This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
APItest:t/utf8_warn_base.pl: Clarify some comments
[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 # This test is split into this number of files.
703 my $num_test_files = $ENV{TEST_JOBS} || 1;
704 $num_test_files = 10 if $num_test_files > 10;
705
706 # We only really need to test utf8n_to_uvchr_msgs() once with this flag.
707 my $tested_CHECK_ONLY = 0;
708
709 my $test_count = -1;
710 foreach my $test (@tests) {
711   $test_count++;
712   next if $test_count % $num_test_files != $::TEST_CHUNK;
713
714   my ($testname, $bytes, $allowed_uv, $needed_to_discern_len) = @$test;
715
716   my $length = length $bytes;
717   my $initially_overlong = $testname =~ /overlong/;
718   my $initially_orphan   = $testname =~ /orphan/;
719   my $will_overflow = $allowed_uv < 0;
720
721   my $uv_string = sprintf(($allowed_uv < 0x100) ? "%02X" : "%04X", $allowed_uv);
722   my $display_bytes = display_bytes($bytes);
723
724   my $controlling_warning_category;
725   my $utf8n_flag_to_warn;
726   my $utf8n_flag_to_disallow;
727   my $uvchr_flag_to_warn;
728   my $uvchr_flag_to_disallow;
729
730   # We want to test that the independent flags are actually independent.
731   # For example, that a surrogate doesn't trigger a non-character warning,
732   # and conversely, turning off an above-Unicode flag doesn't suppress a
733   # surrogate warning.  Earlier versions of this file used nested loops to
734   # test all possible combinations.  But that creates lots of tests, making
735   # this run too long.  What is now done instead is to use the complement of
736   # the category we are testing to greatly reduce the combinatorial
737   # explosion.  For example, if we have a surrogate and we aren't expecting
738   # a warning about it, we set all the flags for non-surrogates to raise
739   # warnings.  If one shows up, it indicates the flags aren't independent.
740   my $utf8n_flag_to_warn_complement;
741   my $utf8n_flag_to_disallow_complement;
742   my $uvchr_flag_to_warn_complement;
743   my $uvchr_flag_to_disallow_complement;
744
745   # Many of the code points being tested are middling in that if code point
746   # edge cases work, these are very likely to as well.  Because this test
747   # file takes a while to execute, we skip testing the edge effects of code
748   # points deemed middling, while testing their basics and continuing to
749   # fully test the non-middling code points.
750   my $skip_most_tests = 0;
751
752   my $cp_message_qr;      # Pattern that matches the message raised when
753                           # that message contains the problematic code
754                           # point.  The message is the same (currently) both
755                           # when going from/to utf8.
756   my $non_cp_trailing_text;   # The suffix text when the message doesn't
757                               # contain a code point.  (This is a result of
758                               # some sort of malformation that means we
759                               # can't get an exact code poin
760   my $extended_cp_message_qr = qr/\QCode point 0x$uv_string is not Unicode,\E
761                       \Q requires a Perl extension, and so is not\E
762                       \Q portable\E/x;
763   my $extended_non_cp_trailing_text
764                       = "is a Perl extension, and so is not portable";
765
766   # What bytes should have been used to specify a code point that has been
767   # specified as an overlong.
768   my $correct_bytes_for_overlong;
769
770   # Is this test malformed from the beginning?  If so, we know to generally
771   # expect that the tests will show it isn't valid.
772   my $initially_malformed = 0;
773
774   if ($initially_overlong || $initially_orphan) {
775       $non_cp_trailing_text = "if you see this, there is an error";
776       $cp_message_qr = qr/\Q$non_cp_trailing_text\E/;
777       $initially_malformed = 1;
778       $utf8n_flag_to_warn     = 0;
779       $utf8n_flag_to_disallow = 0;
780
781       $utf8n_flag_to_warn_complement =     $::UTF8_WARN_SURROGATE;
782       $utf8n_flag_to_disallow_complement = $::UTF8_DISALLOW_SURROGATE;
783       if (! $will_overflow && $allowed_uv <= 0x10FFFF) {
784           $utf8n_flag_to_warn_complement     |= $::UTF8_WARN_SUPER;
785           $utf8n_flag_to_disallow_complement |= $::UTF8_DISALLOW_SUPER;
786           if (($allowed_uv & 0xFFFF) != 0xFFFF) {
787               $utf8n_flag_to_warn_complement      |= $::UTF8_WARN_NONCHAR;
788               $utf8n_flag_to_disallow_complement  |= $::UTF8_DISALLOW_NONCHAR;
789           }
790       }
791       if (! is_extended_utf8($bytes)) {
792           $utf8n_flag_to_warn_complement |= $::UTF8_WARN_PERL_EXTENDED;
793           $utf8n_flag_to_disallow_complement  |= $::UTF8_DISALLOW_PERL_EXTENDED;
794       }
795
796       $controlling_warning_category = 'utf8';
797
798       if ($initially_overlong) {
799           if (! defined $needed_to_discern_len) {
800               $needed_to_discern_len = overlong_discern_len($bytes);
801           }
802           $correct_bytes_for_overlong = display_bytes_no_quotes(chr $allowed_uv);
803       }
804   }
805   elsif($will_overflow || $allowed_uv > 0x10FFFF) {
806
807       # Set the SUPER flags; later, we test for PERL_EXTENDED as well.
808       $utf8n_flag_to_warn     = $::UTF8_WARN_SUPER;
809       $utf8n_flag_to_disallow = $::UTF8_DISALLOW_SUPER;
810       $uvchr_flag_to_warn     = $::UNICODE_WARN_SUPER;
811       $uvchr_flag_to_disallow = $::UNICODE_DISALLOW_SUPER;;
812
813       # Below, we add the flags for non-perl_extended to the code points
814       # that don't fit that category.  Special tests are done for this
815       # category in the inner loop.
