This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
ext/XS-APItest/t/utf8_warn_base.pl: Fix comment typos
[perl5.git] / ext / XS-APItest / t / utf8_warn_base.pl
1 #!perl -w
2
3 # This is a base file to be used by various .t's in its directory
4 # It tests various malformed UTF-8 sequences and some code points that are
5 # "problematic", and verifies that the correct warnings/flags etc are
6 # generated when using them.  For the code points, it also takes the UTF-8 and
7 # perturbs it to be malformed in various ways, and tests that this gets
8 # appropriately detected.
9
10 use strict;
11 use Test::More;
12
13 BEGIN {
14     use_ok('XS::APItest');
15     require 'charset_tools.pl';
16     require './t/utf8_setup.pl';
17 };
18
19 $|=1;
20
21 use XS::APItest;
22
23 my @warnings_gotten;
24
25 use warnings 'utf8';
26 local $SIG{__WARN__} = sub { my @copy = @_;
27                              push @warnings_gotten, map { chomp; $_ } @copy;
28                            };
29
30 my $highest_non_extended_utf8_cp = (isASCII) ? 0x7FFFFFFF : 0x3FFFFFFF;
31 my $native_lowest_continuation_chr = I8_to_native(chr $::lowest_continuation);
32
33 # C5 is chosen as it is valid for both ASCII and EBCDIC platforms
34 my $known_start_byte = I8_to_native("\xC5");
35
36 sub requires_extended_utf8($) {
37
38     # Returns a boolean as to whether or not the code point parameter fits
39     # into 31 bits, subject to the convention that a negative code point
40     # stands for one that overflows the word size, so won't fit in 31 bits.
41
42     return shift > $highest_non_extended_utf8_cp;
43 }
44
45 sub is_extended_utf8($) {
46
47     # Returns a boolean as to whether or not the input UTF-8 sequence uses
48     # Perl extended UTF-8.
49
50     my $byte = substr(shift, 0, 1);
51     return ord $byte >= 0xFE if isASCII;
52     return $byte == I8_to_native("\xFF");
53 }
54
55 sub overflow_discern_len($) {
56
57     # Returns how many bytes are needed to tell if a non-overlong UTF-8
58     # sequence is for a code point that won't fit in the platform's word size.
59     # Only the length of the sequence representing a single code point is
60     # needed.
61
62     if (isASCII) {
63         return ($::is64bit) ? 3 : 1;
64
65         # Below is needed for code points above IV_MAX
66         #return ($::is64bit) ? 3 : ((shift == $::max_bytes)
67         #                           ? 1
68         #                           : 2);
69     }
70
71     return ($::is64bit) ? 2 : 8;
72 }
73
74 sub overlong_discern_len($) {
75
76     # Returns how many bytes are needed to tell if the input UTF-8 sequence
77     # for a code point is overlong
78
79     my $string = shift;
80     my $length = length $string;
81     my $byte = ord native_to_I8(substr($string, 0, 1));
82     if (isASCII) {
83         return ($byte >= 0xFE)
84                 ? ((! $::is64bit)
85                     ? 1
86                     : ($byte == 0xFF) ? 7 : 2)
87                 : (($length == 2) ? 1 : 2);
88         # Below is needed for code points above IV_MAX
89         #return ($length == $::max_bytes)
90         #          # This is constrained to 1 on 32-bit machines, as it
91         #          # overflows there
92         #        ? (($::is64bit) ? 7 : 1)
93         #        : (($length == 2) ? 1 : 2);
94     }
95
96     return ($length == $::max_bytes) ? 8 : (($length <= 3) ? 1 : 2);
97 }
98
99 my @tests;
100 {
101     no warnings qw(portable overflow);
102     @tests = (
103         # $testname,
104         # $bytes,                  UTF-8 string
105         # $allowed_uv,             code point $bytes evaluates to; -1 if
106         #                          overflows
107         # $needed_to_discern_len   optional, how long an initial substring do
108         #                          we need to tell that the string must be for
109         #                          a code point in the category it falls in,
110         #                          like being a surrogate; 0 indicates we need
111         #                          the whole string.  Some categories have a
112         #                          default that is used if this is omitted.
113         [ "orphan continuation byte malformation",
114             I8_to_native("$::I8c"),
115             0xFFFD,
116             1,
117         ],
118         [ "overlong malformation, lowest 2-byte",
119             (isASCII) ? "\xc0\x80" : I8_to_native("\xc0\xa0"),
120             0,   # NUL
121         ],
122         [ "overlong malformation, highest 2-byte",
123             (isASCII) ? "\xc1\xbf" : I8_to_native("\xc4\xbf"),
124             (isASCII) ? 0x7F : 0xFF,
125         ],
126         [ "overlong malformation, lowest 3-byte",
127             (isASCII) ? "\xe0\x80\x80" : I8_to_native("\xe0\xa0\xa0"),
128             0,   # NUL
129         ],
130         [ "overlong malformation, highest 3-byte",
131             (isASCII) ? "\xe0\x9f\xbf" : I8_to_native("\xe0\xbf\xbf"),
132             (isASCII) ? 0x7FF : 0x3FF,
133         ],
134         [ "lowest surrogate",
135             (isASCII) ? "\xed\xa0\x80" : I8_to_native("\xf1\xb6\xa0\xa0"),
136             0xD800,
137         ],
138         [ "a middle surrogate",
139             (isASCII) ? "\xed\xa4\x8d" : I8_to_native("\xf1\xb6\xa8\xad"),
140             0xD90D,
141         ],
142         [ "highest surrogate",
143             (isASCII) ? "\xed\xbf\xbf" : I8_to_native("\xf1\xb7\xbf\xbf"),
144             0xDFFF,
145         ],
146         [ "first of 32 consecutive non-character code points",
147             (isASCII) ? "\xef\xb7\x90" : I8_to_native("\xf1\xbf\xae\xb0"),
148             0xFDD0,
149         ],
150         [ "a mid non-character code point of the 32 consecutive ones",
151             (isASCII) ? "\xef\xb7\xa0" : I8_to_native("\xf1\xbf\xaf\xa0"),
152             0xFDE0,
153         ],
154         [ "final of 32 consecutive non-character code points",
155             (isASCII) ? "\xef\xb7\xaf" : I8_to_native("\xf1\xbf\xaf\xaf"),
156             0xFDEF,
157         ],
158         [ "non-character code point U+FFFE",
159             (isASCII) ? "\xef\xbf\xbe" : I8_to_native("\xf1\xbf\xbf\xbe"),
160             0xFFFE,
161         ],
162         [ "non-character code point U+FFFF",
163             (isASCII) ? "\xef\xbf\xbf" : I8_to_native("\xf1\xbf\xbf\xbf"),
164             0xFFFF,
165         ],
166         [ "overlong malformation, lowest 4-byte",
167             (isASCII) ? "\xf0\x80\x80\x80" : I8_to_native("\xf0\xa0\xa0\xa0"),
168             0,   # NUL
169         ],
170         [ "overlong malformation, highest 4-byte",
171             (isASCII) ? "\xf0\x8F\xbf\xbf" : I8_to_native("\xf0\xaf\xbf\xbf"),
172             (isASCII) ? 0xFFFF : 0x3FFF,
173         ],
174         [ "non-character code point U+1FFFE",
175             (isASCII) ? "\xf0\x9f\xbf\xbe" : I8_to_native("\xf3\xbf\xbf\xbe"),
176             0x1FFFE,
177         ],
178         [ "non-character code point U+1FFFF",
179             (isASCII) ? "\xf0\x9f\xbf\xbf" : I8_to_native("\xf3\xbf\xbf\xbf"),
180             0x1FFFF,
181         ],
182         [ "non-character code point U+2FFFE",
183             (isASCII) ? "\xf0\xaf\xbf\xbe" : I8_to_native("\xf5\xbf\xbf\xbe"),
184             0x2FFFE,
185         ],
186         [ "non-character code point U+2FFFF",
187             (isASCII) ? "\xf0\xaf\xbf\xbf" : I8_to_native("\xf5\xbf\xbf\xbf"),
188             0x2FFFF,
189         ],
190         [ "non-character code point U+3FFFE",
191             (isASCII) ? "\xf0\xbf\xbf\xbe" : I8_to_native("\xf7\xbf\xbf\xbe"),
192             0x3FFFE,
193         ],
194         [ "non-character code point U+3FFFF",
195             (isASCII) ? "\xf0\xbf\xbf\xbf" : I8_to_native("\xf7\xbf\xbf\xbf"),
196             0x3FFFF,
197         ],
198         [ "non-character code point U+4FFFE",
199             (isASCII)
200             ?               "\xf1\x8f\xbf\xbe"
201             : I8_to_native("\xf8\xa9\xbf\xbf\xbe"),
202             0x4FFFE,
203         ],
204         [ "non-character code point U+4FFFF",
205             (isASCII)
206             ?               "\xf1\x8f\xbf\xbf"
207             : I8_to_native("\xf8\xa9\xbf\xbf\xbf"),
208             0x4FFFF,
209         ],
210         [ "non-character code point U+5FFFE",
211             (isASCII)
212             ?              "\xf1\x9f\xbf\xbe"
213             : I8_to_native("\xf8\xab\xbf\xbf\xbe"),
214             0x5FFFE,
215         ],
216         [ "non-character code point U+5FFFF",
217             (isASCII)
218             ?              "\xf1\x9f\xbf\xbf"
219             : I8_to_native("\xf8\xab\xbf\xbf\xbf"),
220             0x5FFFF,
221         ],
222         [ "non-character code point U+6FFFE",
223             (isASCII)
224             ?              "\xf1\xaf\xbf\xbe"
225             : I8_to_native("\xf8\xad\xbf\xbf\xbe"),
226             0x6FFFE,
227         ],
228         [ "non-character code point U+6FFFF",
229             (isASCII)
230             ?              "\xf1\xaf\xbf\xbf"
231             : I8_to_native("\xf8\xad\xbf\xbf\xbf"),
232             0x6FFFF,
233         ],
234         [ "non-character code point U+7FFFE",
235             (isASCII)
236             ?              "\xf1\xbf\xbf\xbe"
237             : I8_to_native("\xf8\xaf\xbf\xbf\xbe"),
238             0x7FFFE,
239         ],
240         [ "non-character code point U+7FFFF",
241             (isASCII)
242             ?              "\xf1\xbf\xbf\xbf"
243             : I8_to_native("\xf8\xaf\xbf\xbf\xbf"),
244             0x7FFFF,
245         ],
246         [ "non-character code point U+8FFFE",
247             (isASCII)
248             ?              "\xf2\x8f\xbf\xbe"
249             : I8_to_native("\xf8\xb1\xbf\xbf\xbe"),
250             0x8FFFE,
251         ],
252         [ "non-character code point U+8FFFF",
253             (isASCII)
254             ?              "\xf2\x8f\xbf\xbf"
255             : I8_to_native("\xf8\xb1\xbf\xbf\xbf"),
256             0x8FFFF,
257         ],
258         [ "non-character code point U+9FFFE",
259             (isASCII)
260             ?              "\xf2\x9f\xbf\xbe"
261             : I8_to_native("\xf8\xb3\xbf\xbf\xbe"),
262             0x9FFFE,
263         ],
264         [ "non-character code point U+9FFFF",
265             (isASCII)
266             ?              "\xf2\x9f\xbf\xbf"
267             : I8_to_native("\xf8\xb3\xbf\xbf\xbf"),
268             0x9FFFF,
269         ],
270         [ "non-character code point U+AFFFE",
271             (isASCII)
272             ?              "\xf2\xaf\xbf\xbe"
273             : I8_to_native("\xf8\xb5\xbf\xbf\xbe"),
274             0xAFFFE,
275         ],
276         [ "non-character code point U+AFFFF",
277             (isASCII)
278             ?              "\xf2\xaf\xbf\xbf"
279             : I8_to_native("\xf8\xb5\xbf\xbf\xbf"),
280             0xAFFFF,
281         ],
282         [ "non-character code point U+BFFFE",
283             (isASCII)
284             ?              "\xf2\xbf\xbf\xbe"
285             : I8_to_native("\xf8\xb7\xbf\xbf\xbe"),
286             0xBFFFE,
287         ],
288         [ "non-character code point U+BFFFF",
289             (isASCII)
290             ?              "\xf2\xbf\xbf\xbf"
291             : I8_to_native("\xf8\xb7\xbf\xbf\xbf"),
292             0xBFFFF,
