list of things that the key folds to.  If the 'thing' is a
163# single code point, it is that ordinal.  If it is a multi-char fold, it is an
164# ordered list of the code points in that fold.  Here's an example for 'S':
165#  '83' => [ 115, 383 ]
166#
167# And one for a multi-char fold: \xDF
168#  223 => [
169#            [  # 'ss'
170#                83,
171#                83
172#            ],
173#            [  # 'SS'
174#                115,
175#                115
176#            ],
177#            [  # LATIN SMALL LETTER LONG S
178#                383,
179#                383
180#            ],
181#          7838 # LATIN_CAPITAL_LETTER_SHARP_S
182#        ],
183
184my %neg_tests;  # Same format, but we expect these tests to fail
185
186my %folds; # keys are code points that fold; values are either 0 or 1 which
187           # in turn are keys with their values each a list of code points the
188           # code point key folds to.  The folds under 1 are the ones that are
189           # valid in this run; the ones under 0 are ones valid under other
190           # circumstances.
191
192my %inverse_folds;  # keys are strings of the folded-to; then come a layer of
193                    # 0 or 1, like %folds.  The lowest values are lists of
194                    # characters that fold to them
195
196# Here's a portion of an %inverse_folds in a run where Turkic folds are not
197# legal, so \x{130} doesn't fold to 'i' in this run.
198#         'h' => {
199#                  '1' => [
200#                           'H'
201#                         ]
202#                },
203#         "h\x{331}" => {
204#                         '1' => [
205#                                  "\x{1e96}"
206#                                ]
207#                       },
208#         'i' => {
209#                  '0' => [
210#                           "\x{130}"
211#                         ],
212#                  '1' => [
213#                           'I'
214#                         ]
215#                },
216#         "i\x{307}" => {
217#                         '1' => [
218#                                  "\x{130}"
219#                                ]
220#                       },
221#         'j' => {
222#                  '1' => [
223#                           'J'
224#                         ]
225#                },
226
227sub add_test($$@) {
228    my ($tests_ref, $to, @from) = @_;
229
230    # Called to cause the input to be tested by adding to $%tests_ref.  @from
231    # is the list of characters that fold to the string $to.  @from should be
232    # sorted so the lowest code point is first....
233    # The input is in string form; %tests uses code points, so have to
234    # convert.
235
236    my $to_chars = length $to;
237    my @test_to;        # List of tests for $to
238
239    if ($to_chars == 1) {
240        @test_to = ord $to;
241    }
242    else {
243        push @test_to, [ map { ord $_ } split "", $to ];
244
245        # For multi-char folds, we also test that things that can fold to each
246        # individual character in the fold also work.  If we were testing
247        # comprehensively, we would try every combination of upper and lower
248        # case in the fold, but it will have to suffice to avoid running
249        # forever to make sure that each thing that folds to these is tested
250        # at least once.  Because of complement matching ([^...]), we need to
251        # do both the folded, and the folded-from.
252        # We first look at each character in the multi-char fold, and save how
253        # many characters fold to it; and also the maximum number of such
254        # folds
255        my @folds_to_count;     # 0th char in fold is index 0 ...
256        my $max_folds_to = 0;
257
258        for (my $i = 0; $i < $to_chars; $i++) {
259            my $to_char = substr($to, $i, 1);
260            if (exists $inverse_folds{$to_char}{1}) {
261                $folds_to_count[$i] = scalar @{$inverse_folds{$to_char}{1}};
262                $max_folds_to = $folds_to_count[$i] if $max_folds_to < $folds_to_count[$i];
263            }
264            else {
265                $folds_to_count[$i] = 0;
266            }
267        }
268
269        # We will need to generate as many tests as the maximum number of
270        # folds, so that each fold will have at least one test.
271        # For example, consider character X which folds to the three character
272        # string 'xyz'.  If 2 things fold to x (X and x), 4 to y (Y, Y'
273        # (Y-prime), Y'' (Y-prime-prime), and y), and 1 thing to z (itself), 4
274        # tests will be generated:
275        #   xyz
276        #   XYz
277        #   xY'z
278        #   xY''z
279        for (my $i = 0; $i < $max_folds_to; $i++) {
280            my @this_test_to;   # Assemble a single test
281
282            # For each character in the multi-char fold ...
283            for (my $j = 0; $j < $to_chars; $j++) {
284                my $this_char = substr($to, $j, 1);
285
286                # Use its corresponding inverse fold, if available.
287                if (   $i < $folds_to_count[$j]
288                    && exists $inverse_folds{$this_char}{1})
289                  {
290                    push @this_test_to, ord $inverse_folds{$this_char}{1}[$i];
291                }
292                else {  # Or else itself.
293                    push @this_test_to, ord $this_char;
294                }
295            }
296
297            # Add this test to the list
298            push @test_to, [ @this_test_to ];
299        }
300
301        # Here, have assembled all the tests for the multi-char fold.  Sort so
302        # lowest code points are first for consistency and aesthetics in
303        # output.  We know there are at least two characters in the fold, but
304        # I haven't bothered to worry about sorting on an optional third
305        # character if the first two are identical.