816       $utf8n_flag_to_warn_complement     = $::UTF8_WARN_NONCHAR
817                                           |$::UTF8_WARN_SURROGATE;
818       $utf8n_flag_to_disallow_complement = $::UTF8_DISALLOW_NONCHAR
819                                           |$::UTF8_DISALLOW_SURROGATE;
820       $uvchr_flag_to_warn_complement     = $::UNICODE_WARN_NONCHAR
821                                           |$::UNICODE_WARN_SURROGATE;
822       $uvchr_flag_to_disallow_complement = $::UNICODE_DISALLOW_NONCHAR
823                                           |$::UNICODE_DISALLOW_SURROGATE;
824       $controlling_warning_category = 'non_unicode';
825
826       if ($will_overflow) {  # This is realy a malformation
827           $non_cp_trailing_text = "if you see this, there is an error";
828           $cp_message_qr = qr/\Q$non_cp_trailing_text\E/;
829           $initially_malformed = 1;
830           if (! defined $needed_to_discern_len) {
831               $needed_to_discern_len = overflow_discern_len($length);
832           }
833       }
834       elsif (requires_extended_utf8($allowed_uv)) {
835           $cp_message_qr = $extended_cp_message_qr;
836           $non_cp_trailing_text = $extended_non_cp_trailing_text;
837           $needed_to_discern_len = 1 unless defined $needed_to_discern_len;
838       }
839       else {
840           $cp_message_qr = qr/\QCode point 0x$uv_string is not Unicode,\E
841                               \Q may not be portable\E/x;
842           $non_cp_trailing_text = "is for a non-Unicode code point, may not"
843                               . " be portable";
844           $utf8n_flag_to_warn_complement     |= $::UTF8_WARN_PERL_EXTENDED;
845           $utf8n_flag_to_disallow_complement
846                                           |= $::UTF8_DISALLOW_PERL_EXTENDED;
847           $uvchr_flag_to_warn_complement |= $::UNICODE_WARN_PERL_EXTENDED;
848           $uvchr_flag_to_disallow_complement
849                                       |= $::UNICODE_DISALLOW_PERL_EXTENDED;
850       }
851   }
852   elsif ($allowed_uv >= 0xD800 && $allowed_uv <= 0xDFFF) {
853       $cp_message_qr = qr/UTF-16 surrogate U\+$uv_string/;
854       $non_cp_trailing_text = "is for a surrogate";
855       $needed_to_discern_len = 2 unless defined $needed_to_discern_len;
856       $skip_most_tests = 1 if $allowed_uv > 0xD800 && $allowed_uv < 0xDFFF;
857
858       $utf8n_flag_to_warn     = $::UTF8_WARN_SURROGATE;
859       $utf8n_flag_to_disallow = $::UTF8_DISALLOW_SURROGATE;
860       $uvchr_flag_to_warn     = $::UNICODE_WARN_SURROGATE;
861       $uvchr_flag_to_disallow = $::UNICODE_DISALLOW_SURROGATE;;
862
863       $utf8n_flag_to_warn_complement     = $::UTF8_WARN_NONCHAR
864                                           |$::UTF8_WARN_SUPER
865                                           |$::UTF8_WARN_PERL_EXTENDED;
866       $utf8n_flag_to_disallow_complement = $::UTF8_DISALLOW_NONCHAR
867                                           |$::UTF8_DISALLOW_SUPER
868                                           |$::UTF8_DISALLOW_PERL_EXTENDED;
869       $uvchr_flag_to_warn_complement     = $::UNICODE_WARN_NONCHAR
870                                           |$::UNICODE_WARN_SUPER
871                                           |$::UNICODE_WARN_PERL_EXTENDED;
872       $uvchr_flag_to_disallow_complement = $::UNICODE_DISALLOW_NONCHAR
873                                           |$::UNICODE_DISALLOW_SUPER
874                                           |$::UNICODE_DISALLOW_PERL_EXTENDED;
875       $controlling_warning_category = 'surrogate';
876   }
877   elsif (   ($allowed_uv >= 0xFDD0 && $allowed_uv <= 0xFDEF)
878           || ($allowed_uv & 0xFFFE) == 0xFFFE)
879   {
880       $cp_message_qr = qr/\QUnicode non-character U+$uv_string\E
881                           \Q is not recommended for open interchange\E/x;
882       $non_cp_trailing_text = "if you see this, there is an error";
883       $needed_to_discern_len = $length unless defined $needed_to_discern_len;
884       if (   ($allowed_uv > 0xFDD0 && $allowed_uv < 0xFDEF)
885           || ($allowed_uv > 0xFFFF && $allowed_uv < 0x10FFFE))
886       {
887           $skip_most_tests = 1;
888       }
889
890       $utf8n_flag_to_warn     = $::UTF8_WARN_NONCHAR;
891       $utf8n_flag_to_disallow = $::UTF8_DISALLOW_NONCHAR;
892       $uvchr_flag_to_warn     = $::UNICODE_WARN_NONCHAR;
893       $uvchr_flag_to_disallow = $::UNICODE_DISALLOW_NONCHAR;;
894
895       $utf8n_flag_to_warn_complement     = $::UTF8_WARN_SURROGATE
896                                           |$::UTF8_WARN_SUPER
897                                           |$::UTF8_WARN_PERL_EXTENDED;
898       $utf8n_flag_to_disallow_complement = $::UTF8_DISALLOW_SURROGATE
899                                           |$::UTF8_DISALLOW_SUPER
900                                           |$::UTF8_DISALLOW_PERL_EXTENDED;
901       $uvchr_flag_to_warn_complement     = $::UNICODE_WARN_SURROGATE
902                                           |$::UNICODE_WARN_SUPER
903                                           |$::UNICODE_WARN_PERL_EXTENDED;
904       $uvchr_flag_to_disallow_complement = $::UNICODE_DISALLOW_SURROGATE
905                                           |$::UNICODE_DISALLOW_SUPER
906                                           |$::UNICODE_DISALLOW_PERL_EXTENDED;
907
908       $controlling_warning_category = 'nonchar';
909   }
910   else {
911       die "Can't figure out what type of warning to test for $testname"
912   }
913
914   die 'Didn\'t set $needed_to_discern_len for ' . $testname
915                                       unless defined $needed_to_discern_len;
916
917   # We try various combinations of malformations that can occur
918   foreach my $short (0, 1) {
919     next if $skip_most_tests && $short;
920     foreach my $unexpected_noncont (0, 1) {
921       next if $skip_most_tests && $unexpected_noncont;
922       foreach my $overlong (0, 1) {
923         next if $overlong && $skip_most_tests;
924         next if $initially_overlong && ! $overlong;
925
926         # If we're creating an overlong, it can't be longer than the
927         # maximum length, so skip if we're already at that length.