293         ],
294         [ "non-character code point U+CFFFE",
295             (isASCII)
296             ?              "\xf3\x8f\xbf\xbe"
297             : I8_to_native("\xf8\xb9\xbf\xbf\xbe"),
298             0xCFFFE,
299         ],
300         [ "non-character code point U+CFFFF",
301             (isASCII)
302             ?              "\xf3\x8f\xbf\xbf"
303             : I8_to_native("\xf8\xb9\xbf\xbf\xbf"),
304             0xCFFFF,
305         ],
306         [ "non-character code point U+DFFFE",
307             (isASCII)
308             ?              "\xf3\x9f\xbf\xbe"
309             : I8_to_native("\xf8\xbb\xbf\xbf\xbe"),
310             0xDFFFE,
311         ],
312         [ "non-character code point U+DFFFF",
313             (isASCII)
314             ?              "\xf3\x9f\xbf\xbf"
315             : I8_to_native("\xf8\xbb\xbf\xbf\xbf"),
316             0xDFFFF,
317         ],
318         [ "non-character code point U+EFFFE",
319             (isASCII)
320             ?              "\xf3\xaf\xbf\xbe"
321             : I8_to_native("\xf8\xbd\xbf\xbf\xbe"),
322             0xEFFFE,
323         ],
324         [ "non-character code point U+EFFFF",
325             (isASCII)
326             ?              "\xf3\xaf\xbf\xbf"
327             : I8_to_native("\xf8\xbd\xbf\xbf\xbf"),
328             0xEFFFF,
329         ],
330         [ "non-character code point U+FFFFE",
331             (isASCII)
332             ?              "\xf3\xbf\xbf\xbe"
333             : I8_to_native("\xf8\xbf\xbf\xbf\xbe"),
334             0xFFFFE,
335         ],
336         [ "non-character code point U+FFFFF",
337             (isASCII)
338             ?              "\xf3\xbf\xbf\xbf"
339             : I8_to_native("\xf8\xbf\xbf\xbf\xbf"),
340             0xFFFFF,
341         ],
342         [ "non-character code point U+10FFFE",
343             (isASCII)
344             ?              "\xf4\x8f\xbf\xbe"
345             : I8_to_native("\xf9\xa1\xbf\xbf\xbe"),
346             0x10FFFE,
347         ],
348         [ "non-character code point U+10FFFF",
349             (isASCII)
350             ?              "\xf4\x8f\xbf\xbf"
351             : I8_to_native("\xf9\xa1\xbf\xbf\xbf"),
352             0x10FFFF,
353         ],
354         [ "first non_unicode",
355             (isASCII)
356             ?              "\xf4\x90\x80\x80"
357             : I8_to_native("\xf9\xa2\xa0\xa0\xa0"),
358             0x110000,
359             2,
360         ],
361         [ "non_unicode whose first byte tells that",
362             (isASCII)
363             ?              "\xf5\x80\x80\x80"
364             : I8_to_native("\xfa\xa0\xa0\xa0\xa0"),
365             (isASCII) ? 0x140000 : 0x200000,
366             1,
367         ],
368         [ "overlong malformation, lowest 5-byte",
369             (isASCII)
370             ?              "\xf8\x80\x80\x80\x80"
371             : I8_to_native("\xf8\xa0\xa0\xa0\xa0"),
372             0,   # NUL
373         ],
374         [ "overlong malformation, highest 5-byte",
375             (isASCII)
376             ?              "\xf8\x87\xbf\xbf\xbf"
377             : I8_to_native("\xf8\xa7\xbf\xbf\xbf"),
378             (isASCII) ? 0x1FFFFF : 0x3FFFF,
379         ],
380         [ "overlong malformation, lowest 6-byte",
381             (isASCII)
382             ?              "\xfc\x80\x80\x80\x80\x80"
383             : I8_to_native("\xfc\xa0\xa0\xa0\xa0\xa0"),
384             0,   # NUL
385         ],
386         [ "overlong malformation, highest 6-byte",
387             (isASCII)
388             ?              "\xfc\x83\xbf\xbf\xbf\xbf"
389             : I8_to_native("\xfc\xa3\xbf\xbf\xbf\xbf"),
390             (isASCII) ? 0x3FFFFFF : 0x3FFFFF,
391         ],
392         [ "overlong malformation, lowest 7-byte",
393             (isASCII)
394             ?              "\xfe\x80\x80\x80\x80\x80\x80"
395             : I8_to_native("\xfe\xa0\xa0\xa0\xa0\xa0\xa0"),
396             0,   # NUL
397         ],
398         [ "overlong malformation, highest 7-byte",
399             (isASCII)
400             ?              "\xfe\x81\xbf\xbf\xbf\xbf\xbf"
401             : I8_to_native("\xfe\xa1\xbf\xbf\xbf\xbf\xbf"),
402             (isASCII) ? 0x7FFFFFFF : 0x3FFFFFF,
403         ],
404         [ "highest 31 bit code point",
405             (isASCII)
406             ?  "\xfd\xbf\xbf\xbf\xbf\xbf"
407             : I8_to_native(
408                "\xff\xa0\xa0\xa0\xa0\xa0\xa0\xa1\xbf\xbf\xbf\xbf\xbf\xbf"),
409             0x7FFFFFFF,
410             1,
411         ],
412         [ "lowest 32 bit code point",
413             (isASCII)
414             ?  "\xfe\x82\x80\x80\x80\x80\x80"
415             : I8_to_native(
416                 "\xff\xa0\xa0\xa0\xa0\xa0\xa0\xa2\xa0\xa0\xa0\xa0\xa0\xa0"),
417             ($::is64bit) ? 0x80000000 : -1,   # Overflows on 32-bit systems
418             1,
419         ],
420         # Used when UV_MAX is allowed as a code point
421         #[ "highest 32 bit code point",
422         #    (isASCII)
423         #    ?  "\xfe\x83\xbf\xbf\xbf\xbf\xbf"
424         #    : I8_to_native(
425         #       "\xff\xa0\xa0\xa0\xa0\xa0\xa0\xa3\xbf\xbf\xbf\xbf\xbf\xbf"),
426         #    0xFFFFFFFF,
427         #],
428         #[ "Lowest 33 bit code point",
429         #    (isASCII)
430         #    ?  "\xfe\x84\x80\x80\x80\x80\x80"
431         #    : I8_to_native(
432         #        "\xff\xa0\xa0\xa0\xa0\xa0\xa0\xa4\xa0\xa0\xa0\xa0\xa0\xa0"),
433         #    ($::is64bit) ? 0x100000000 : 0x0,   # Overflows on 32-bit systems
434         #],
435     );
436
437     if (! $::is64bit) {
438         if (isASCII) {
439             push @tests,
440                 [ "overlong malformation, but naively looks like overflow",
441                     "\xff\x80\x80\x80\x80\x80\x80\x81\xbf\xbf\xbf\xbf\xbf",
442                     0x7FFFFFFF,
443                 ],
444                 # Used when above IV_MAX are allowed.
445                 #[ "overlong malformation, but naively looks like overflow",
446                 #    "\xff\x80\x80\x80\x80\x80\x80\x83\xbf\xbf\xbf\xbf\xbf",
447                 #    0xFFFFFFFF,
448                 #],
449                 [ "overflow that old algorithm failed to detect",
450                     "\xfe\x86\x80\x80\x80\x80\x80",
451                     -1,
452                 ];
453         }
454     }
455
456     push @tests,
457         [ "overlong malformation, lowest max-byte",
458             (isASCII)
459              ?      "\xff\x80\x80\x80\x80\x80\x80\x80\x80\x80\x80\x80\x80"
460              : I8_to_native(
461                     "\xff\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
462             0,   # NUL
463         ],
464         [ "overlong malformation, highest max-byte",
465             (isASCII)    # 2**36-1 on ASCII; 2**30-1 on EBCDIC
466              ?      "\xff\x80\x80\x80\x80\x80\x80\xbf\xbf\xbf\xbf\xbf\xbf"
467              : I8_to_native(
468                     "\xff\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xbf\xbf\xbf\xbf\xbf\xbf"),
469             (isASCII) ? (($::is64bit) ? 0xFFFFFFFFF : -1) : 0x3FFFFFFF,
470         ];
471
472     if (isASCII) {
473         push @tests,
474             [ "Lowest code point requiring 13 bytes to represent", # 2**36
475                 "\xff\x80\x80\x80\x80\x80\x81\x80\x80\x80\x80\x80\x80",
476                 ($::is64bit) ? 0x1000000000 : -1,    # overflows on 32bit
477             ],
478     };
479
480     if ($::is64bit) {
481         push @tests,
482             [ "highest 63 bit code point",
483               (isASCII)
484               ? "\xff\x80\x87\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf"
485               : I8_to_native(
486                 "\xff\xa7\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf"),
487               0x7FFFFFFFFFFFFFFF,
488               (isASCII) ? 1 : 2,
489             ],
490             [ "first 64 bit code point",
491               (isASCII)
492               ? "\xff\x80\x88\x80\x80\x80\x80\x80\x80\x80\x80\x80\x80"
493               : I8_to_native(
494                 "\xff\xa8\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
495               -1,
496             ];
497             # Used when UV_MAX is allowed as a code point
498             #[ "highest 64 bit code point",
499             #  (isASCII)
500             #  ? "\xff\x80\x8f\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf"
501             #  : I8_to_native(
502             #    "\xff\xaf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf"),
503             #  0xFFFFFFFFFFFFFFFF,
504             #  (isASCII) ? 1 : 2,
505             #],
506             #[ "first 65 bit code point",
507             #  (isASCII)
508             #  ? "\xff\x80\x9f\x80\x80\x80\x80\x80\x80\x80\x80\x80\x80"
509             #  : I8_to_native(
510             #    "\xff\xb0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
511             #  0,
512             #];
513         if (isASCII) {
514             push @tests,
515                 [ "overflow that old algorithm failed to detect",
516                     "\xff\x80\x90\x90\x90\xbf\xbf\xbf\xbf\xbf\xbf\xbf\xbf",
517                     -1,
518                 ];
519         }
520         else {
521             push @tests,    # These could falsely show wrongly in a naive
522                             # implementation
523                 [ "requires at least 32 bits",
524                     I8_to_native(
525                     "\xff\xa0\xa0\xa0\xa0\xa0\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
526                     0x800000000,
527                       40000000
528                 ],
529                 [ "requires at least 32 bits",
530                     I8_to_native(
531                     "\xff\xa0\xa0\xa0\xa0\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
532                     0x10000000000,
533                 ],
534                 [ "requires at least 32 bits",
535                     I8_to_native(
536                     "\xff\xa0\xa0\xa0\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
537                     0x200000000000,
538                 ],
539                 [ "requires at least 32 bits",
540                     I8_to_native(
541                     "\xff\xa0\xa0\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
542                     0x4000000000000,
543                 ],
544                 [ "requires at least 32 bits",
545                     I8_to_native(
546                     "\xff\xa0\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
547                     0x80000000000000,
548                 ],
549                 [ "requires at least 32 bits",
550                     I8_to_native(
551                     "\xff\xa1\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0\xa0"),
552                     0x1000000000000000,
553                 ];
554         }
555     }
556 }
557
558 sub flags_to_text($$)
559 {
560     my ($flags, $flags_to_text_ref) = @_;
561
562     # Returns a string containing a mnemonic representation of the bits that
563     # are set in the $flags.  These are assumed to be flag bits.  The return
564     # looks like "FOO|BAR|BAZ".  The second parameter is a reference to an
565     # array that gives the textual representation of all the possible flags.