306        @test_to = sort { ($a->[0] == $b->[0])
307                           ? $a->[1] <=> $b->[1]
308                           : $a->[0] <=> $b->[0]
309                        } @test_to;
310    }
311
312
313    # This test is from n bytes to m bytes.  Record that so won't try to add
314    # another test that does the same.
315    use bytes;
316    my $to_bytes = length $to;
317    foreach my $from_map (@from) {
318        $has_test_by_byte_count{length $from_map}{$to_bytes} = $to;
319    }
320    no bytes;
321
322    my $ord_smallest_from = ord shift @from;
323    if (exists $tests_ref->{$ord_smallest_from}) {
324        die "There are already tests for $ord_smallest_from"
325    };
326
327    # Add in the fold tests,
328    push @{$tests_ref->{$ord_smallest_from}}, @test_to;
329
330    # Then any remaining froms in the equivalence class.
331    push @{$tests_ref->{$ord_smallest_from}}, map { ord $_ } @from;
332}
333
334# Get the Unicode rules and construct inverse mappings from them
335
336use Unicode::UCD;
337my $file="../lib/unicore/CaseFolding.txt";
338
339# Use the Unicode data file if we are on an ASCII platform (which its data is
340# for), and it is in the modern format (starting in Unicode 3.1.0) and it is
341# available.  This avoids being affected by potential bugs introduced by other
342# layers of Perl
343if ($::IS_ASCII
344    && pack("C*", split /\./, Unicode::UCD::UnicodeVersion()) ge v3.1.0
345    && open my $fh, "<", $file)
346{
347    # We process the file in reverse order because its easier to see the T
348    # entry first and then know that the next line we process is the
349    # corresponding one for non-T.
350    my @rules = <$fh>;
351    my $prev_was_turkic = 0;
352    while (defined ($_ = pop @rules)) {
353        chomp;
354
355        # Lines look like (though without the initial '#')
356        #0130; F; 0069 0307; # LATIN CAPITAL LETTER I WITH DOT ABOVE
357
358        # Get rid of comments, ignore blank or comment-only lines
359        my $line = $_ =~ s/ (?: \s* \# .* )? $ //rx;
360        next unless length $line;
361        my ($hex_from, $fold_type, @hex_folded) = split /[\s;]+/, $line;
362
363        next if $fold_type eq 'S';  # If Unicode's tables are correct, the F
364                                    # should be a superset of S
365        next if $fold_type eq 'I';  # Perl doesn't do old Turkish folding
366
367        my $test_type;
368        if ($fold_type eq 'T') {
369            $test_type = 0 + $use_turkic_rules;
370            $prev_was_turkic = 1;
371        }
372        elsif ($prev_was_turkic) {
373            $test_type = 0 + ! $use_turkic_rules;
374            $prev_was_turkic = 0;
375        }
376        else {
377            $test_type = 1;
378            $prev_was_turkic = 0;
379        }
380
381        my $from = hex $hex_from;
382        my @to = map { hex $_ } @hex_folded;
383        push @{$folds{$from}{$test_type}}, @to;
384
385        my $folded_str = pack ("U0U*", @to);
386        push @{$inverse_folds{$folded_str}{$test_type}}, chr $from;
387    }
388}
389else {  # Here, can't use the .txt file: read the Unicode rules file and
390        # construct inverse mappings from it
391
392    skip_all "Don't know how to generate turkic rules on this platform"
393                                                            if $use_turkic_rules;
394    my ($invlist_ref, $invmap_ref, undef, $default)
395                                    = Unicode::UCD::prop_invmap('Case_Folding');
396    for my $i (0 .. @$invlist_ref - 1 - 1) {
397        next if $invmap_ref->[$i] == $default;
398
399        # Make into an array if not so already, so can treat uniformly below
400        $invmap_ref->[$i] = [ $invmap_ref->[$i] ] if ! ref $invmap_ref->[$i];
401
402        # Each subsequent element of the range requires adjustment of +1 from
403        # the previous element
404        my $adjust = -1;
405        for my $j ($invlist_ref->[$i] .. $invlist_ref->[$i+1] -1) {
406            $adjust++;
407            my @to = map { $_ + $adjust } @{$invmap_ref->[$i]};
408            push @{$folds{$j}{1}}, @to;
409            my $folded_str = join "", map { chr } @to;
410            utf8::upgrade($folded_str);
411            #note (sprintf "%d: %04X: %s", __LINE__, $j, join " ",
412            #    map { sprintf "%04X", $_  + $adjust } @{$invmap_ref->[$i]});
413            push @{$inverse_folds{$folded_str}{1}}, chr $j;
414        }
415    }
416}
417
418# Analyze the data and generate tests to get adequate test coverage.  We sort
419# things so that smallest code points are done first.