928         next if   (! $initially_overlong && $overlong)
929                   &&  $length >= $::max_bytes;
930
931         my $this_cp_message_qr = $cp_message_qr;
932         my $this_non_cp_trailing_text = $non_cp_trailing_text;
933
934         foreach my $malformed_allow_type (0..2) {
935           # 0 don't allow this malformation; ignored if no malformation
936           # 1 allow, with REPLACEMENT CHARACTER returned
937           # 2 allow, with intended code point returned.  All malformations
938           #   other than overlong can't determine the intended code point,
939           #   so this isn't valid for them.
940           next if     $malformed_allow_type == 2
941                   && ($will_overflow || $short || $unexpected_noncont);
942           next if $skip_most_tests && $malformed_allow_type;
943
944           # Here we are in the innermost loop for malformations.  So we
945           # know which ones are in effect.  Can now change the input to be
946           # appropriately malformed.  We also can set up certain other
947           # things now, like whether we expect a return flag from this
948           # malformation, and which flag.
949
950           my $this_bytes = $bytes;
951           my $this_length = $length;
952           my $this_expected_len = $length;
953           my $this_needed_to_discern_len = $needed_to_discern_len;
954
955           my @malformation_names;
956           my @expected_malformation_warnings;
957           my @expected_malformation_return_flags;
958
959           # Contains the flags for any allowed malformations.  Currently no
960           # combinations of on/off are tested for.  It's either all are
961           # allowed, or none are.
962           my $allow_flags = 0;
963           my $overlong_is_in_perl_extended_utf8 = 0;
964           my $dont_use_overlong_cp = 0;
965
966           if ($initially_orphan) {
967               next if $overlong || $short || $unexpected_noncont;
968           }
969
970           if ($overlong) {
971               if (! $initially_overlong) {
972                   my $new_expected_len;
973
974                   # To force this malformation, we convert the original start
975                   # byte into a continuation byte with the same data bits as
976                   # originally. ...
977                   my $start_byte = substr($this_bytes, 0, 1);
978                   my $converted_to_continuation_byte
979                                           = start_byte_to_cont($start_byte);
980
981                   # ... Then we prepend it with a known overlong sequence.
982                   # This should evaluate to the exact same code point as the
983                   # original.  We try to avoid an overlong using Perl
984                   # extended UTF-8.  The code points are the highest
985                   # representable as overlongs on the respective platform
986                   # without using extended UTF-8.
987                   if (native_to_I8($start_byte) lt "\xFC") {
988                       $start_byte = I8_to_native("\xFC");
989                       $new_expected_len = 6;
990                   }
991                   elsif (! isASCII && native_to_I8($start_byte) lt "\xFE") {
992
993                       # FE is not extended UTF-8 on EBCDIC
994                       $start_byte = I8_to_native("\xFE");
995                       $new_expected_len = 7;
996                   }
997                   else {  # Must use extended UTF-8.  On ASCII platforms, we
998                           # could express some overlongs here starting with
999                           # \xFE, but there's no real reason to do so.
1000                       $overlong_is_in_perl_extended_utf8 = 1;
1001                       $start_byte = I8_to_native("\xFF");
1002                       $new_expected_len = $::max_bytes;
1003                       $this_cp_message_qr = $extended_cp_message_qr;
1004
1005                       # The warning that gets raised doesn't include the
1006                       # code point in the message if the code point can be
1007                       # expressed without using extended UTF-8, but the
1008                       # particular overlong sequence used is in extended
1009                       # UTF-8.  To do otherwise would be confusing to the
1010                       # user, as it would claim the code point requires
1011                       # extended, when it doesn't.
1012                       $dont_use_overlong_cp = 1
1013                                   unless requires_extended_utf8($allowed_uv);
1014                       $this_non_cp_trailing_text
1015                                             = $extended_non_cp_trailing_text;
1016                   }
1017
1018                   # Splice in the revise continuation byte, preceded by the
1019                   # start byte and the proper number of the lowest
1020                   # continuation bytes.
1021                   $this_bytes =   $start_byte
1022                               . ($native_lowest_continuation_chr
1023                                   x (  $new_expected_len
1024                                       - 1
1025                                       - length($this_bytes)))
1026                               .  $converted_to_continuation_byte
1027                               .  substr($this_bytes, 1);
1028                   $this_length = length($this_bytes);
1029                   $this_needed_to_discern_len =    $new_expected_len
1030                                               - (  $this_expected_len
1031                                               - $this_needed_to_discern_len);
1032                   $this_expected_len = $new_expected_len;
1033               }
1034           }
1035
1036           if ($short) {
1037
1038               # To force this malformation, just tell the test to not look
1039               # as far as it should into the input.