566     # Element 0 is the text for the bit 0 flag; element 1 for bit 1; ....  If
567     # no bits at all are set the string "0" is returned;
568
569     my @flag_text;
570     my $shift = 0;
571
572     return "0" if $flags == 0;
573
574     while ($flags) {
575         #diag sprintf "%x", $flags;
576         if ($flags & 1) {
577             push @flag_text, $flags_to_text_ref->[$shift];
578         }
579         $shift++;
580         $flags >>= 1;
581     }
582
583     return join "|", @flag_text;
584 }
585
586 # Possible flag returns from utf8n_to_uvchr_error().  These should have G_,
587 # instead of A_, D_, but the prefixes will be used in a a later commit, so
588 # minimize churn by having them here.
589 my @utf8n_flags_to_text =  ( qw(
590         A_EMPTY
591         A_CONTINUATION
592         A_NON_CONTINUATION
593         A_SHORT
594         A_LONG
595         A_LONG_AND_ITS_VALUE
596         PLACEHOLDER
597         A_OVERFLOW
598         D_SURROGATE
599         W_SURROGATE
600         D_NONCHAR
601         W_NONCHAR
602         D_SUPER
603         W_SUPER
604         D_PERL_EXTENDED
605         W_PERL_EXTENDED
606         CHECK_ONLY
607         NO_CONFIDENCE_IN_CURLEN_
608     ) );
609
610 sub utf8n_display_call($)
611 {
612     # Converts an eval string that calls test_utf8n_to_uvchr into a more human
613     # readable form, and returns it.  Doesn't work if the byte string contains
614     # an apostrophe.  The return will look something like:
615     #   test_utf8n_to_uvchr_error('$bytes', $length, $flags)
616     #diag $_[0];
617
618     $_[0] =~ / ^ ( [^(]* \( ) ' ( [^']*? ) ' ( .+ , \D* ) ( \d+ ) \) $ /x;
619     my $text1 = $1;     # Everything before the byte string
620     my $bytes = $2;
621     my $text2 = $3;     # Includes the length
622     my $flags = $4;
623
624     return $text1
625          . display_bytes($bytes)
626          . $text2
627          . flags_to_text($flags, \@utf8n_flags_to_text)
628          . ')';
629 }
630
631 my @uvchr_flags_to_text =  ( qw(
632         W_SURROGATE
633         W_NONCHAR
634         W_SUPER
635         W_PERL_EXTENDED
636         D_SURROGATE
637         D_NONCHAR
638         D_SUPER
639         D_PERL_EXTENDED
640 ) );
641
642 sub uvchr_display_call($)
643 {
644     # Converts an eval string that calls test_uvchr_to_utf8 into a more human
645     # readable form, and returns it.  The return will look something like:
646     #   test_uvchr_to_utf8n_flags($uv, $flags)
647     #diag $_[0];
648
649
650     $_[0] =~ / ^ ( [^(]* \( ) ( \d+ ) , \s* ( \d+ ) \) $ /x;
651     my $text = $1;
652     my $cp = sprintf "%X", $2;
653     my $flags = $3;
654
655     return "${text}0x$cp, " . flags_to_text($flags, \@uvchr_flags_to_text) . ')';
656 }
657
658 sub do_warnings_test(@)
659 {
660     my @expected_warnings = @_;
661
662     # Compares the input expected warnings array with @warnings_gotten,
663     # generating a pass for each found, removing it from @warnings_gotten.
664     # Any discrepancies generate test failures.  Returns TRUE if no
665     # discrepcancies; otherwise FALSE.
666
667     my $succeeded = 1;
668
669     if (@expected_warnings == 0) {
670         if (! is(@warnings_gotten, 0, "    Expected and got no warnings")) {
671             output_warnings(@warnings_gotten);
672             $succeeded = 0;
673         }
674         return $succeeded;
675     }
676
677     # Check that we got all the expected warnings,
678     # removing each one found
679   WARNING:
680     foreach my $expected (@expected_warnings) {
681         foreach (my $i = 0; $i < @warnings_gotten; $i++) {
682             if ($warnings_gotten[$i] =~ $expected) {
683                 pass("    Expected and got warning: "
684                     . " $warnings_gotten[$i]");
685                 splice @warnings_gotten, $i, 1;
686                 next WARNING;
687             }
688         }
689         fail("    Expected a warning that matches "
690             . $expected . " but didn't get it");
691         $succeeded = 0;
692     }
693
694     if (! is(@warnings_gotten, 0, "    Got no unexpected warnings")) {
695         output_warnings(@warnings_gotten);
696         $succeeded = 0;
697     }
698
699     return $succeeded;
700 }
701
702 my $min_cont = (isASCII) ? 0x80 : 0xA0;
703 my $continuation_shift = (isASCII) ? 6 : 5;
704 my $continuation_mask = (1 << $continuation_shift) - 1;
705
706 sub isUTF8_CHAR($$) {   # Uses first principals to determine if this is legal
707                         # (Doesn't work if overflows)
708     my ($string, $length) = @_;
709
710     # Uses first principals to calculate if $string is legal
711
712     return 0 if $length <= 0;
713
714     my $first = ord substr($string, 0, 1);
715
716     # Invariant
717     return 1 if $length == 1 && $first < $min_cont;
718
719     return 0 if $first < 0xC0;  # Starts with continuation
720
721     # Calculate the number of leading 1 bits
722     my $utf8skip = 0;
723     my $bits = $first;
724     do {
725         $utf8skip++;
726         $bits = ($bits << 1) & 0xFF;
727     } while ($bits & 0x80);
728
729     return 0 if $utf8skip != $length;
730
731     # Accumulate the $code point.  The remaining bits in the start byte count
732     # towards it
733     my $cp = $bits >> $utf8skip;
734
735     for my $i (1 .. $length - 1) {
736         my $ord = ord substr($string, $i, 1);
737
738         # Wrong if not a continuation
739         return 0 if $ord < $min_cont || $ord >= 0xC0;
740
741         $cp = ($cp << $continuation_shift)
742             | ($ord & $continuation_mask);
743     }
744
745     # If the calculated value can be expressed in fewer bytes than were passed
746     # in, is an illegal overlong.  XXX if 'chr' is not working properly, this
747     # may not be right
748     my $chr = chr $cp;
749     utf8::upgrade($chr);
750
751     use bytes;
752     return 0 if length $chr < $length;
753
754     return 1;
755 }
756
757 sub start_mark($) {
758     my $len = shift;
759     return 0xFF if $len >  7;
760     return (0xFF & (0xFE << (7 - $len)));
761 }
762
763 sub start_mask($) {
764     my $len = shift;
765     return 0 if $len >  7;
766     return 0x1F >> ($len - 2);
767 }
768
769 # This test is split into this number of files.
770 my $num_test_files = $ENV{TEST_JOBS} || 1;
771 $num_test_files = 10 if $num_test_files > 10;
772
773 # We only really need to test utf8n_to_uvchr_msgs() once with this flag.
774 my $tested_CHECK_ONLY = 0;
775
776 my $test_count = -1;
777
778 # By setting this environment variable to this particular value, we test
779 # essentially all combinations of potential UTF-8, so that can get a
780 # comprehensive test of the decoding routine.  This test assumes the routine
781 # that does the translation from code point to UTF-8 is working.  An assert
782 # can be used in the routine to make sure that the dfa is working precisely
783 # correctly, and any flaws in it aren't being masked by the remainder of the
784 # function.
785 if ($::TEST_CHUNK == 0
786 && $ENV{PERL_DEBUG_FULL_TEST}
787 && $ENV{PERL_DEBUG_FULL_TEST} == 97)
788 {
789     my $min_cont_mask = $min_cont | 0xF;
790     my @bytes = (   0,  # Placeholder to signify to use an empty string ""
791                 ord 'A',# We assume that all the invariant characters are
792                         # properly in the same class, so this is an exemplar
793                         # character
794                 $min_cont .. 0xFF   # But test every non-invariant individually
795                 );
796     my $shift = (isASCII) ? 6 : 5;
797     my $mark = $min_cont;
798     my $mask = (1 << $shift) - 1;
799     for my $byte1 (@bytes) {
800         for my $byte2 (@bytes) {
801             last if $byte2 && ! $byte1;      # Don't test empty preceding byte
802
803             last if $byte2 && $byte1 < 0xC0; # No need to test more than a
804                                              # single byte unless start byte
805                                              # indicates those.
806
807             for my $byte3 (@bytes) {
808                 last if $byte3 && ! $byte2;
809                 last if $byte3 && $byte1 < 0xE0;    # Only test 3 bytes for
810                                                     # 3-byte start byte
811
812                 # If the preceding byte is a start byte, it should fail, and
813                 # there is no need to test illegal bytes that follow.
814                 # Instead, limit ourselves to just a few legal bytes that
815                 # could follow.  This cuts down tremendously on the number of
816                 # tests executed.
817                 next if $byte2 >= 0xC0
818                      && $byte3 >= $min_cont
819                      && ($byte3 & $min_cont_mask) != $min_cont;
820
821                 for my $byte4 (@bytes) {
822                     last if $byte4 && ! $byte3;
823                     last if $byte4 && $byte1 < 0xF0;  # Only test 4 bytes for
824                                                       # 4 byte strings
825
826                     # Like for byte 3, we limit things that come after a
827                     # mispositioned start-byte to just a few things that
828                     # otherwise would be legal
829                     next if ($byte2 >= 0xC0 || $byte3 >= 0xC0)
830                           && $byte4 >= $min_cont
831                           && ($byte4 & $min_cont_mask) != $min_cont;
832
833                     for my $byte5 (@bytes) {
834                         last if $byte5 && ! $byte4;
835                         last if $byte5 && $byte1 < 0xF8;  # Only test 5 bytes for
836                                                           # 5 byte strings
837
838                         # Like for byte 4, we limit things that come after a
839                         # mispositioned start-byte to just a few things that
840                         # otherwise would be legal
841                         next if (   $byte2 >= 0xC0
842                                  || $byte3 >= 0xC0
843                                  || $byte4 >= 0xC0)
844                               && $byte4 >= $min_cont
845                               && ($byte4 & $min_cont_mask) != $min_cont;
846
847                         my $string = "";
848                         $string .= chr $byte1 if $byte1;
849                         $string .= chr $byte2 if $byte2;
850                         $string .= chr $byte3 if $byte3;
851                         $string .= chr $byte4 if $byte4;
852                         $string .= chr $byte5 if $byte5;
853
854                         my $length = length $string;
855                         next unless $length;
856                         last if $byte1 >= ((isASCII) ? 0xF6 : 0xFA);
857
858                         my $native = I8_to_native($string);
859                         my $is_valid = isUTF8_CHAR($native, $length);
860                         my $got_valid = test_isUTF8_CHAR($native, $length);
861                         my $got_strict
862                                     = test_isSTRICT_UTF8_CHAR($native, $length);
863                         my $got_C9
864                                  = test_isC9_STRICT_UTF8_CHAR($native, $length);
865                         my $ret = test_utf8n_to_uvchr_msgs($native, $length,
866                                             $::UTF8_WARN_ILLEGAL_INTERCHANGE);
867                         my $is_strict = $is_valid;
868                         my $is_C9 = $is_valid;
869
870                         if ($is_valid) {
871
872                             # Here, is legal UTF-8.  Verify that it returned
873                             # the correct code point, and if so, that it
874                             # correctly classifies the result.