420foreach my $to (sort { $a cmp $b } keys %inverse_folds)
421{
422TO:
423  foreach my $tests_ref (\%tests, \%neg_tests) {
424    my $test_type = ($tests_ref == \%tests) ? 1 : 0;
425
426    next unless exists $inverse_folds{$to}{$test_type};
427
428    # Within each fold, sort so that the smallest code points are done first
429    @{$inverse_folds{$to}{$test_type}} = sort { $a cmp $b } @{$inverse_folds{$to}{$test_type}};
430    my @from = @{$inverse_folds{$to}{$test_type}};
431
432    # Just add it to the tests if doing complete coverage
433    if (! $skip_apparently_redundant) {
434        add_test($tests_ref, $to, @from);
435        next TO;
436    }
437
438    my $to_chars = length $to;
439    my $to_range_type = range_type(substr($to, 0, 1));
440
441    # If this is required to be tested, do so.  We check for these first, as
442    # they will take up slots of byte-to-byte combinations that we otherwise
443    # would have to have other tests to get.
444    foreach my $from_map (@from) {
445        if (exists $be_sure_to_test{$from_map}) {
446            add_test($tests_ref, $to, @from);
447            next TO;
448        }
449    }
450
451    # If the fold contains heterogeneous range types, is suspect and should be
452    # tested.
453    if ($to_chars > 1) {
454        foreach my $char (split "", $to) {
455            if (range_type($char) != $to_range_type) {
456                add_test($tests_ref, $to, @from);
457                next TO;
458            }
459        }
460    }
461
462    # If the mapping crosses range types, is suspect and should be tested
463    foreach my $from_map (@from) {
464        if (range_type($from_map) != $to_range_type) {
465            add_test($tests_ref, $to, @from);
466            next TO;
467        }
468    }
469
470    # Here, all components of the mapping are in the same range type.  For
471    # single character folds, we test one case in each range type that has 2
472    # particpants, 3 particpants, etc.
473    if ($to_chars == 1) {
474        if (! exists $has_test_by_participants{scalar @from}{$to_range_type}) {
475            add_test($tests_ref, $to, @from);
476            $has_test_by_participants{scalar @from}{$to_range_type} = $to;
477            next TO;
478        }
479    }
480
481    # We also test all combinations of mappings from m to n bytes.  This is
482    # because the regex optimizer cares.  (Don't bother worrying about that
483    # Latin1 chars will occupy a different number of bytes under utf8, as
484    # there are plenty of other cases that catch these byte numbers.)
485    use bytes;
486    my $to_bytes = length $to;
487    foreach my $from_map (@from) {
488        if (! exists $has_test_by_byte_count{length $from_map}{$to_bytes}) {
489            add_test($tests_ref, $to, @from);
490            next TO;
491        }
492    }
493  }
494}
495
496# For each range type, test additionally a character that folds to itself
497add_test(\%tests, ":", ":");
498add_test(\%tests, chr utf8::unicode_to_native(0xF7), chr utf8::unicode_to_native(0xF7));
499add_test(\%tests, chr 0x2C7, chr 0x2C7);
500
501# To cut down on the number of tests
502my $has_tested_aa_above_latin1;
503my $has_tested_latin1_aa;
504my $has_tested_ascii_aa;
505my $has_tested_l_above_latin1;
506my $has_tested_above_latin1_l;
507my $has_tested_ascii_l;
508my $has_tested_above_latin1_d;
509my $has_tested_ascii_d;
510my $has_tested_non_latin1_d;
511my $has_tested_above_latin1_a;
512my $has_tested_ascii_a;
513my $has_tested_non_latin1_a;
514
515# For use by pairs() in generating combinations
516sub prefix {
517    my $p = shift;
518    map [ $p, $_ ], @_
519}
520
521# Returns all ordered combinations of pairs of elements from the input array.
522# It doesn't return pairs like (a, a), (b, b).  Change the slice to an array
523# to do that.  This was just to have fewer tests.
524sub pairs (@) {
525    #print STDERR __LINE__, ": ", join(" XXX ", map { sprintf "%04X", $_ } @_), "\n";
526    map { prefix $_[$_], @_[0..$_-1, $_+1..$#_] } 0..$#_
527}
528
529# Finally ready to do the tests
530foreach my $tests_ref (\%neg_tests, \%tests) {
531foreach my $test (sort { numerically } keys %{$tests_ref}) {
532
533  my $previous_target;
534  my $previous_pattern;
535  my @pairs = pairs(sort numerically $test, @{$tests_ref->{$test}});
536
537  # Each fold can be viewed as a closure of all the characters that
538  # participate in it.  Look at each possible pairing from a closure, with the
539  # first member of the pair the target string to match against, and the
540  # second member forming the pattern.  Thus each fold member gets tested as
541  # the string, and the pattern with every other member in the opposite role.
542  while (my $pair = shift @pairs) {
543    my ($target, $pattern) = @$pair;
544
545    # When testing a char that doesn't fold, we can get the same
546    # permutation twice; so skip all but the first.