1040               $this_length--;
1041               $this_expected_len--;
1042
1043               $allow_flags |= $::UTF8_ALLOW_SHORT if $malformed_allow_type;
1044           }
1045
1046           if ($unexpected_noncont) {
1047
1048               # To force this malformation, change the final continuation
1049               # byte into a start byte.
1050               my $pos = ($short) ? -2 : -1;
1051               substr($this_bytes, $pos, 1) = $known_start_byte;
1052               $this_expected_len--;
1053           }
1054
1055           # The whole point of a test that is malformed from the beginning
1056           # is to test for that malformation.  If we've modified things so
1057           # much that we don't have enough information to detect that
1058           # malformation, there's no point in testing.
1059           next if    $initially_malformed
1060                   && $this_expected_len < $this_needed_to_discern_len;
1061
1062           # Here, we've transformed the input with all of the desired
1063           # non-overflow malformations.  We are now in a position to
1064           # construct any potential warnings for those malformations.  But
1065           # it's a pain to get the detailed messages exactly right, so for
1066           # now XXX, only do so for those that return an explicit code
1067           # point.
1068
1069           if ($initially_orphan) {
1070               push @malformation_names, "orphan continuation";
1071               push @expected_malformation_return_flags,
1072                                                   $::UTF8_GOT_CONTINUATION;
1073               $allow_flags |= $::UTF8_ALLOW_CONTINUATION
1074                                                   if $malformed_allow_type;
1075               push @expected_malformation_warnings, qr/unexpected continuation/;
1076           }
1077
1078           if ($overlong) {
1079               push @malformation_names, 'overlong';
1080               push @expected_malformation_return_flags, $::UTF8_GOT_LONG;
1081
1082               # If one of the other malformation types is also in effect, we
1083               # don't know what the intended code point was.
1084               if ($short || $unexpected_noncont || $will_overflow) {
1085                   push @expected_malformation_warnings, qr/overlong/;
1086               }
1087               else {
1088                   my $wrong_bytes = display_bytes_no_quotes(
1089                                         substr($this_bytes, 0, $this_length));
1090                   if (! defined $correct_bytes_for_overlong) {
1091                       $correct_bytes_for_overlong
1092                                           = display_bytes_no_quotes($bytes);
1093                   }
1094                   my $prefix = (   $allowed_uv > 0x10FFFF
1095                                 || ! isASCII && $allowed_uv < 256)
1096                                 ? "0x"
1097                                 : "U+";
1098                   push @expected_malformation_warnings,
1099                           qr/\QMalformed UTF-8 character: $wrong_bytes\E
1100                               \Q (overlong; instead use\E
1101                               \Q $correct_bytes_for_overlong to\E
1102                               \Q represent $prefix$uv_string)/x;
1103               }
1104
1105               if ($malformed_allow_type == 2) {
1106                   $allow_flags |= $::UTF8_ALLOW_LONG_AND_ITS_VALUE;
1107               }
1108               elsif ($malformed_allow_type) {
1109                   $allow_flags |= $::UTF8_ALLOW_LONG;
1110               }
1111           }
1112           if ($short) {
1113               push @malformation_names, 'short';
1114               push @expected_malformation_return_flags, $::UTF8_GOT_SHORT;
1115               push @expected_malformation_warnings, qr/too short/;
1116           }
1117           if ($unexpected_noncont) {
1118               push @malformation_names, 'unexpected non-continuation';
1119               push @expected_malformation_return_flags,
1120                               $::UTF8_GOT_NON_CONTINUATION;
1121               $allow_flags |= $::UTF8_ALLOW_NON_CONTINUATION
1122                                                   if $malformed_allow_type;
1123               push @expected_malformation_warnings,
1124                                       qr/unexpected non-continuation byte/;
1125           }
1126
1127           # The overflow malformation is done differently than other
1128           # malformations.  It comes from manually typed tests in the test
1129           # array.  We now make it be treated like one of the other
1130           # malformations.  But some has to be deferred until the inner loop
1131           my $overflow_msg_pattern;
1132           if ($will_overflow) {
1133               push @malformation_names, 'overflow';
1134
1135               $overflow_msg_pattern = display_bytes_no_quotes(
1136                                   substr($this_bytes, 0, $this_expected_len));
1137               $overflow_msg_pattern = qr/\QMalformed UTF-8 character:\E
1138                                           \Q $overflow_msg_pattern\E
1139                                           \Q (overflows)\E/x;
1140               push @expected_malformation_return_flags, $::UTF8_GOT_OVERFLOW;
1141               $allow_flags |= $::UTF8_ALLOW_OVERFLOW if $malformed_allow_type;
1142           }
1143
1144           # And we can create the malformation-related text for the the test
1145           # names we eventually will generate.