875                             my $cp = $ret->[0];
876
877                             my $should_be_string;
878                             if ($length == 1) {
879                                 $should_be_string = chr $cp;
880                             }
881                             else {
882
883                                 # Starting with the code point, use first
884                                 # principals to find the equivalent I8 string
885                                 my @bytes;
886                                 my $uv = $cp;
887                                 for (my $i = $length - 1; $i > 0; $i--) {
888                                     $bytes[$i] = chr I8_to_native(($uv & $mask)
889                                                                   | $mark);
890                                     $uv >>= $shift;
891                                 }
892                                 $bytes[0] = chr I8_to_native((   $uv
893                                                         & start_mask($length))
894                                             | start_mark($length));
895                                 $should_be_string = join "", @bytes;
896                             }
897
898                             # If the original string and the inverse are the
899                             # same, it worked.
900                             if (is($native, $should_be_string,
901                                     "utf8n_to_uvchr_msgs("
902                                  .  display_bytes($native)
903                                  . ") returns correct uv=0x"
904                                  . sprintf ("%x", $cp)))
905                             {
906                                 my $is_surrogate = $cp >= 0xD800
907                                                 && $cp <= 0xDFFF;
908                                 my $got_surrogate
909                                     = ($ret->[2] & $::UTF8_GOT_SURROGATE) != 0;
910                                 $is_strict = 0 if $is_surrogate;
911                                 $is_C9 = 0 if $is_surrogate;
912
913                                 my $is_super = $cp > 0x10FFFF;
914                                 my $got_super
915                                         = ($ret->[2] & $::UTF8_GOT_SUPER) != 0;
916                                 $is_strict = 0 if $is_super;
917                                 $is_C9 = 0 if $is_super;
918
919                                 my $is_nonchar = ! $is_super
920                                     && (   ($cp & 0xFFFE) == 0xFFFE
921                                         || ($cp >= 0xFDD0 && $cp <= 0xFDEF));
922                                 my $got_nonchar
923                                       = ($ret->[2] & $::UTF8_GOT_NONCHAR) != 0;
924                                 $is_strict = 0 if $is_nonchar;
925
926                                 is($got_surrogate, $is_surrogate,
927                                     "    And correctly flagged it as"
928                                   . ((! $is_surrogate) ? " not" : "")
929                                   . " being a surrogate");
930                                 is($got_super, $is_super,
931                                     "    And correctly flagged it as"
932                                   . ((! $is_super) ? " not" : "")
933                                   . " being above Unicode");
934                                 is($got_nonchar, $is_nonchar,
935                                     "    And correctly flagged it as"
936                                   . ((! $is_nonchar) ? " not" : "")
937                                   . " being a non-char");
938                             }
939
940                             # This is how we exit the loop normally if things
941                             # are working.  The fail-safe code above is used
942                             # when they aren't.
943                             goto done if $cp > 0x140001;
944                         }
945                         else {
946                             is($ret->[0], 0, "utf8n_to_uvchr_msgs("
947                                             . display_bytes($native)
948                                             . ") correctly returns error");
949                             if (! ($ret->[2] & ($::UTF8_GOT_SHORT
950                                                |$::UTF8_GOT_NON_CONTINUATION
951                                                |$::UTF8_GOT_LONG)))
952                             {
953                                 is($ret->[2] & ( $::UTF8_GOT_NONCHAR
954                                                 |$::UTF8_GOT_SURROGATE
955                                                 |$::UTF8_GOT_SUPER), 0,
956                                 "    And isn't a surrogate, non-char, nor"
957                                 . " above Unicode");
958                              }
959                         }
960
961                         is($got_valid == 0, $is_valid == 0,
962                            "    And isUTF8_CHAR() correctly returns "
963                          . (($got_valid == 0) ? "0" : "non-zero"));
964                         is($got_strict == 0, $is_strict == 0,
965                            "    And isSTRICT_UTF8_CHAR() correctly returns "
966                          . (($got_strict == 0) ? "0" : "non-zero"));
967                         is($got_C9 == 0, $is_C9 == 0,
968                            "    And isC9_UTF8_CHAR() correctly returns "
969                          . (($got_C9 == 0) ? "0" : "non-zero"));
970                     }
971                 }
972             }
973         }
974     }
975   done:
976 }
977
978 foreach my $test (@tests) {
979   $test_count++;
980   next if $test_count % $num_test_files != $::TEST_CHUNK;
981
982   my ($testname, $bytes, $allowed_uv, $needed_to_discern_len) = @$test;
983
984   my $length = length $bytes;
985   my $initially_overlong = $testname =~ /overlong/;
986   my $initially_orphan   = $testname =~ /orphan/;
987   my $will_overflow = $allowed_uv < 0;
988
989   my $uv_string = sprintf(($allowed_uv < 0x100) ? "%02X" : "%04X", $allowed_uv);
990   my $display_bytes = display_bytes($bytes);
991
992   my $controlling_warning_category;
993   my $utf8n_flag_to_warn;
994   my $utf8n_flag_to_disallow;
995   my $uvchr_flag_to_warn;
996   my $uvchr_flag_to_disallow;
997
998   # We want to test that the independent flags are actually independent.
999   # For example, that a surrogate doesn't trigger a non-character warning,
1000   # and conversely, turning off an above-Unicode flag doesn't suppress a
1001   # surrogate warning.  Earlier versions of this file used nested loops to
1002   # test all possible combinations.  But that creates lots of tests, making
1003   # this run too long.  What is now done instead is to use the complement of
1004   # the category we are testing to greatly reduce the combinatorial
1005   # explosion.  For example, if we have a surrogate and we aren't expecting
1006   # a warning about it, we set all the flags for non-surrogates to raise
1007   # warnings.  If one shows up, it indicates the flags aren't independent.
1008   my $utf8n_flag_to_warn_complement;
1009   my $utf8n_flag_to_disallow_complement;
1010   my $uvchr_flag_to_warn_complement;
1011   my $uvchr_flag_to_disallow_complement;
1012
1013   # Many of the code points being tested are middling in that if code point
1014   # edge cases work, these are very likely to as well.  Because this test
1015   # file takes a while to execute, we skip testing the edge effects of code
1016   # points deemed middling, while testing their basics and continuing to
1017   # fully test the non-middling code points.
1018   my $skip_most_tests = 0;
1019
1020   my $cp_message_qr;      # Pattern that matches the message raised when
1021                           # that message contains the problematic code
1022                           # point.  The message is the same (currently) both
1023                           # when going from/to utf8.
1024   my $non_cp_trailing_text;   # The suffix text when the message doesn't
1025                               # contain a code point.  (This is a result of
1026                               # some sort of malformation that means we
1027                               # can't get an exact code poin
1028   my $extended_cp_message_qr = qr/\QCode point 0x$uv_string is not Unicode,\E
1029                       \Q requires a Perl extension, and so is not\E
1030                       \Q portable\E/x;
1031   my $extended_non_cp_trailing_text
1032                       = "is a Perl extension, and so is not portable";
1033
1034   # What bytes should have been used to specify a code point that has been
1035   # specified as an overlong.
1036   my $correct_bytes_for_overlong;
1037
1038   # Is this test malformed from the beginning?  If so, we know to generally
1039   # expect that the tests will show it isn't valid.
1040   my $initially_malformed = 0;
1041
1042   if ($initially_overlong || $initially_orphan) {
1043       $non_cp_trailing_text = "if you see this, there is an error";
1044       $cp_message_qr = qr/\Q$non_cp_trailing_text\E/;
1045       $initially_malformed = 1;
1046       $utf8n_flag_to_warn     = 0;
1047       $utf8n_flag_to_disallow = 0;
1048
1049       $utf8n_flag_to_warn_complement =     $::UTF8_WARN_SURROGATE;
1050       $utf8n_flag_to_disallow_complement = $::UTF8_DISALLOW_SURROGATE;
1051       if (! $will_overflow && $allowed_uv <= 0x10FFFF) {
1052           $utf8n_flag_to_warn_complement     |= $::UTF8_WARN_SUPER;
1053           $utf8n_flag_to_disallow_complement |= $::UTF8_DISALLOW_SUPER;
1054           if (($allowed_uv & 0xFFFF) != 0xFFFF) {
1055               $utf8n_flag_to_warn_complement      |= $::UTF8_WARN_NONCHAR;
1056               $utf8n_flag_to_disallow_complement  |= $::UTF8_DISALLOW_NONCHAR;
1057           }
1058       }
1059       if (! is_extended_utf8($bytes)) {
1060           $utf8n_flag_to_warn_complement |= $::UTF8_WARN_PERL_EXTENDED;
1061           $utf8n_flag_to_disallow_complement  |= $::UTF8_DISALLOW_PERL_EXTENDED;
1062       }
1063
1064       $controlling_warning_category = 'utf8';
1065
1066       if ($initially_overlong) {
1067           if (! defined $needed_to_discern_len) {
1068               $needed_to_discern_len = overlong_discern_len($bytes);
1069           }
1070           $correct_bytes_for_overlong = display_bytes_no_quotes(chr $allowed_uv);
1071       }
1072   }
1073   elsif($will_overflow || $allowed_uv > 0x10FFFF) {
1074
1075       # Set the SUPER flags; later, we test for PERL_EXTENDED as well.
1076       $utf8n_flag_to_warn     = $::UTF8_WARN_SUPER;
1077       $utf8n_flag_to_disallow = $::UTF8_DISALLOW_SUPER;
1078       $uvchr_flag_to_warn     = $::UNICODE_WARN_SUPER;
1079       $uvchr_flag_to_disallow = $::UNICODE_DISALLOW_SUPER;;
1080
1081       # Below, we add the flags for non-perl_extended to the code points
1082       # that don't fit that category.  Special tests are done for this
1083       # category in the inner loop.