547    next if $previous_target
548            && $previous_target == $target
549            && $previous_pattern == $pattern;
550    ($previous_target, $previous_pattern) = ($target, $pattern);
551
552    # Each side may be either a single char or a string.  Extract each into an
553    # array (perhaps of length 1)
554    my @target, my @pattern;
555    @target = (ref $target) ? @$target : $target;
556    @pattern = (ref $pattern) ? @$pattern : $pattern;
557
558    # We are testing just folds to/from a single character.  If our pairs
559    # happens to generate multi/multi, skip.
560    next if @target > 1 && @pattern > 1;
561
562    # Get in hex form.
563    my @x_target = map { sprintf "\\x{%04X}", $_ } @target;
564    my @x_pattern = map { sprintf "\\x{%04X}", $_ } @pattern;
565
566    my $target_above_latin1 = grep { $_ > 255 } @target;
567    my $pattern_above_latin1 = grep { $_ > 255 } @pattern;
568    my $target_has_ascii = grep { utf8::native_to_unicode($_) < 128 } @target;
569    my $pattern_has_ascii = grep { utf8::native_to_unicode($_) < 128 } @pattern;
570    my $target_only_ascii = ! grep { utf8::native_to_unicode($_) > 127 } @target;
571    my $pattern_only_ascii = ! grep { utf8::native_to_unicode($_) > 127 } @pattern;
572    my $target_has_latin1 = grep { $_ < 256 } @target;
573    my $target_has_upper_latin1
574                = grep { $_ < 256 && utf8::native_to_unicode($_) > 127 } @target;
575    my $pattern_has_upper_latin1
576                = grep { $_ < 256 && utf8::native_to_unicode($_) > 127 } @pattern;
577    my $pattern_has_latin1 = grep { $_ < 256 } @pattern;
578    my $is_self = @target == 1 && @pattern == 1 && $target[0] == $pattern[0];
579
580    # We don't test multi-char folding into other multi-chars.  We are testing
581    # a code point that folds to or from other characters.  Find the single
582    # code point for diagnostic purposes.  (If both are single, choose the
583    # target string)
584    my $ord = @target == 1 ? $target[0] : $pattern[0];
585    my $progress = sprintf "%04X: \"%s\" and /%s/",
586                            $test,
587                            join("", @x_target),
588                            join("", @x_pattern);
589    #note $progress;
590
591    # Now grind out tests, using various combinations.
592    {
593      my $charset_mod = lc $charset;
594      my $current_locale = ($has_LC_CTYPE)
595                           ? setlocale(&POSIX::LC_CTYPE)
596                           : 'C';
597      $current_locale = 'C locale' if $current_locale eq 'C';
598      $okays = 0;
599      $this_iteration = 0;
600
601      # To cut down somewhat on the enormous quantity of tests this currently
602      # runs, skip some for some of the character sets whose results aren't
603      # likely to differ from others.  But run all tests on the code points
604      # that don't fold, plus one other set in each range group.
605      if (! $is_self) {
606
607        # /aa should only affect things with folds in the ASCII range.  But, try
608        # it on one set in the other ranges just to make sure it doesn't break
609        # them.
610        if ($charset eq 'aa') {
611
612          # It may be that this $pair of code points to test are both
613          # non-ascii, but if either of them actually fold to ascii, that is
614          # suspect and should be tested.  So for /aa, use whether their folds
615          # are ascii or not
616          my $target_has_ascii = $target_has_ascii;
617          my $pattern_has_ascii = $pattern_has_ascii;
618          if (! $target_has_ascii) {
619            foreach my $cp (@target) {
620              if (exists $folds{$cp}{1}
621                  && grep { utf8::native_to_unicode($_) < 128 } @{$folds{$cp}{1}} )
622              {
623                  $target_has_ascii = 1;
624                  last;
625              }
626            }
627          }
628          if (! $pattern_has_ascii) {
629            foreach my $cp (@pattern) {
630              if (exists $folds{$cp}{1}
631                  && grep { utf8::native_to_unicode($_) < 128 } @{$folds{$cp}}{1} )
632              {
633                  $pattern_has_ascii = 1;
634                  last;
635              }
636            }
637          }
638
639          if (! $target_has_ascii && ! $pattern_has_ascii) {
640            if ($target_above_latin1 || $pattern_above_latin1) {
641              next if defined $has_tested_aa_above_latin1
642                      && $has_tested_aa_above_latin1 != $test;
643              $has_tested_aa_above_latin1 = $test;
644            }
645            next if defined $has_tested_latin1_aa
646                    && $has_tested_latin1_aa != $test;
647            $has_tested_latin1_aa = $test;
648          }
649          elsif ($target_only_ascii && $pattern_only_ascii) {
650
651              # And, except for one set just to make sure, skip tests
652              # where both elements in the pair are ASCII.  If one works for
653              # aa, the others are likely too.  This skips tests where the
654              # fold is from non-ASCII to ASCII, but this part of the test
655              # is just about the ASCII components.