1146           my $malformations_name = "";
1147           if (@malformation_names) {
1148               $malformations_name .= "dis" unless $malformed_allow_type;
1149               $malformations_name .= "allowed ";
1150               $malformations_name .= "malformation";
1151               $malformations_name .= "s" if @malformation_names > 1;
1152               $malformations_name .= ": ";
1153               $malformations_name .=  join "/", @malformation_names;
1154               $malformations_name =  " ($malformations_name)";
1155           }
1156
1157           # Done setting up the malformation related stuff
1158
1159           {   # First test the isFOO calls
1160               use warnings; # XXX no warnings 'deprecated';   # Make sure these don't raise warnings
1161               undef @warnings_gotten;
1162
1163               my $ret = test_isUTF8_CHAR($this_bytes, $this_length);
1164               my $ret_flags
1165                       = test_isUTF8_CHAR_flags($this_bytes, $this_length, 0);
1166               if ($malformations_name) {
1167                   is($ret, 0, "For $testname$malformations_name: isUTF8_CHAR() returns 0");
1168                   is($ret_flags, 0, "    And isUTF8_CHAR_flags() returns 0");
1169               }
1170               else {
1171                   is($ret, $this_length, "For $testname: isUTF8_CHAR() returns"
1172                                         . " expected length: $this_length");
1173                   is($ret_flags, $this_length,
1174                       "    And isUTF8_CHAR_flags(...,0) returns expected"
1175                     . " length: $this_length");
1176               }
1177               is(scalar @warnings_gotten, 0,
1178                   "    And neither isUTF8_CHAR() nor isUTF8_CHAR()_flags"
1179                 . " generated any warnings")
1180               or output_warnings(@warnings_gotten);
1181
1182               undef @warnings_gotten;
1183               $ret = test_isSTRICT_UTF8_CHAR($this_bytes, $this_length);
1184               if ($malformations_name) {
1185                   is($ret, 0, "    And isSTRICT_UTF8_CHAR() returns 0");
1186               }
1187               else {
1188                   my $expected_ret
1189                               = (   $testname =~ /surrogate|non-character/
1190                                   || $allowed_uv > 0x10FFFF)
1191                                 ? 0
1192                                 : $this_length;
1193                   is($ret, $expected_ret,
1194                       "    And isSTRICT_UTF8_CHAR() returns expected"
1195                     . " length: $expected_ret");
1196                   $ret = test_isUTF8_CHAR_flags($this_bytes, $this_length,
1197                                       $::UTF8_DISALLOW_ILLEGAL_INTERCHANGE);
1198                   is($ret, $expected_ret,
1199                       "    And isUTF8_CHAR_flags('"
1200                     . "DISALLOW_ILLEGAL_INTERCHANGE') acts like"
1201                     . " isSTRICT_UTF8_CHAR");
1202               }
1203               is(scalar @warnings_gotten, 0,
1204                       "    And neither isSTRICT_UTF8_CHAR() nor"
1205                     . " isUTF8_CHAR_flags generated any warnings")
1206               or output_warnings(@warnings_gotten);
1207
1208               undef @warnings_gotten;
1209               $ret = test_isC9_STRICT_UTF8_CHAR($this_bytes, $this_length);
1210               if ($malformations_name) {
1211                   is($ret, 0, "    And isC9_STRICT_UTF8_CHAR() returns 0");
1212               }
1213               else {
1214                   my $expected_ret = (   $testname =~ /surrogate/
1215                                       || $allowed_uv > 0x10FFFF)
1216                                       ? 0
1217                                       : $this_expected_len;
1218                   is($ret, $expected_ret, "    And isC9_STRICT_UTF8_CHAR()"
1219                                         . " returns expected length:"
1220                                         . " $expected_ret");
1221                   $ret = test_isUTF8_CHAR_flags($this_bytes, $this_length,
1222                                   $::UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE);
1223                   is($ret, $expected_ret,
1224                       "    And isUTF8_CHAR_flags('"
1225                     . "DISALLOW_ILLEGAL_C9_INTERCHANGE') acts like"
1226                     . " isC9_STRICT_UTF8_CHAR");
1227               }
1228               is(scalar @warnings_gotten, 0,
1229                       "    And neither isC9_STRICT_UTF8_CHAR() nor"
1230                     . " isUTF8_CHAR_flags generated any warnings")
1231               or output_warnings(@warnings_gotten);
1232
1233               foreach my $disallow_type (0..2) {
1234                   # 0 is don't disallow this type of code point
1235                   # 1 is do disallow
1236                   # 2 is do disallow, but only code points requiring
1237                   #   perl-extended-UTF8
1238
1239                   my $disallow_flags;
1240                   my $expected_ret;
1241
1242                   if ($malformations_name) {
1243
1244                       # Malformations are by default disallowed, so testing
1245                       # with $disallow_type equal to 0 is sufficicient.
1246                       next if $disallow_type;
1247
1248                       $disallow_flags = 0;
1249                       $expected_ret = 0;
1250                   }
1251                   elsif ($disallow_type == 1) {
1252                       $disallow_flags = $utf8n_flag_to_disallow;
1253                       $expected_ret = 0;
1254                   }
1255                   elsif ($disallow_type == 2) {
1256                       next if ! requires_extended_utf8($allowed_uv);
1257                       $disallow_flags = $::UTF8_DISALLOW_PERL_EXTENDED;
1258                       $expected_ret = 0;
1259                   }
1260                   else {  # type is 0
1261                       $disallow_flags = $utf8n_flag_to_disallow_complement;
1262                       $expected_ret = $this_length;
1263                   }
1264
1265                   $ret = test_isUTF8_CHAR_flags($this_bytes, $this_length,
1266                                                 $disallow_flags);
1267                   is($ret, $expected_ret,
1268                             "    And isUTF8_CHAR_flags($display_bytes,"
1269                           . " $disallow_flags) returns $expected_ret")
1270                     or diag "The flags mean "
1271                           . flags_to_text($disallow_flags,
1272                                           \@utf8n_flags_to_text);
1273                   is(scalar @warnings_gotten, 0,
1274                           "    And isUTF8_CHAR_flags(...) generated"
1275                         . " no warnings")
1276                     or output_warnings(@warnings_gotten);
1277
1278                   # Test partial character handling, for each byte not a
1279                   # full character
1280                   my $did_test_partial = 0;
1281                   for (my $j = 1; $j < $this_length - 1; $j++) {
1282                       $did_test_partial = 1;
1283                       my $partial = substr($this_bytes, 0, $j);
1284                       my $ret_should_be;
1285                       my $comment;
1286                       if ($disallow_type || $malformations_name) {
1287                           $ret_should_be = 0;
1288                           $comment = "disallowed";
1289
1290                           # The number of bytes required to tell if a
1291                           # sequence has something wrong is the smallest of
1292                           # all the things wrong with it.  We start with the
1293                           # number for this type of code point, if that is
1294                           # disallowed; or the whole length if not.  The
1295                           # latter is what a couple of the malformations
1296                           # require.