1084       $utf8n_flag_to_warn_complement     = $::UTF8_WARN_NONCHAR
1085                                           |$::UTF8_WARN_SURROGATE;
1086       $utf8n_flag_to_disallow_complement = $::UTF8_DISALLOW_NONCHAR
1087                                           |$::UTF8_DISALLOW_SURROGATE;
1088       $uvchr_flag_to_warn_complement     = $::UNICODE_WARN_NONCHAR
1089                                           |$::UNICODE_WARN_SURROGATE;
1090       $uvchr_flag_to_disallow_complement = $::UNICODE_DISALLOW_NONCHAR
1091                                           |$::UNICODE_DISALLOW_SURROGATE;
1092       $controlling_warning_category = 'non_unicode';
1093
1094       if ($will_overflow) {  # This is realy a malformation
1095           $non_cp_trailing_text = "if you see this, there is an error";
1096           $cp_message_qr = qr/\Q$non_cp_trailing_text\E/;
1097           $initially_malformed = 1;
1098           if (! defined $needed_to_discern_len) {
1099               $needed_to_discern_len = overflow_discern_len($length);
1100           }
1101       }
1102       elsif (requires_extended_utf8($allowed_uv)) {
1103           $cp_message_qr = $extended_cp_message_qr;
1104           $non_cp_trailing_text = $extended_non_cp_trailing_text;
1105           $needed_to_discern_len = 1 unless defined $needed_to_discern_len;
1106       }
1107       else {
1108           $cp_message_qr = qr/\QCode point 0x$uv_string is not Unicode,\E
1109                               \Q may not be portable\E/x;
1110           $non_cp_trailing_text = "is for a non-Unicode code point, may not"
1111                               . " be portable";
1112           $utf8n_flag_to_warn_complement     |= $::UTF8_WARN_PERL_EXTENDED;
1113           $utf8n_flag_to_disallow_complement
1114                                           |= $::UTF8_DISALLOW_PERL_EXTENDED;
1115           $uvchr_flag_to_warn_complement |= $::UNICODE_WARN_PERL_EXTENDED;
1116           $uvchr_flag_to_disallow_complement
1117                                       |= $::UNICODE_DISALLOW_PERL_EXTENDED;
1118       }
1119   }
1120   elsif ($allowed_uv >= 0xD800 && $allowed_uv <= 0xDFFF) {
1121       $cp_message_qr = qr/UTF-16 surrogate U\+$uv_string/;
1122       $non_cp_trailing_text = "is for a surrogate";
1123       $needed_to_discern_len = 2 unless defined $needed_to_discern_len;
1124       $skip_most_tests = 1 if $allowed_uv > 0xD800 && $allowed_uv < 0xDFFF;
1125
1126       $utf8n_flag_to_warn     = $::UTF8_WARN_SURROGATE;
1127       $utf8n_flag_to_disallow = $::UTF8_DISALLOW_SURROGATE;
1128       $uvchr_flag_to_warn     = $::UNICODE_WARN_SURROGATE;
1129       $uvchr_flag_to_disallow = $::UNICODE_DISALLOW_SURROGATE;;
1130
1131       $utf8n_flag_to_warn_complement     = $::UTF8_WARN_NONCHAR
1132                                           |$::UTF8_WARN_SUPER
1133                                           |$::UTF8_WARN_PERL_EXTENDED;
1134       $utf8n_flag_to_disallow_complement = $::UTF8_DISALLOW_NONCHAR
1135                                           |$::UTF8_DISALLOW_SUPER
1136                                           |$::UTF8_DISALLOW_PERL_EXTENDED;
1137       $uvchr_flag_to_warn_complement     = $::UNICODE_WARN_NONCHAR
1138                                           |$::UNICODE_WARN_SUPER
1139                                           |$::UNICODE_WARN_PERL_EXTENDED;
1140       $uvchr_flag_to_disallow_complement = $::UNICODE_DISALLOW_NONCHAR
1141                                           |$::UNICODE_DISALLOW_SUPER
1142                                           |$::UNICODE_DISALLOW_PERL_EXTENDED;
1143       $controlling_warning_category = 'surrogate';
1144   }
1145   elsif (   ($allowed_uv >= 0xFDD0 && $allowed_uv <= 0xFDEF)
1146           || ($allowed_uv & 0xFFFE) == 0xFFFE)
1147   {
1148       $cp_message_qr = qr/\QUnicode non-character U+$uv_string\E
1149                           \Q is not recommended for open interchange\E/x;
1150       $non_cp_trailing_text = "if you see this, there is an error";
1151       $needed_to_discern_len = $length unless defined $needed_to_discern_len;
1152       if (   ($allowed_uv > 0xFDD0 && $allowed_uv < 0xFDEF)
1153           || ($allowed_uv > 0xFFFF && $allowed_uv < 0x10FFFE))
1154       {
1155           $skip_most_tests = 1;
1156       }
1157
1158       $utf8n_flag_to_warn     = $::UTF8_WARN_NONCHAR;
1159       $utf8n_flag_to_disallow = $::UTF8_DISALLOW_NONCHAR;
1160       $uvchr_flag_to_warn     = $::UNICODE_WARN_NONCHAR;
1161       $uvchr_flag_to_disallow = $::UNICODE_DISALLOW_NONCHAR;;
1162
1163       $utf8n_flag_to_warn_complement     = $::UTF8_WARN_SURROGATE
1164                                           |$::UTF8_WARN_SUPER
1165                                           |$::UTF8_WARN_PERL_EXTENDED;
1166       $utf8n_flag_to_disallow_complement = $::UTF8_DISALLOW_SURROGATE
1167                                           |$::UTF8_DISALLOW_SUPER
1168                                           |$::UTF8_DISALLOW_PERL_EXTENDED;
1169       $uvchr_flag_to_warn_complement     = $::UNICODE_WARN_SURROGATE
1170                                           |$::UNICODE_WARN_SUPER
1171                                           |$::UNICODE_WARN_PERL_EXTENDED;
1172       $uvchr_flag_to_disallow_complement = $::UNICODE_DISALLOW_SURROGATE
1173                                           |$::UNICODE_DISALLOW_SUPER
1174                                           |$::UNICODE_DISALLOW_PERL_EXTENDED;
1175
1176       $controlling_warning_category = 'nonchar';
1177   }
1178   else {
1179       die "Can't figure out what type of warning to test for $testname"
1180   }
1181
1182   die 'Didn\'t set $needed_to_discern_len for ' . $testname
1183                                       unless defined $needed_to_discern_len;
1184
1185   # We try various combinations of malformations that can occur
1186   foreach my $short (0, 1) {
1187     next if $skip_most_tests && $short;
1188     foreach my $unexpected_noncont (0, 1) {
1189       next if $skip_most_tests && $unexpected_noncont;
1190       foreach my $overlong (0, 1) {
1191         next if $overlong && $skip_most_tests;
1192         next if $initially_overlong && ! $overlong;
1193
1194         # If we're creating an overlong, it can't be longer than the
1195         # maximum length, so skip if we're already at that length.
1196         next if   (! $initially_overlong && $overlong)
1197                   &&  $length >= $::max_bytes;
1198
1199         my $this_cp_message_qr = $cp_message_qr;
1200         my $this_non_cp_trailing_text = $non_cp_trailing_text;
1201
1202         foreach my $malformed_allow_type (0..2) {
1203           # 0 don't allow this malformation; ignored if no malformation
1204           # 1 allow, with REPLACEMENT CHARACTER returned
1205           # 2 allow, with intended code point returned.  All malformations
1206           #   other than overlong can't determine the intended code point,
1207           #   so this isn't valid for them.
1208           next if     $malformed_allow_type == 2
1209                   && ($will_overflow || $short || $unexpected_noncont);
1210           next if $skip_most_tests && $malformed_allow_type;
1211
1212           # Here we are in the innermost loop for malformations.  So we
1213           # know which ones are in effect.  Can now change the input to be
1214           # appropriately malformed.  We also can set up certain other
1215           # things now, like whether we expect a return flag from this
1216           # malformation, and which flag.
1217
1218           my $this_bytes = $bytes;
1219           my $this_length = $length;
1220           my $this_expected_len = $length;
1221           my $this_needed_to_discern_len = $needed_to_discern_len;
1222
1223           my @malformation_names;
1224           my @expected_malformation_warnings;
1225           my @expected_malformation_return_flags;
1226
1227           # Contains the flags for any allowed malformations.  Currently no
1228           # combinations of on/off are tested for.  It's either all are
1229           # allowed, or none are.
1230           my $allow_flags = 0;
1231           my $overlong_is_in_perl_extended_utf8 = 0;
1232           my $dont_use_overlong_cp = 0;
1233
1234           if ($initially_orphan) {
1235               next if $overlong || $short || $unexpected_noncont;
1236           }
1237
1238           if ($overlong) {
1239               if (! $initially_overlong) {
1240                   my $new_expected_len;
1241
1242                   # To force this malformation, we convert the original start
1243                   # byte into a continuation byte with the same data bits as
1244                   # originally. ...
1245                   my $start_byte = substr($this_bytes, 0, 1);
1246                   my $converted_to_continuation_byte
1247                                           = start_byte_to_cont($start_byte);
1248
1249                   # ... Then we prepend it with a known overlong sequence.
1250                   # This should evaluate to the exact same code point as the
1251                   # original.  We try to avoid an overlong using Perl
1252                   # extended UTF-8.  The code points are the highest
1253                   # representable as overlongs on the respective platform
1254                   # without using extended UTF-8.
1255                   if (native_to_I8($start_byte) lt "\xFC") {
1256                       $start_byte = I8_to_native("\xFC");
1257                       $new_expected_len = 6;
1258                   }
1259                   elsif (! isASCII && native_to_I8($start_byte) lt "\xFE") {
1260
1261                       # FE is not extended UTF-8 on EBCDIC
1262                       $start_byte = I8_to_native("\xFE");
1263                       $new_expected_len = 7;
1264                   }
1265                   else {  # Must use extended UTF-8.  On ASCII platforms, we
1266                           # could express some overlongs here starting with
1267                           # \xFE, but there's no real reason to do so.
1268                       $overlong_is_in_perl_extended_utf8 = 1;
1269                       $start_byte = I8_to_native("\xFF");
1270                       $new_expected_len = $::max_bytes;
1271                       $this_cp_message_qr = $extended_cp_message_qr;
1272
1273                       # The warning that gets raised doesn't include the
1274                       # code point in the message if the code point can be
1275                       # expressed without using extended UTF-8, but the
1276                       # particular overlong sequence used is in extended
1277                       # UTF-8.  To do otherwise would be confusing to the
1278                       # user, as it would claim the code point requires
1279                       # extended, when it doesn't.
1280                       $dont_use_overlong_cp = 1
1281                                   unless requires_extended_utf8($allowed_uv);
1282                       $this_non_cp_trailing_text
1283                                             = $extended_non_cp_trailing_text;
1284                   }
1285
1286                   # Splice in the revise continuation byte, preceded by the
1287                   # start byte and the proper number of the lowest
1288                   # continuation bytes.
1289                   $this_bytes =   $start_byte
1290                               . ($native_lowest_continuation_chr
1291                                   x (  $new_expected_len
1292                                       - 1
1293                                       - length($this_bytes)))
1294                               .  $converted_to_continuation_byte
1295                               .  substr($this_bytes, 1);
1296                   $this_length = length($this_bytes);
1297                   $this_needed_to_discern_len =    $new_expected_len
1298                                               - (  $this_expected_len
1299                                               - $this_needed_to_discern_len);
1300                   $this_expected_len = $new_expected_len;
1301               }
1302           }
1303
1304           if ($short) {
1305
1306               # To force this malformation, just tell the test to not look
1307               # as far as it should into the input.
1308               $this_length--;
1309               $this_expected_len--;
1310
1311               $allow_flags |= $::UTF8_ALLOW_SHORT if $malformed_allow_type;
1312           }
1313
1314           if ($unexpected_noncont) {
1315
1316               # To force this malformation, change the final continuation
1317               # byte into a start byte.
1318               my $pos = ($short) ? -2 : -1;
1319               substr($this_bytes, $pos, 1) = $known_start_byte;
1320               $this_expected_len--;
1321           }
1322
1323           # The whole point of a test that is malformed from the beginning
1324           # is to test for that malformation.  If we've modified things so
1325           # much that we don't have enough information to detect that
1326           # malformation, there's no point in testing.
1327           next if    $initially_malformed
1328                   && $this_expected_len < $this_needed_to_discern_len;
1329
1330           # Here, we've transformed the input with all of the desired
1331           # non-overflow malformations.  We are now in a position to
1332           # construct any potential warnings for those malformations.  But
1333           # it's a pain to get the detailed messages exactly right, so for
1334           # now XXX, only do so for those that return an explicit code
1335           # point.