656              next if defined $has_tested_ascii_l
657                      && $has_tested_ascii_l != $test;
658              $has_tested_ascii_l = $test;
659          }
660        }
661        elsif ($charset eq 'l') {
662
663          # For l, don't need to test beyond one set those things that are
664          # all above latin1, because unlikely to have different successes
665          # than /u.  But, for the same reason as described in the /aa above,
666          # it is suspect and should be tested, if either of the folds are to
667          # latin1.
668          my $target_has_latin1 = $target_has_latin1;
669          my $pattern_has_latin1 = $pattern_has_latin1;
670          if (! $target_has_latin1) {
671            foreach my $cp (@target) {
672              if (exists $folds{$cp}{1}
673                  && grep { $_ < 256 } @{$folds{$cp}{1}} )
674              {
675                $target_has_latin1 = 1;
676                last;
677              }
678            }
679          }
680          if (! $pattern_has_latin1) {
681            foreach my $cp (@pattern) {
682              if (exists $folds{$cp}{1}
683                  && grep { $_ < 256 } @{$folds{$cp}{1}} )
684              {
685                $pattern_has_latin1 = 1;
686                last;
687              }
688            }
689          }
690          if (! $target_has_latin1 && ! $pattern_has_latin1) {
691            next if defined $has_tested_above_latin1_l
692                    && $has_tested_above_latin1_l != $test;
693            $has_tested_above_latin1_l = $test;
694          }
695          elsif ($target_only_ascii && $pattern_only_ascii) {
696
697              # And, except for one set just to make sure, skip tests
698              # where both elements in the pair are ASCII.  This is
699              # essentially the same reasoning as above for /aa.
700              next if defined $has_tested_ascii_l
701                      && $has_tested_ascii_l != $test;
702              $has_tested_ascii_l = $test;
703          }
704        }
705        elsif ($charset eq 'd') {
706          # Similarly for d.  Beyond one test (besides self) each, we  don't
707          # test pairs that are both ascii; or both above latin1, or are
708          # combinations of ascii and above latin1.
709          if (! $target_has_upper_latin1 && ! $pattern_has_upper_latin1) {
710            if ($target_has_ascii && $pattern_has_ascii) {
711              next if defined $has_tested_ascii_d
712                      && $has_tested_ascii_d != $test;
713              $has_tested_ascii_d = $test
714            }
715            elsif (! $target_has_latin1 && ! $pattern_has_latin1) {
716              next if defined $has_tested_above_latin1_d
717                      && $has_tested_above_latin1_d != $test;
718              $has_tested_above_latin1_d = $test;
719            }
720            else {
721              next if defined $has_tested_non_latin1_d
722                      && $has_tested_non_latin1_d != $test;
723              $has_tested_non_latin1_d = $test;
724            }
725          }
726        }
727        elsif ($charset eq 'a') {
728          # Similarly for a.  This should match identically to /u, so wasn't
729          # tested at all until a bug was found that was thereby missed.
730          # As a compromise, beyond one test (besides self) each, we  don't
731          # test pairs that are both ascii; or both above latin1, or are
732          # combinations of ascii and above latin1.
733          if (! $target_has_upper_latin1 && ! $pattern_has_upper_latin1) {
734            if ($target_has_ascii && $pattern_has_ascii) {
735              next if defined $has_tested_ascii_a
736                      && $has_tested_ascii_a != $test;
737              $has_tested_ascii_a = $test
738            }
739            elsif (! $target_has_latin1 && ! $pattern_has_latin1) {
740              next if defined $has_tested_above_latin1_a
741                      && $has_tested_above_latin1_a != $test;
742              $has_tested_above_latin1_a = $test;
743            }
744            else {
745              next if defined $has_tested_non_latin1_a
746                      && $has_tested_non_latin1_a != $test;
747              $has_tested_non_latin1_a = $test;
748            }
749          }
750        }
751      }
752
753      foreach my $utf8_target (0, 1) {    # Both utf8 and not, for
754                                          # code points < 256
755        my $upgrade_target = "";
756
757        # These must already be in utf8 because the string to match has
758        # something above latin1.  So impossible to test if to not to be in
759        # utf8; and otherwise, no upgrade is needed.
760        next if $target_above_latin1 && ! $utf8_target;
761        $upgrade_target = ' utf8::upgrade($c);' if ! $target_above_latin1 && $utf8_target;
762
763        foreach my $utf8_pattern (0, 1) {
764          next if $pattern_above_latin1 && ! $utf8_pattern;
765
766          # Our testing of 'l' uses the POSIX locale, which is ASCII-only
767          my $uni_semantics = $charset ne 'l' && (    $utf8_target
768                                                  ||  $charset eq 'u'
769                                                  ||  $charset eq 'L'
770                                                  || ($charset eq 'd' && $utf8_pattern)
771                                                  ||  $charset =~ /a/);
772          my $upgrade_pattern = "";
773          $upgrade_pattern = ' utf8::upgrade($rhs);'
774            if ! $pattern_above_latin1 && $utf8_pattern;
775
776          my $lhs = join "", @x_target;
777          my $lhs_str = eval qq{"$lhs"}; fail($@) if $@;
778          my @rhs = @x_pattern;
779          my $rhs = join "", @rhs;
780
781          # Unicode created a folding rule that partially emulates what
782          # happens in a Turkish locale, by using combining characters.  The
783          # result is close enough to what really should happen, that it can
784          # pass many of the tests, but not all.  So, if we have a rule that
785          # is expecting failure, it may pass instead.  The code in the block
786          # below is good enough for skipping the tests, and khw tried to make
787          # it general, but should the rules be revised (unlikely at this
788          # point), this might need to be tweaked.