1297                           my $needed_to_tell = ($disallow_type)
1298                                                 ? $this_needed_to_discern_len
1299                                                 : $this_expected_len;
1300
1301                           # Then we see if the malformations that are
1302                           # detectable early in the string are present.
1303                           if ($overlong) {
1304                               my $dl = overlong_discern_len($this_bytes);
1305                               $needed_to_tell = $dl if $dl < $needed_to_tell;
1306                           }
1307                           if ($will_overflow) {
1308                               my $dl = overflow_discern_len($length);
1309                               $needed_to_tell = $dl if $dl < $needed_to_tell;
1310                           }
1311
1312                           if ($j < $needed_to_tell) {
1313                               $ret_should_be = 1;
1314                               $comment .= ", but need $needed_to_tell"
1315                                         . " bytes to discern:";
1316                           }
1317                       }
1318                       else {
1319                           $ret_should_be = 1;
1320                           $comment = "allowed";
1321                       }
1322
1323                       undef @warnings_gotten;
1324
1325                       $ret = test_is_utf8_valid_partial_char_flags($partial,
1326                                                       $j, $disallow_flags);
1327                       is($ret, $ret_should_be,
1328                           "    And is_utf8_valid_partial_char_flags("
1329                           . display_bytes($partial)
1330                           . ", $disallow_flags), $comment: returns"
1331                           . " $ret_should_be")
1332                       or diag "The flags mean "
1333                       . flags_to_text($disallow_flags, \@utf8n_flags_to_text);
1334                   }
1335
1336                   if ($did_test_partial) {
1337                       is(scalar @warnings_gotten, 0,
1338                           "    And is_utf8_valid_partial_char_flags()"
1339                           . " generated no warnings for any of the lengths")
1340                         or output_warnings(@warnings_gotten);
1341                   }
1342               }
1343           }
1344
1345           # Now test the to/from UTF-8 calls.  There are several orthogonal
1346           # variables involved.  We test most possible combinations
1347
1348           foreach my $do_disallow (0, 1) {
1349             if ($do_disallow) {
1350               next if $initially_overlong || $initially_orphan;
1351             }
1352             else {
1353               next if $skip_most_tests;
1354             }
1355
1356             # This tests three functions.  utf8n_to_uvchr_error,
1357             # utf8n_to_uvchr_msgs, and uvchr_to_utf8_flags.  But only the
1358             # first two are variants of each other.  We use a loop
1359             # 'which_func' to determine which of these.  uvchr_to_utf8_flags
1360             # is done separately at the end of each iteration, only when
1361             # which_func is 0.  which_func is numeric in part so we don't
1362             # have to type in the function name and risk misspelling it
1363             # somewhere, and also it sets whether we are expecting warnings
1364             # or not in certain places.  The _msgs() version of the function
1365             # expects warnings even if lexical ones are turned off, so by
1366             # making its which_func == 1, we can say we want warnings;
1367             # whereas the other one with the value 0, doesn't get them.
1368             for my $which_func (0, 1) {
1369               my $func = ($which_func)
1370                           ? 'utf8n_to_uvchr_msgs'
1371                           : 'utf8n_to_uvchr_error';
1372
1373               # We classify the warnings into certain "interesting" types,
1374               # described later
1375               foreach my $warning_type (0..4) {
1376                 next if $skip_most_tests && $warning_type != 1;
1377                 foreach my $use_warn_flag (0, 1) {
1378                     if ($use_warn_flag) {
1379                         next if $initially_overlong || $initially_orphan;
1380
1381                         # Since foo_msgs() expects warnings even when lexical
1382                         # ones are turned off, we can skip testing it when
1383                         # they are turned on, with little likelihood of
1384                         # missing an error case.
1385                         next if $which_func;
1386                     }
1387                     else {
1388                         next if $skip_most_tests;
1389                     }
1390
1391                     # Finally, here is the inner loop
1392
1393                     my $this_utf8n_flag_to_warn = $utf8n_flag_to_warn;
1394                     my $this_utf8n_flag_to_disallow = $utf8n_flag_to_disallow;
1395                     my $this_uvchr_flag_to_warn = $uvchr_flag_to_warn;
1396                     my $this_uvchr_flag_to_disallow = $uvchr_flag_to_disallow;
1397
1398                     my $eval_warn;
1399                     my $expect_regular_warnings;
1400                     my $expect_warnings_for_malformed;
1401                     my $expect_warnings_for_overflow;
1402
1403                     if ($warning_type == 0) {
1404                         $eval_warn = "use warnings";
1405                         $expect_regular_warnings = $use_warn_flag;
1406
1407                         # We ordinarily expect overflow warnings here.  But it
1408                         # is somewhat more complicated, and the final
1409                         # determination is deferred to one place in the file
1410                         # where we handle overflow.
1411                         $expect_warnings_for_overflow = 1;
1412
1413                         # We would ordinarily expect malformed warnings in
1414                         # this case, but not if malformations are allowed.