1336
1337           if ($initially_orphan) {
1338               push @malformation_names, "orphan continuation";
1339               push @expected_malformation_return_flags,
1340                                                   $::UTF8_GOT_CONTINUATION;
1341               $allow_flags |= $::UTF8_ALLOW_CONTINUATION
1342                                                   if $malformed_allow_type;
1343               push @expected_malformation_warnings, qr/unexpected continuation/;
1344           }
1345
1346           if ($overlong) {
1347               push @malformation_names, 'overlong';
1348               push @expected_malformation_return_flags, $::UTF8_GOT_LONG;
1349
1350               # If one of the other malformation types is also in effect, we
1351               # don't know what the intended code point was.
1352               if ($short || $unexpected_noncont || $will_overflow) {
1353                   push @expected_malformation_warnings, qr/overlong/;
1354               }
1355               else {
1356                   my $wrong_bytes = display_bytes_no_quotes(
1357                                         substr($this_bytes, 0, $this_length));
1358                   if (! defined $correct_bytes_for_overlong) {
1359                       $correct_bytes_for_overlong
1360                                           = display_bytes_no_quotes($bytes);
1361                   }
1362                   my $prefix = (   $allowed_uv > 0x10FFFF
1363                                 || ! isASCII && $allowed_uv < 256)
1364                                 ? "0x"
1365                                 : "U+";
1366                   push @expected_malformation_warnings,
1367                           qr/\QMalformed UTF-8 character: $wrong_bytes\E
1368                               \Q (overlong; instead use\E
1369                               \Q $correct_bytes_for_overlong to\E
1370                               \Q represent $prefix$uv_string)/x;
1371               }
1372
1373               if ($malformed_allow_type == 2) {
1374                   $allow_flags |= $::UTF8_ALLOW_LONG_AND_ITS_VALUE;
1375               }
1376               elsif ($malformed_allow_type) {
1377                   $allow_flags |= $::UTF8_ALLOW_LONG;
1378               }
1379           }
1380           if ($short) {
1381               push @malformation_names, 'short';
1382               push @expected_malformation_return_flags, $::UTF8_GOT_SHORT;
1383               push @expected_malformation_warnings, qr/too short/;
1384           }
1385           if ($unexpected_noncont) {
1386               push @malformation_names, 'unexpected non-continuation';
1387               push @expected_malformation_return_flags,
1388                               $::UTF8_GOT_NON_CONTINUATION;
1389               $allow_flags |= $::UTF8_ALLOW_NON_CONTINUATION
1390                                                   if $malformed_allow_type;
1391               push @expected_malformation_warnings,
1392                                       qr/unexpected non-continuation byte/;
1393           }
1394
1395           # The overflow malformation is done differently than other
1396           # malformations.  It comes from manually typed tests in the test
1397           # array.  We now make it be treated like one of the other
1398           # malformations.  But some has to be deferred until the inner loop
1399           my $overflow_msg_pattern;
1400           if ($will_overflow) {
1401               push @malformation_names, 'overflow';
1402
1403               $overflow_msg_pattern = display_bytes_no_quotes(
1404                                   substr($this_bytes, 0, $this_expected_len));
1405               $overflow_msg_pattern = qr/\QMalformed UTF-8 character:\E
1406                                           \Q $overflow_msg_pattern\E
1407                                           \Q (overflows)\E/x;
1408               push @expected_malformation_return_flags, $::UTF8_GOT_OVERFLOW;
1409               $allow_flags |= $::UTF8_ALLOW_OVERFLOW if $malformed_allow_type;
1410           }
1411
1412           # And we can create the malformation-related text for the the test
1413           # names we eventually will generate.
1414           my $malformations_name = "";
1415           if (@malformation_names) {
1416               $malformations_name .= "dis" unless $malformed_allow_type;
1417               $malformations_name .= "allowed ";
1418               $malformations_name .= "malformation";
1419               $malformations_name .= "s" if @malformation_names > 1;
1420               $malformations_name .= ": ";
1421               $malformations_name .=  join "/", @malformation_names;
1422               $malformations_name =  " ($malformations_name)";
1423           }
1424
1425           # Done setting up the malformation related stuff
1426
1427           {   # First test the isFOO calls
1428               use warnings; # XXX no warnings 'deprecated';   # Make sure these don't raise warnings
1429               undef @warnings_gotten;
1430
1431               my $ret = test_isUTF8_CHAR($this_bytes, $this_length);
1432               my $ret_flags
1433                       = test_isUTF8_CHAR_flags($this_bytes, $this_length, 0);
1434               if ($malformations_name) {
1435                   is($ret, 0, "For $testname$malformations_name: isUTF8_CHAR() returns 0");
1436                   is($ret_flags, 0, "    And isUTF8_CHAR_flags() returns 0");
1437               }
1438               else {
1439                   is($ret, $this_length, "For $testname: isUTF8_CHAR() returns"
1440                                         . " expected length: $this_length");
1441                   is($ret_flags, $this_length,
1442                       "    And isUTF8_CHAR_flags(...,0) returns expected"
1443                     . " length: $this_length");
1444               }
1445               is(scalar @warnings_gotten, 0,
1446                   "    And neither isUTF8_CHAR() nor isUTF8_CHAR()_flags"
1447                 . " generated any warnings")
1448               or output_warnings(@warnings_gotten);
1449
1450               undef @warnings_gotten;
1451               $ret = test_isSTRICT_UTF8_CHAR($this_bytes, $this_length);
1452               if ($malformations_name) {
1453                   is($ret, 0, "    And isSTRICT_UTF8_CHAR() returns 0");
1454               }
1455               else {
1456                   my $expected_ret
1457                               = (   $testname =~ /surrogate|non-character/
1458                                   || $allowed_uv > 0x10FFFF)
1459                                 ? 0
1460                                 : $this_length;
1461                   is($ret, $expected_ret,
1462                       "    And isSTRICT_UTF8_CHAR() returns expected"
1463                     . " length: $expected_ret");
1464                   $ret = test_isUTF8_CHAR_flags($this_bytes, $this_length,
1465                                       $::UTF8_DISALLOW_ILLEGAL_INTERCHANGE);
1466                   is($ret, $expected_ret,
1467                       "    And isUTF8_CHAR_flags('"
1468                     . "DISALLOW_ILLEGAL_INTERCHANGE') acts like"
1469                     . " isSTRICT_UTF8_CHAR");
1470               }
1471               is(scalar @warnings_gotten, 0,
1472                       "    And neither isSTRICT_UTF8_CHAR() nor"
1473                     . " isUTF8_CHAR_flags generated any warnings")
1474               or output_warnings(@warnings_gotten);
1475
1476               undef @warnings_gotten;
1477               $ret = test_isC9_STRICT_UTF8_CHAR($this_bytes, $this_length);
1478               if ($malformations_name) {
1479                   is($ret, 0, "    And isC9_STRICT_UTF8_CHAR() returns 0");
1480               }
1481               else {
1482                   my $expected_ret = (   $testname =~ /surrogate/
1483                                       || $allowed_uv > 0x10FFFF)
1484                                       ? 0
1485                                       : $this_expected_len;
1486                   is($ret, $expected_ret, "    And isC9_STRICT_UTF8_CHAR()"
1487                                         . " returns expected length:"
1488                                         . " $expected_ret");
1489                   $ret = test_isUTF8_CHAR_flags($this_bytes, $this_length,
1490                                   $::UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE);
1491                   is($ret, $expected_ret,
1492                       "    And isUTF8_CHAR_flags('"
1493                     . "DISALLOW_ILLEGAL_C9_INTERCHANGE') acts like"
1494                     . " isC9_STRICT_UTF8_CHAR");
1495               }
1496               is(scalar @warnings_gotten, 0,
1497                       "    And neither isC9_STRICT_UTF8_CHAR() nor"
1498                     . " isUTF8_CHAR_flags generated any warnings")
1499               or output_warnings(@warnings_gotten);
1500
1501               foreach my $disallow_type (0..2) {
1502                   # 0 is don't disallow this type of code point
1503                   # 1 is do disallow
1504                   # 2 is do disallow, but only code points requiring
1505                   #   perl-extended-UTF8
1506
1507                   my $disallow_flags;
1508                   my $expected_ret;
1509
1510                   if ($malformations_name) {
1511
1512                       # Malformations are by default disallowed, so testing
1513                       # with $disallow_type equal to 0 is sufficicient.
1514                       next if $disallow_type;
1515
1516                       $disallow_flags = 0;
1517                       $expected_ret = 0;
1518                   }
1519                   elsif ($disallow_type == 1) {
1520                       $disallow_flags = $utf8n_flag_to_disallow;
1521                       $expected_ret = 0;
1522                   }
1523                   elsif ($disallow_type == 2) {
1524                       next if ! requires_extended_utf8($allowed_uv);
1525                       $disallow_flags = $::UTF8_DISALLOW_PERL_EXTENDED;
1526                       $expected_ret = 0;
1527                   }
1528                   else {  # type is 0
1529                       $disallow_flags = $utf8n_flag_to_disallow_complement;
1530                       $expected_ret = $this_length;
1531                   }
1532
1533                   $ret = test_isUTF8_CHAR_flags($this_bytes, $this_length,
1534                                                 $disallow_flags);
1535                   is($ret, $expected_ret,
1536                             "    And isUTF8_CHAR_flags($display_bytes,"
1537                           . " $disallow_flags) returns $expected_ret")
1538                     or diag "The flags mean "
1539                           . flags_to_text($disallow_flags,
1540                                           \@utf8n_flags_to_text);
1541                   is(scalar @warnings_gotten, 0,
1542                           "    And isUTF8_CHAR_flags(...) generated"
1543                         . " no warnings")
1544                     or output_warnings(@warnings_gotten);
1545
1546                   # Test partial character handling, for each byte not a
1547                   # full character
1548                   my $did_test_partial = 0;
1549                   for (my $j = 1; $j < $this_length - 1; $j++) {
1550                       $did_test_partial = 1;
1551                       my $partial = substr($this_bytes, 0, $j);
1552                       my $ret_should_be;
1553                       my $comment;
1554                       if ($disallow_type || $malformations_name) {
1555                           $ret_should_be = 0;
1556                           $comment = "disallowed";
1557
1558                           # The number of bytes required to tell if a
1559                           # sequence has something wrong is the smallest of
1560                           # all the things wrong with it.  We start with the
1561                           # number for this type of code point, if that is
1562                           # disallowed; or the whole length if not.  The
1563                           # latter is what a couple of the malformations
1564                           # require.
1565                           my $needed_to_tell = ($disallow_type)
1566                                                 ? $this_needed_to_discern_len
1567                                                 : $this_expected_len;
1568
1569                           # Then we see if the malformations that are
1570                           # detectable early in the string are present.