789          if ($tests_ref == \%neg_tests) {
790            my ($shorter_ref, $longer_ref);
791
792            # Convert the $rhs to a string, like we already did for the lhs
793            my $rhs_str = eval qq{"$rhs"}; fail($@) if $@;
794
795            # If the lengths of the two sides are equal, we don't want to do
796            # this; this is only to bypass the combining characters affecting
797            # things
798            if (length $lhs_str != length $rhs_str) {
799
800              # Find the shorter and longer of the pair
801              if (length $lhs_str < length $rhs_str) {
802                  $shorter_ref = \$lhs_str;
803                  $longer_ref = \$rhs_str;
804              }
805              else {
806                  $shorter_ref = \$rhs_str;
807                  $longer_ref = \$lhs_str;
808              }
809
810              # If the shorter string is entirely contained in the longer, we
811              # have generated a test that is likely to succeed, and the
812              # reasons it would fail have nothing to do with folding.  But we
813              # are expecting it to fail, and so our test is invalid.  Skip
814              # it.
815              next if index($$longer_ref, $$shorter_ref) >= 0;
816
817
818              # The above eliminates about half the failure cases.  This gets
819              # the rest.  If the shorter string is a single character and has
820              # a fold legal in this run to a character that is in the longer
821              # string, it is also likely to succeed under /i.  So again our
822              # computed test is bogus.
823              if (   length $$shorter_ref == 1
824                  && exists $folds{ord $$shorter_ref}{1})
825              {
826                my @folded_to = @{$folds{ord $$shorter_ref}{1}};
827                next if   @folded_to == 1
828                       && index($$longer_ref, chr $folded_to[0]) >= 0;
829              }
830            }
831          }
832
833          my $should_fail = (! $uni_semantics && $ord < 256 && ! $is_self && utf8::native_to_unicode($ord) >= 128)
834                            || ($charset eq 'aa' && $target_has_ascii != $pattern_has_ascii)
835                            || ($charset eq 'l' && $target_has_latin1 != $pattern_has_latin1)
836                            || $tests_ref == \%neg_tests;
837
838          # Do simple tests of referencing capture buffers, named and
839          # numbered.
840          my $op = '=~';
841          $op = '!~' if $should_fail;
842
843          my $todo = 0;  # No longer any todo's
844          my $eval = "my \$c = \"$lhs$rhs\"; my \$rhs = \"$rhs\"; "
845                   . $upgrade_pattern
846                   . " my \$p = qr/(?$charset_mod:^(\$rhs)\\1\$)/i;"
847                   . "$upgrade_target \$c $op \$p";
848          run_test($eval, $todo, ($charset_mod eq 'l'), "");
849
850          $eval = "my \$c = \"$lhs$rhs\"; my \$rhs = \"$rhs\"; "
851                . $upgrade_pattern
852                . " my \$p = qr/(?$charset_mod:^(?<grind>\$rhs)\\k<grind>\$)/i;"
853                . "$upgrade_target \$c $op \$p";
854          run_test($eval, $todo, ($charset_mod eq 'l'), "");
855
856          if ($lhs ne $rhs) {
857            $eval = "my \$c = \"$rhs$lhs\"; my \$rhs = \"$rhs\"; "
858                  . $upgrade_pattern
859                  . " my \$p = qr/(?$charset_mod:^(\$rhs)\\1\$)/i;"
860                  . "$upgrade_target \$c $op \$p";
861            run_test($eval, "", ($charset_mod eq 'l'), "");
862
863            $eval = "my \$c = \"$rhs$lhs\"; my \$rhs = \"$rhs\"; "
864                  . $upgrade_pattern
865                  . " my \$p = qr/(?$charset_mod:^(?<grind>\$rhs)\\k<grind>\$)/i;"
866                  . "$upgrade_target \$c $op \$p";
867            run_test($eval, "", ($charset_mod eq 'l'), "");
868          }
869
870          # See if works on what could be a simple trie.