1415                         $expect_warnings_for_malformed
1416                                                 = $malformed_allow_type == 0;
1417                     }
1418                     elsif ($warning_type == 1) {
1419                         $eval_warn = "no warnings";
1420                         $expect_regular_warnings = $which_func;
1421                         $expect_warnings_for_overflow = $which_func;
1422                         $expect_warnings_for_malformed = $which_func;
1423                     }
1424                     elsif ($warning_type == 2) {
1425                         $eval_warn = "no warnings; use warnings 'utf8'";
1426                         $expect_regular_warnings = $use_warn_flag;
1427                         $expect_warnings_for_overflow = 1;
1428                         $expect_warnings_for_malformed
1429                                                 = $malformed_allow_type == 0;
1430                     }
1431                     elsif ($warning_type == 3) {
1432                         $eval_warn = "no warnings; use warnings"
1433                                    . " '$controlling_warning_category'";
1434                         $expect_regular_warnings = $use_warn_flag;
1435                         $expect_warnings_for_overflow
1436                             = $controlling_warning_category eq 'non_unicode';
1437                         $expect_warnings_for_malformed = $which_func;
1438                     }
1439                     elsif ($warning_type == 4) {  # Like type 3, but uses the
1440                                                   # PERL_EXTENDED flags
1441                         # The complement flags were set up so that the
1442                         # PERL_EXTENDED flags have been tested that they don't
1443                         # trigger wrongly for too small code points.  And the
1444                         # flags have been set up so that those small code
1445                         # points are tested for being above Unicode.  What's
1446                         # left to test is that the large code points do
1447                         # trigger the PERL_EXTENDED flags.
1448                         next if ! requires_extended_utf8($allowed_uv);
1449                         next if $controlling_warning_category ne 'non_unicode';
1450                         $eval_warn = "no warnings; use warnings 'non_unicode'";
1451                         $expect_regular_warnings = 1;
1452                         $expect_warnings_for_overflow = 1;
1453                         $expect_warnings_for_malformed = 0;
1454                         $this_utf8n_flag_to_warn = $::UTF8_WARN_PERL_EXTENDED;
1455                         $this_utf8n_flag_to_disallow
1456                                              = $::UTF8_DISALLOW_PERL_EXTENDED;
1457                         $this_uvchr_flag_to_warn
1458                                               = $::UNICODE_WARN_PERL_EXTENDED;
1459                         $this_uvchr_flag_to_disallow
1460                                           = $::UNICODE_DISALLOW_PERL_EXTENDED;
1461                     }
1462                     else {
1463                        die "Unexpected warning type '$warning_type'";
1464                     }
1465
1466                     # We only need to test the case where all warnings are
1467                     # enabled (type 0) to see if turning off the warning flag
1468                     # causes things to not be output.  If those pass, then
1469                     # turning on some sub-category of warnings, or turning off
1470                     # warnings altogether are extremely likely to not output
1471                     # warnings either, given how the warnings subsystem is
1472                     # supposed to work, and this file assumes it does work.
1473                     next if $warning_type != 0 && ! $use_warn_flag;
1474
1475                     # The convention is that the 'got' flag is the same value
1476                     # as the disallow one.  If this were violated, the tests
1477                     # here should start failing.
1478                     my $return_flag = $this_utf8n_flag_to_disallow;
1479
1480                     # If we aren't expecting warnings/disallow for this, turn
1481                     # on all the other flags.  That makes sure that they all
1482                     # are independent of this flag, and so we don't need to
1483                     # test them individually.
1484                     my $this_warning_flags
1485                             = ($use_warn_flag)
1486                               ? $this_utf8n_flag_to_warn
1487                               : ($overlong_is_in_perl_extended_utf8
1488                                 ? ($utf8n_flag_to_warn_complement
1489                                     & ~$::UTF8_WARN_PERL_EXTENDED)
1490                                 :  $utf8n_flag_to_warn_complement);
1491                     my $this_disallow_flags
1492                             = ($do_disallow)
1493                               ? $this_utf8n_flag_to_disallow
1494                               : ($overlong_is_in_perl_extended_utf8
1495                                  ? ($utf8n_flag_to_disallow_complement
1496                                     & ~$::UTF8_DISALLOW_PERL_EXTENDED)
1497                                  :  $utf8n_flag_to_disallow_complement);
1498                     my $expected_uv = $allowed_uv;
1499                     my $this_uv_string = $uv_string;
1500
1501                     my @expected_return_flags
1502                                         = @expected_malformation_return_flags;
1503                     my @expected_warnings;
1504                     push @expected_warnings, @expected_malformation_warnings
1505                                             if $expect_warnings_for_malformed;
1506
1507                     # The overflow malformation is done differently than other
1508                     # malformations.  It comes from manually typed tests in
1509                     # the test array, but it also is above Unicode and uses
1510                     # Perl extended UTF-8, so affects some of the flags being
1511                     # tested.  We now make it be treated like one of the other
1512                     # generated malformations.
1513                     if ($will_overflow) {
1514
1515                         # An overflow is (way) above Unicode, and overrides
1516                         # everything else.
1517                         $expect_regular_warnings = 0;
1518
1519                         # Earlier, we tentatively calculated whether this
1520                         # should emit a message or not.  It's tentative
1521                         # because, even if we ordinarily would output it, we
1522                         # don't if malformations are allowed -- except an
1523                         # overflow is also a SUPER and PERL_EXTENDED, and if
1524                         # warnings for those are enabled, the overflow
1525                         # warning does get raised.
1526                         if (   $expect_warnings_for_overflow
1527                             && (    $malformed_allow_type == 0
1528                                 ||   (   $this_warning_flags
1529                                       & ($::UTF8_WARN_SUPER
1530                                         |$::UTF8_WARN_PERL_EXTENDED))))
1531                         {
1532                             push @expected_warnings, $overflow_msg_pattern;
1533                         }
1534                     }
1535
1536                     # It may be that the malformations have shortened the
1537                     # amount of input we look at so much that we can't tell
1538                     # what the category the code point was in.  Otherwise, set
1539                     # up the expected return flags based on the warnings and
1540                     # disallowments.