1571                           if ($overlong) {
1572                               my $dl = overlong_discern_len($this_bytes);
1573                               $needed_to_tell = $dl if $dl < $needed_to_tell;
1574                           }
1575                           if ($will_overflow) {
1576                               my $dl = overflow_discern_len($length);
1577                               $needed_to_tell = $dl if $dl < $needed_to_tell;
1578                           }
1579
1580                           if ($j < $needed_to_tell) {
1581                               $ret_should_be = 1;
1582                               $comment .= ", but need $needed_to_tell"
1583                                         . " bytes to discern:";
1584                           }
1585                       }
1586                       else {
1587                           $ret_should_be = 1;
1588                           $comment = "allowed";
1589                       }
1590
1591                       undef @warnings_gotten;
1592
1593                       $ret = test_is_utf8_valid_partial_char_flags($partial,
1594                                                       $j, $disallow_flags);
1595                       is($ret, $ret_should_be,
1596                           "    And is_utf8_valid_partial_char_flags("
1597                           . display_bytes($partial)
1598                           . ", $disallow_flags), $comment: returns"
1599                           . " $ret_should_be")
1600                       or diag "The flags mean "
1601                       . flags_to_text($disallow_flags, \@utf8n_flags_to_text);
1602                   }
1603
1604                   if ($did_test_partial) {
1605                       is(scalar @warnings_gotten, 0,
1606                           "    And is_utf8_valid_partial_char_flags()"
1607                           . " generated no warnings for any of the lengths")
1608                         or output_warnings(@warnings_gotten);
1609                   }
1610               }
1611           }
1612
1613           # Now test the to/from UTF-8 calls.  There are several orthogonal
1614           # variables involved.  We test most possible combinations
1615
1616           foreach my $do_disallow (0, 1) {
1617             if ($do_disallow) {
1618               next if $initially_overlong || $initially_orphan;
1619             }
1620             else {
1621               next if $skip_most_tests;
1622             }
1623
1624             # This tests four functions: utf8n_to_uvchr_error,
1625             # utf8n_to_uvchr_msgs, uvchr_to_utf8_flags, and
1626             # uvchr_to_utf8_msgs.  The first two are variants of each other,
1627             # and the final two also form a pair.  We use a loop 'which_func'
1628             # to determine which of each pair is being tested.  The main loop
1629             # tests either the first and third, or the 2nd and fourth.
1630             # which_func is sets whether we are expecting warnings or not in
1631             # certain places.  The _msgs() version of the functions expects
1632             # warnings even if lexical ones are turned off, so by making its
1633             # which_func == 1, we can say we want warnings; whereas the other
1634             # one with the value 0, doesn't get them.
1635             for my $which_func (0, 1) {
1636               my $utf8_func = ($which_func)
1637                           ? 'utf8n_to_uvchr_msgs'
1638                           : 'utf8n_to_uvchr_error';
1639
1640               # We classify the warnings into certain "interesting" types,
1641               # described later
1642               foreach my $warning_type (0..4) {
1643                 next if $skip_most_tests && $warning_type != 1;
1644                 foreach my $use_warn_flag (0, 1) {
1645                     if ($use_warn_flag) {
1646                         next if $initially_overlong || $initially_orphan;
1647
1648                         # Since foo_msgs() expects warnings even when lexical
1649                         # ones are turned off, we can skip testing it when
1650                         # they are turned on, with little likelihood of
1651                         # missing an error case.
1652                         next if $which_func;
1653                     }
1654                     else {
1655                         next if $skip_most_tests;
1656                     }
1657
1658                     # Finally, here is the inner loop
1659
1660                     my $this_utf8n_flag_to_warn = $utf8n_flag_to_warn;
1661                     my $this_utf8n_flag_to_disallow = $utf8n_flag_to_disallow;
1662                     my $this_uvchr_flag_to_warn = $uvchr_flag_to_warn;
1663                     my $this_uvchr_flag_to_disallow = $uvchr_flag_to_disallow;
1664
1665                     my $eval_warn;
1666                     my $expect_regular_warnings;
1667                     my $expect_warnings_for_malformed;
1668                     my $expect_warnings_for_overflow;
1669
1670                     if ($warning_type == 0) {
1671                         $eval_warn = "use warnings";
1672                         $expect_regular_warnings = $use_warn_flag;
1673
1674                         # We ordinarily expect overflow warnings here.  But it
1675                         # is somewhat more complicated, and the final
1676                         # determination is deferred to one place in the file
1677                         # where we handle overflow.
1678                         $expect_warnings_for_overflow = 1;
1679
1680                         # We would ordinarily expect malformed warnings in
1681                         # this case, but not if malformations are allowed.
1682                         $expect_warnings_for_malformed
1683                                                 = $malformed_allow_type == 0;
1684                     }
1685                     elsif ($warning_type == 1) {
1686                         $eval_warn = "no warnings";
1687                         $expect_regular_warnings = $which_func;
1688                         $expect_warnings_for_overflow = $which_func;
1689                         $expect_warnings_for_malformed = $which_func;
1690                     }
1691                     elsif ($warning_type == 2) {
1692                         $eval_warn = "no warnings; use warnings 'utf8'";
1693                         $expect_regular_warnings = $use_warn_flag;
1694                         $expect_warnings_for_overflow = 1;
1695                         $expect_warnings_for_malformed
1696                                                 = $malformed_allow_type == 0;
1697                     }
1698                     elsif ($warning_type == 3) {
1699                         $eval_warn = "no warnings; use warnings"
1700                                    . " '$controlling_warning_category'";
1701                         $expect_regular_warnings = $use_warn_flag;
1702                         $expect_warnings_for_overflow
1703                             = $controlling_warning_category eq 'non_unicode';
1704                         $expect_warnings_for_malformed = $which_func;
1705                     }
1706                     elsif ($warning_type == 4) {  # Like type 3, but uses the
1707                                                   # PERL_EXTENDED flags
1708                         # The complement flags were set up so that the
1709                         # PERL_EXTENDED flags have been tested that they don't
1710                         # trigger wrongly for too small code points.  And the
1711                         # flags have been set up so that those small code
1712                         # points are tested for being above Unicode.  What's
1713                         # left to test is that the large code points do
1714                         # trigger the PERL_EXTENDED flags.
1715                         next if ! requires_extended_utf8($allowed_uv);
1716                         next if $controlling_warning_category ne 'non_unicode';
1717                         $eval_warn = "no warnings; use warnings 'non_unicode'";
1718                         $expect_regular_warnings = 1;
1719                         $expect_warnings_for_overflow = 1;
1720                         $expect_warnings_for_malformed = 0;
1721                         $this_utf8n_flag_to_warn = $::UTF8_WARN_PERL_EXTENDED;
1722                         $this_utf8n_flag_to_disallow
1723                                              = $::UTF8_DISALLOW_PERL_EXTENDED;
1724                         $this_uvchr_flag_to_warn
1725                                               = $::UNICODE_WARN_PERL_EXTENDED;
1726                         $this_uvchr_flag_to_disallow
1727                                           = $::UNICODE_DISALLOW_PERL_EXTENDED;
1728                     }
1729                     else {
1730                        die "Unexpected warning type '$warning_type'";
1731                     }
1732
1733                     # We only need to test the case where all warnings are
1734                     # enabled (type 0) to see if turning off the warning flag
1735                     # causes things to not be output.  If those pass, then
1736                     # turning on some sub-category of warnings, or turning off
1737                     # warnings altogether are extremely likely to not output
1738                     # warnings either, given how the warnings subsystem is
1739                     # supposed to work, and this file assumes it does work.
1740                     next if $warning_type != 0 && ! $use_warn_flag;
1741
1742                     # The convention is that the 'got' flag is the same value
1743                     # as the disallow one.  If this were violated, the tests
1744                     # here should start failing.
1745                     my $return_flag = $this_utf8n_flag_to_disallow;
1746
1747                     # If we aren't expecting warnings/disallow for this, turn
1748                     # on all the other flags.  That makes sure that they all
1749                     # are independent of this flag, and so we don't need to
1750                     # test them individually.
1751                     my $this_warning_flags
1752                             = ($use_warn_flag)
1753                               ? $this_utf8n_flag_to_warn
1754                               : ($overlong_is_in_perl_extended_utf8
1755                                 ? ($utf8n_flag_to_warn_complement
1756                                     & ~$::UTF8_WARN_PERL_EXTENDED)
1757                                 :  $utf8n_flag_to_warn_complement);
1758                     my $this_disallow_flags
1759                             = ($do_disallow)
1760                               ? $this_utf8n_flag_to_disallow
1761                               : ($overlong_is_in_perl_extended_utf8
1762                                  ? ($utf8n_flag_to_disallow_complement
1763                                     & ~$::UTF8_DISALLOW_PERL_EXTENDED)
1764                                  :  $utf8n_flag_to_disallow_complement);
1765                     my $expected_uv = $allowed_uv;
1766                     my $this_uv_string = $uv_string;
1767
1768                     my @expected_return_flags
1769                                         = @expected_malformation_return_flags;
1770                     my @expected_warnings;
1771                     push @expected_warnings, @expected_malformation_warnings
1772                                             if $expect_warnings_for_malformed;
1773
1774                     # The overflow malformation is done differently than other
1775                     # malformations.  It comes from manually typed tests in
1776                     # the test array, but it also is above Unicode and uses
1777                     # Perl extended UTF-8, so affects some of the flags being
1778                     # tested.  We now make it be treated like one of the other
1779                     # generated malformations.
1780                     if ($will_overflow) {
1781
1782                         # An overflow is (way) above Unicode, and overrides
1783                         # everything else.
1784                         $expect_regular_warnings = 0;
1785
1786                         # Earlier, we tentatively calculated whether this
1787                         # should emit a message or not.  It's tentative
1788                         # because, even if we ordinarily would output it, we
1789                         # don't if malformations are allowed -- except an
1790                         # overflow is also a SUPER and PERL_EXTENDED, and if
1791                         # warnings for those are enabled, the overflow
1792                         # warning does get raised.
1793                         if (   $expect_warnings_for_overflow
1794                             && (    $malformed_allow_type == 0
1795                                 ||   (   $this_warning_flags
1796                                       & ($::UTF8_WARN_SUPER
1797                                         |$::UTF8_WARN_PERL_EXTENDED))))
1798                         {
1799                             push @expected_warnings, $overflow_msg_pattern;
1800                         }
1801                     }
1802
1803                     # It may be that the malformations have shortened the
1804                     # amount of input we look at so much that we can't tell
1805                     # what the category the code point was in.  Otherwise, set
1806                     # up the expected return flags based on the warnings and
1807                     # disallowments.
1808                     if ($this_expected_len < $this_needed_to_discern_len) {
1809                         $expect_regular_warnings = 0;
1810                     }
1811                     elsif (   ($this_warning_flags & $this_utf8n_flag_to_warn)
1812                            || (  $this_disallow_flags
1813                                & $this_utf8n_flag_to_disallow))
1814                     {
1815                         push @expected_return_flags, $return_flag;
1816                     }
1817
1818                     # Finish setting up the expected warning.
1819                     if ($expect_regular_warnings) {
1820
1821                         # So far the array contains warnings generated by
1822                         # malformations.  Add the expected regular one.
1823                         unshift @expected_warnings, $this_cp_message_qr;
1824
1825                         # But it may need to be modified, because either of
1826                         # these malformations means we can't determine the
1827                         # expected code point.
1828                         if (   $short || $unexpected_noncont
1829                             || $dont_use_overlong_cp)
1830                         {
1831                             my $first_byte = substr($this_bytes, 0, 1);
1832                             $expected_warnings[0] = display_bytes(
1833                                     substr($this_bytes, 0, $this_expected_len));
1834                             $expected_warnings[0]
1835                                 = qr/[Aa]\Qny UTF-8 sequence that starts with\E
1836                                      \Q $expected_warnings[0]\E
1837                                      \Q $this_non_cp_trailing_text\E/x;
1838                         }
1839                     }
1840
1841                     # Is effectively disallowed if we've set up a malformation
1842                     # (unless malformations are allowed), even if the flag
1843                     # indicates it is allowed.  Fix up test name to indicate
1844                     # this as well
1845                     my $disallowed = 0;
1846                     if (   $this_disallow_flags & $this_utf8n_flag_to_disallow
1847                         && $this_expected_len >= $this_needed_to_discern_len)
1848                     {
1849                         $disallowed = 1;
1850                     }
1851                     if ($malformations_name) {
1852                         if ($malformed_allow_type == 0) {
1853                             $disallowed = 1;
1854                         }
1855                         elsif ($malformed_allow_type == 1) {
1856
1857                             # Even if allowed, the malformation returns the
1858                             # REPLACEMENT CHARACTER.