871          my $alternate;
872          {
873            # Keep the alternate | branch the same length as the tested one so
874            # that it's length doesn't influence things
875            my $evaled = eval "\"$rhs\"";   # Convert e.g. \x{foo} into its
876                                            # chr equivalent
877            use bytes;
878            $alternate = 'q' x length $evaled;
879          }
880          $eval = "my \$c = \"$lhs\"; my \$rhs = \"$rhs\"; "
881                . $upgrade_pattern
882                . " my \$p = qr/\$rhs|$alternate/i$charset_mod;"
883                . "$upgrade_target \$c $op \$p";
884          run_test($eval, "", ($charset_mod eq 'l'), "");
885
886          # Check that works when the folded character follows something that
887          # is quantified.  This test knows the regex code internals to the
888          # extent that it knows this is a potential problem, and that there
889          # are three different types of quantifiers generated: 1) The thing
890          # being quantified matches a single character; 2) it matches more
891          # than one character, but is fixed width; 3) it can match a variable
892          # number of characters.  (It doesn't know that case 3 shouldn't
893          # matter, since it doesn't do anything special for the character
894          # following the quantifier; nor that some of the different
895          # quantifiers execute the same underlying code, as these tests are
896          # quick, and this insulates these tests from changes in the
897          # implementation.)
898          for my $quantifier ('?', '??', '*', '*?', '+', '+?', '{1,2}', '{1,2}?') {
899            $eval = "my \$c = \"_$lhs\"; my \$rhs = \"$rhs\"; $upgrade_pattern "
900                  . "my \$p = qr/(?$charset_mod:.$quantifier\$rhs)/i;"
901                  . "$upgrade_target \$c $op \$p";
902            run_test($eval, "", ($charset_mod eq 'l'), "");
903            $eval = "my \$c = \"__$lhs\"; my \$rhs = \"$rhs\"; $upgrade_pattern "
904                  . "my \$p = qr/(?$charset_mod:(?:..)$quantifier\$rhs)/i;"
905                  . "$upgrade_target \$c $op \$p";
906            run_test($eval, "", ($charset_mod eq 'l'), "");
907            $eval = "my \$c = \"__$lhs\"; my \$rhs = \"$rhs\"; $upgrade_pattern "
908                  . "my \$p = qr/(?$charset_mod:(?:.|\\R)$quantifier\$rhs)/i;"
909                  . "$upgrade_target \$c $op \$p";
910            run_test($eval, "", ($charset_mod eq 'l'), "");
911          }
912
913          foreach my $bracketed (0, 1) {   # Put rhs in [...], or not
914            next if $bracketed && @pattern != 1;    # bracketed makes these
915                                                    # or's instead of a sequence
916            foreach my $optimize_bracketed (0, 1) {
917              next if $optimize_bracketed && ! $bracketed;
918              foreach my $inverted (0,1) {
919                  next if $inverted && ! $bracketed;  # inversion only valid
920                                                      # in [^...]
921                  next if $inverted && @target != 1;  # [perl #89750] multi-char
922                                                      # not valid in [^...]
923
924                # In some cases, add an extra character that doesn't fold, and
925                # looks ok in the output.
926                my $extra_char = "_";
927                foreach my $prepend ("", $extra_char) {
928                  foreach my $append ("", $extra_char) {
929
930                    # Assemble the rhs.  Put each character in a separate
931                    # bracketed if using charclasses.  This creates a stress on
932                    # the code to span a match across multiple elements
933                    my $rhs = "";
934                    foreach my $rhs_char (@rhs) {
935                        $rhs .= '[' if $bracketed;
936                        $rhs .= '^' if $inverted;
937                        $rhs .=  $rhs_char;
938
939                        # Add a character to the class, so class doesn't get
940                        # optimized out, unless we are testing that optimization
941                        $rhs .= '_' if $optimize_bracketed;
942                        $rhs .= ']' if $bracketed;
943                    }
944
945                    # Add one of: no capturing parens
946                    #             a single set
947                    #             a nested set
948                    # Use quantifiers and extra variable width matches inside
949                    # them to keep some optimizations from happening
950                    foreach my $parend (0, 1, 2) {
951                      my $interior = (! $parend)
952                                      ? $rhs
953                                      : ($parend == 1)
954                                          ? "(${rhs},?)"
955                                          : "((${rhs})+,?)";
956                      foreach my $quantifier ("", '?', '*', '+', '{1,3}') {
957
958                        # Perhaps should be TODOs, as are unimplemented, but
959                        # maybe will never be implemented
960                        next if @pattern != 1 && $quantifier;
961
962                        # A ? or * quantifier normally causes the thing to be
963                        # able to match a null string
964                        my $quantifier_can_match_null = $quantifier eq '?'
965                                                     || $quantifier eq '*';
966
967                        # But since we only quantify the last character in a
968                        # multiple fold, the other characters will have width,
969                        # except if we are quantifying the whole rhs
970                        my $can_match_null = $quantifier_can_match_null
971                                             && (@rhs == 1 || $parend);
972
973                        foreach my $l_anchor ("", '^') { # '\A' didn't change
974                                                         # result)
975                          foreach my $r_anchor ("", '$') { # '\Z', '\z' didn't
976                                                           # change result)
977                            # The folded part can match the null string if it
978                            # isn't required to have width, and there's not
979                            # something on one or both sides that force it to.