1859                             $expected_uv = 0xFFFD;
1860                             $this_uv_string = "0xFFFD"
1861                         }
1862                     }
1863
1864                     my $this_name = "$utf8_func() $testname: ";
1865                     my @scratch_expected_return_flags = @expected_return_flags;
1866                     if (! $initially_malformed) {
1867                         $this_name .= ($disallowed)
1868                                        ? 'disallowed, '
1869                                        : 'allowed, ';
1870                     }
1871                     $this_name .= "$eval_warn";
1872                     $this_name .= ", " . ((  $this_warning_flags
1873                                             & $this_utf8n_flag_to_warn)
1874                                           ? 'with flag for raising warnings'
1875                                           : 'no flag for raising warnings');
1876                     $this_name .= $malformations_name;
1877
1878                     # Do the actual test using an eval
1879                     undef @warnings_gotten;
1880                     my $ret_ref;
1881                     my $this_flags
1882                         = $allow_flags|$this_warning_flags|$this_disallow_flags;
1883                     my $eval_text =      "$eval_warn; \$ret_ref"
1884                             . " = test_$utf8_func("
1885                             . "'$this_bytes', $this_length, $this_flags)";
1886                     eval "$eval_text";
1887                     if (! ok ($@ eq "", "$this_name: eval succeeded"))
1888                     {
1889                         diag "\$@='$@'; call was: "
1890                            . utf8n_display_call($eval_text);
1891                         next;
1892                     }
1893
1894                     if ($disallowed) {
1895                         is($ret_ref->[0], 0, "    And returns 0")
1896                           or diag "Call was: " . utf8n_display_call($eval_text);
1897                     }
1898                     else {
1899                         is($ret_ref->[0], $expected_uv,
1900                                 "    And returns expected uv: "
1901                               . $this_uv_string)
1902                           or diag "Call was: " . utf8n_display_call($eval_text);
1903                     }
1904                     is($ret_ref->[1], $this_expected_len,
1905                                         "    And returns expected length:"
1906                                       . " $this_expected_len")
1907                       or diag "Call was: " . utf8n_display_call($eval_text);
1908
1909                     my $returned_flags = $ret_ref->[2];
1910
1911                     for (my $i = @scratch_expected_return_flags - 1;
1912                          $i >= 0;
1913                          $i--)
1914                     {
1915                       if ($scratch_expected_return_flags[$i] & $returned_flags)
1916                       {
1917                           if ($scratch_expected_return_flags[$i]
1918                                               == $::UTF8_GOT_PERL_EXTENDED)
1919                           {
1920                               pass("    Expected and got return flag for"
1921                                   . " PERL_EXTENDED");
1922                           }
1923                                   # The first entries in this are
1924                                   # malformations
1925                           elsif ($i > @malformation_names - 1)  {
1926                               pass("    Expected and got return flag"
1927                                   . " for " . $controlling_warning_category);
1928                           }
1929                           else {
1930                               pass("    Expected and got return flag for "
1931                                   . $malformation_names[$i]
1932                                   . " malformation");
1933                           }
1934                           $returned_flags
1935                                       &= ~$scratch_expected_return_flags[$i];
1936                           splice @scratch_expected_return_flags, $i, 1;
1937                       }
1938                     }
1939
1940                     if (! is($returned_flags, 0,
1941                        "    Got no unexpected return flags"))
1942                     {
1943                         diag "The unexpected flags gotten were: "
1944                            . (flags_to_text($returned_flags,
1945                                             \@utf8n_flags_to_text)
1946                                 # We strip off any prefixes from the flag
1947                                 # names
1948                              =~ s/ \b [A-Z] _ //xgr);
1949                         diag "Call was: " . utf8n_display_call($eval_text);
1950                     }
1951
1952                     if (! is (scalar @scratch_expected_return_flags, 0,
1953                         "    Got all expected return flags"))
1954                     {
1955                         diag "The expected flags not gotten were: "
1956                            . (flags_to_text(eval join("|",
1957                                                 @scratch_expected_return_flags),
1958                                             \@utf8n_flags_to_text)
1959                                 # We strip off any prefixes from the flag
1960                                 # names
1961                              =~ s/ \b [A-Z] _ //xgr);
1962                         diag "Call was: " . utf8n_display_call($eval_text);
1963                     }
1964
1965                     if ($which_func) {
1966                         my @returned_warnings;
1967                         for my $element_ref (@{$ret_ref->[3]}) {
1968                             push @returned_warnings, $element_ref->{'text'};
1969                             my $text = $element_ref->{'text'};
1970                             my $flag = $element_ref->{'flag_bit'};
1971                             my $category = $element_ref->{'warning_category'};
1972
1973                             if (! ok(($flag & ($flag-1)) == 0,
1974                                       "flag for returned msg is a single bit"))
1975                             {
1976                               diag sprintf("flags are %x; msg=%s", $flag, $text);
1977                             }
1978                             else {
1979                               if (grep { $_ == $flag } @expected_return_flags) {
1980                                   pass("flag for returned msg is expected");
1981                               }
1982                               else {
1983                                   fail("flag ("
1984                                      . flags_to_text($flag, \@utf8n_flags_to_text)
1985                                      . ") for returned msg is expected");
1986                               }
1987                             }
1988
1989                             # In perl space, don't know the category numbers
1990                             isnt($category, 0,
1991                                           "returned category for msg isn't 0");
1992                         }
1993
1994                         ok(@warnings_gotten == 0, "$utf8_func raised no warnings;"
1995                               . " the next tests are for ones in the returned"
1996                               . " variable")
1997                             or diag join "\n", "The unexpected warnings were:",
1998                                                               @warnings_gotten;
1999                         @warnings_gotten = @returned_warnings;
2000                     }
2001
2002                     do_warnings_test(@expected_warnings)
2003                       or diag "Call was: " . utf8n_display_call($eval_text);
2004                     undef @warnings_gotten;
2005
2006                     # Check CHECK_ONLY results when the input is
2007                     # disallowed.  Do this when actually disallowed,
2008                     # not just when the $this_disallow_flags is set.  We only
2009                     # test once utf8n_to_uvchr_msgs() with this.
2010                     if (   $disallowed
2011                         && ($which_func == 0 || ! $tested_CHECK_ONLY))
2012                     {
2013                         $tested_CHECK_ONLY = 1;
2014                         my $this_flags = $this_disallow_flags|$::UTF8_CHECK_ONLY;
2015                         my $eval_text = "use warnings; \$ret_ref ="
2016                                       . " test_$utf8_func('"
2017                                       . "$this_bytes', $this_length,"
2018                                       . " $this_flags)";
2019                         eval $eval_text;
2020                         if (! ok ($@ eq "",
2021                             "    And eval succeeded with CHECK_ONLY"))
2022                         {
2023                             diag "\$@='$@'; Call was: "
2024                                . utf8n_display_call($eval_text);
2025                             next;
2026                         }
2027                         is($ret_ref->[0], 0, "    CHECK_ONLY: Returns 0")
2028                           or diag "Call was: " . utf8n_display_call($eval_text);
2029                         is($ret_ref->[1], -1,
2030                                        "    CHECK_ONLY: returns -1 for length")
2031                           or diag "Call was: " . utf8n_display_call($eval_text);
2032                         if (! is(scalar @warnings_gotten, 0,
2033                                       "    CHECK_ONLY: no warnings generated"))
2034                         {
2035                             diag "Call was: " . utf8n_display_call($eval_text);
2036                             output_warnings(@warnings_gotten);
2037                         }
2038                     }
2039
2040                     # Now repeat some of the above, but for
2041                     # uvchr_to_utf8_flags().  Since this comes from an
2042                     # existing code point, it hasn't overflowed, and isn't
2043                     # malformed.
2044                     next if @malformation_names;
2045
2046                     my $uvchr_func = ($which_func)
2047                                      ? 'uvchr_to_utf8_flags_msgs'
2048                                      : 'uvchr_to_utf8_flags';
2049
2050                     $this_warning_flags = ($use_warn_flag)
2051                                           ? $this_uvchr_flag_to_warn
2052                                           : 0;
2053                     $this_disallow_flags = ($do_disallow)
2054                                            ? $this_uvchr_flag_to_disallow
2055                                            : 0;
2056
2057                     $disallowed = $this_disallow_flags
2058                                 & $this_uvchr_flag_to_disallow;
2059                     $this_name .= ", " . ((  $this_warning_flags
2060                                            & $this_utf8n_flag_to_warn)
2061                                           ? 'with flag for raising warnings'
2062                                           : 'no flag for raising warnings');
2063
2064                     $this_name = "$uvchr_func() $testname: "
2065                                         . (($disallowed)
2066                                            ? 'disallowed'
2067                                            : 'allowed');
2068                     $this_name .= ", $eval_warn";
2069                     $this_name .= ", " . ((  $this_warning_flags
2070                                            & $this_uvchr_flag_to_warn)
2071                                         ? 'with warning flag'
2072                                         : 'no warning flag');
2073
2074                     undef @warnings_gotten;
2075                     my $ret;
2076                     $this_flags = $this_warning_flags|$this_disallow_flags;
2077                     $eval_text = "$eval_warn; \$ret ="
2078                             . " test_$uvchr_func("
2079                             . "$allowed_uv, $this_flags)";
2080                     eval "$eval_text";
2081                     if (! ok ($@ eq "", "$this_name: eval succeeded"))
2082                     {
2083                         diag "\$@='$@'; call was: "
2084                            . uvchr_display_call($eval_text);
2085                         next;
2086                     }
2087
2088                     if ($which_func) {
2089                         if (defined $ret->[1]) {
2090                             my @returned_warnings;
2091                             push @returned_warnings, $ret->[1]{'text'};
2092                             my $text = $ret->[1]{'text'};
2093                             my $flag = $ret->[1]{'flag_bit'};
2094                             my $category = $ret->[1]{'warning_category'};
2095
2096                             if (! ok(($flag & ($flag-1)) == 0,
2097                                         "flag for returned msg is a single bit"))
2098                             {
2099                                 diag sprintf("flags are %x; msg=%s", $flag, $text);
2100                             }
2101                             else {
2102                                 if ($flag & $this_uvchr_flag_to_disallow) {
2103                                     pass("flag for returned msg is expected");
2104                                 }
2105                                 else {
2106                                     fail("flag ("
2107                                         . flags_to_text($flag, \@utf8n_flags_to_text)
2108                                         . ") for returned msg is expected");
2109                                 }
2110                             }
2111
2112                             # In perl space, don't know the category numbers
2113                             isnt($category, 0,
2114                                             "returned category for msg isn't 0");
2115
2116                             ok(@warnings_gotten == 0, "$uvchr_func raised no warnings;"
2117                                 . " the next tests are for ones in the returned"
2118                                 . " variable")
2119                                 or diag join "\n", "The unexpected warnings were:",
2120                                                                 @warnings_gotten;
2121                             @warnings_gotten = @returned_warnings;
2122                         }
2123
2124                         $ret = $ret->[0];
2125                     }
2126
2127                     if ($disallowed) {
2128                         is($ret, undef, "    And returns undef")
2129                           or diag "Call was: " . uvchr_display_call($eval_text);
2130                     }
2131                     else {
2132                         is($ret, $this_bytes, "    And returns expected string")
2133                           or diag "Call was: " . uvchr_display_call($eval_text);
2134                     }
2135
2136                     do_warnings_test(@expected_warnings)
2137                       or diag "Call was: " . uvchr_display_call($eval_text);
2138                 }
2139               }
2140             }
2141           }
2142         }
2143       }
2144     }
2145   }
2146 }
2147
2148 done_testing;