980                            my $both_sides = ($l_anchor && $r_anchor)
981                                              || ($l_anchor && $append)
982                                              || ($r_anchor && $prepend)
983                                              || ($prepend && $append);
984                            my $must_match = ! $can_match_null || $both_sides;
985                            # for performance, but doing this missed many failures
986                            #next unless $must_match;
987                            my $quantified = "(?$charset_mod:$l_anchor$prepend"
988                                           . "$interior${quantifier}$append$r_anchor)";
989                            my $op;
990                            if ($must_match && $should_fail)  {
991                                $op = 0;
992                            } else {
993                                $op = 1;
994                            }
995                            $op = ! $op if $must_match && $inverted;
996
997                            if ($inverted && @target > 1) {
998                              # When doing an inverted match against a
999                              # multi-char target, and there is not something on
1000                              # the left to anchor the match, if it shouldn't
1001                              # succeed, skip, as what will happen (when working
1002                              # correctly) is that it will match the first
1003                              # position correctly, and then be inverted to not
1004                              # match; then it will go to the second position
1005                              # where it won't match, but get inverted to match,
1006                              # and hence succeeding.
1007                              next if ! ($l_anchor || $prepend) && ! $op;
1008
1009                              # Can't ever match for latin1 code points non-uni
1010                              # semantics that have a inverted multi-char fold
1011                              # when there is something on both sides and the
1012                              # quantifier isn't such as to span the required
1013                              # width, which is 2 or 3.
1014                              $op = 0 if $ord < 255
1015                                        && ! $uni_semantics
1016                                        && $both_sides
1017                                        && ( ! $quantifier || $quantifier eq '?')
1018                                        && $parend < 2;
1019
1020                              # Similarly can't ever match when inverting a
1021                              # multi-char fold for /aa and the quantifier
1022                              # isn't sufficient to allow it to span to both
1023                              # sides.
1024                              $op = 0 if $target_has_ascii
1025                                         && $charset eq 'aa'
1026                                         && $both_sides
1027                                         && ( ! $quantifier || $quantifier eq '?')
1028                                         && $parend < 2;
1029
1030                              # Or for /l
1031                              $op = 0 if $target_has_latin1 && $charset eq 'l'
1032                                      && $both_sides
1033                                      && ( ! $quantifier || $quantifier eq '?')
1034                                      && $parend < 2;
1035                            }
1036
1037
1038                            my $desc = "";
1039                            if ($charset_mod eq 'l') {
1040                                $desc .= 'setlocale(LC_CTYPE, "'
1041                                        . POSIX::setlocale(&POSIX::LC_CTYPE)
1042                                        . '"); '
1043                            }
1044                            $desc .= "my \$c = \"$prepend$lhs$append\"; "
1045                                    . "my \$rhs = \"\"; $upgrade_pattern"
1046                                    . "my \$p = qr/$quantified\$rhs/i;"
1047                                    . "$upgrade_target "
1048                                    . "\$c " . ($op ? "=~" : "!~") . " \$p; ";
1049                            if ($DEBUG) {
1050                              $desc .= (
1051                              "; uni_semantics=$uni_semantics, "
1052                              . "should_fail=$should_fail, "
1053                              . "bracketed=$bracketed, "
1054                              . "prepend=$prepend, "
1055                              . "append=$append, "
1056                              . "parend=$parend, "
1057                              . "quantifier=$quantifier, "
1058                              . "l_anchor=$l_anchor, "
1059                              . "r_anchor=$r_anchor; "
1060                              . "pattern_above_latin1=$pattern_above_latin1; "
1061                              . "utf8_pattern=$utf8_pattern"
1062                              );
1063                            }
1064
1065                            my $c = "$prepend$lhs_str$append";
1066                            my $p = "$quantified"; # string copy deliberate
1067                            utf8::upgrade($c) if length($upgrade_target);
1068                            utf8::upgrade($p) if length($upgrade_pattern);
1069                            $p = qr/$p/i;
1070                            my $res = $op ? ($c =~ $p): ($c !~ $p);
1071
1072                            if (!$res || $list_all_tests) {
1073                              # Failed or debug; output the result
1074                              $count++;
1075                              ok($res, "test $count - $desc");
1076                            } else {
1077                              # Just count the test as passed
1078                              $okays++;
1079                            }
1080                            $this_iteration++;
1081                          }
1082                        }
1083                      }
1084                    }
1085                  }
1086                }
1087              }
1088            }
1089          }
1090        }
1091      }
1092      unless($list_all_tests) {
1093        $count++;
1094        is $okays, $this_iteration, "$okays subtests ok for"
1095          . " /$charset_mod"
1096          . (($charset_mod eq 'l') ? " ($current_locale)" : "")
1097          . ', target="' . join("", @x_target) . '",'
1098          . ' pat="' . join("", @x_pattern) . '"';
1099      }
1100    }
1101  }
1102}
1103}
1104
1105plan($count);
1106
11071
1108