1541                     if ($this_expected_len < $this_needed_to_discern_len) {
1542                         $expect_regular_warnings = 0;
1543                     }
1544                     elsif (   ($this_warning_flags & $this_utf8n_flag_to_warn)
1545                            || (  $this_disallow_flags
1546                                & $this_utf8n_flag_to_disallow))
1547                     {
1548                         push @expected_return_flags, $return_flag;
1549                     }
1550
1551                     # Finish setting up the expected warning.
1552                     if ($expect_regular_warnings) {
1553
1554                         # So far the array contains warnings generated by
1555                         # malformations.  Add the expected regular one.
1556                         unshift @expected_warnings, $this_cp_message_qr;
1557
1558                         # But it may need to be modified, because either of
1559                         # these malformations means we can't determine the
1560                         # expected code point.
1561                         if (   $short || $unexpected_noncont
1562                             || $dont_use_overlong_cp)
1563                         {
1564                             my $first_byte = substr($this_bytes, 0, 1);
1565                             $expected_warnings[0] = display_bytes(
1566                                     substr($this_bytes, 0, $this_expected_len));
1567                             $expected_warnings[0]
1568                                 = qr/[Aa]\Qny UTF-8 sequence that starts with\E
1569                                      \Q $expected_warnings[0]\E
1570                                      \Q $this_non_cp_trailing_text\E/x;
1571                         }
1572                     }
1573
1574                     # Is effectively disallowed if we've set up a malformation
1575                     # (unless malformations are allowed), even if the flag
1576                     # indicates it is allowed.  Fix up test name to indicate
1577                     # this as well
1578                     my $disallowed = 0;
1579                     if (   $this_disallow_flags & $this_utf8n_flag_to_disallow
1580                         && $this_expected_len >= $this_needed_to_discern_len)
1581                     {
1582                         $disallowed = 1;
1583                     }
1584                     if ($malformations_name) {
1585                         if ($malformed_allow_type == 0) {
1586                             $disallowed = 1;
1587                         }
1588                         elsif ($malformed_allow_type == 1) {
1589
1590                             # Even if allowed, the malformation returns the
1591                             # REPLACEMENT CHARACTER.
1592                             $expected_uv = 0xFFFD;
1593                             $this_uv_string = "0xFFFD"
1594                         }
1595                     }
1596
1597                     my $this_name = "$func() $testname: ";
1598                     my @scratch_expected_return_flags = @expected_return_flags;
1599                     if (! $initially_malformed) {
1600                         $this_name .= ($disallowed)
1601                                        ? 'disallowed, '
1602                                        : 'allowed, ';
1603                     }
1604                     $this_name .= "$eval_warn";
1605                     $this_name .= ", " . ((  $this_warning_flags
1606                                             & $this_utf8n_flag_to_warn)
1607                                           ? 'with flag for raising warnings'
1608                                           : 'no flag for raising warnings');
1609                     $this_name .= $malformations_name;
1610
1611                     # Do the actual test using an eval
1612                     undef @warnings_gotten;
1613                     my $ret_ref;
1614                     my $this_flags
1615                         = $allow_flags|$this_warning_flags|$this_disallow_flags;
1616                     my $eval_text =      "$eval_warn; \$ret_ref"
1617                             . " = test_$func("
1618                             . "'$this_bytes', $this_length, $this_flags)";
1619                     eval "$eval_text";
1620                     if (! ok ($@ eq "", "$this_name: eval succeeded"))
1621                     {
1622                         diag "\$@='$@'; call was: "
1623                            . utf8n_display_call($eval_text);
1624                         next;
1625                     }
1626
1627                     if ($disallowed) {
1628                         is($ret_ref->[0], 0, "    And returns 0")
1629                           or diag "Call was: " . utf8n_display_call($eval_text);
1630                     }
1631                     else {
1632                         is($ret_ref->[0], $expected_uv,
1633                                 "    And returns expected uv: "
1634                               . $this_uv_string)
1635                           or diag "Call was: " . utf8n_display_call($eval_text);
1636                     }
1637                     is($ret_ref->[1], $this_expected_len,
1638                                         "    And returns expected length:"
1639                                       . " $this_expected_len")
1640                       or diag "Call was: " . utf8n_display_call($eval_text);
1641
1642                     my $returned_flags = $ret_ref->[2];
1643
1644                     for (my $i = @scratch_expected_return_flags - 1;
1645                          $i >= 0;
1646                          $i--)
1647                     {
1648                       if ($scratch_expected_return_flags[$i] & $returned_flags)
1649                       {
1650                           if ($scratch_expected_return_flags[$i]
1651                                               == $::UTF8_GOT_PERL_EXTENDED)
1652                           {
1653                               pass("    Expected and got return flag for"
1654                                   . " PERL_EXTENDED");
1655                           }
1656                                   # The first entries in this are
1657                                   # malformations
1658                           elsif ($i > @malformation_names - 1)  {
1659                               pass("    Expected and got return flag"
1660                                   . " for " . $controlling_warning_category);
1661                           }
1662                           else {
1663                               pass("    Expected and got return flag for "
1664                                   . $malformation_names[$i]
1665                                   . " malformation");
1666                           }
1667                           $returned_flags
1668                                       &= ~$scratch_expected_return_flags[$i];
1669                           splice @scratch_expected_return_flags, $i, 1;
1670                       }
1671                     }
1672
1673                     if (! is($returned_flags, 0,
1674                        "    Got no unexpected return flags"))
1675                     {
1676                         diag "The unexpected flags gotten were: "
1677                            . (flags_to_text($returned_flags,
1678                                             \@utf8n_flags_to_text)
1679                                 # We strip off any prefixes from the flag
1680                                 # names
1681                              =~ s/ \b [A-Z] _ //xgr);
1682                         diag "Call was: " . utf8n_display_call($eval_text);
1683                     }
1684
1685                     if (! is (scalar @scratch_expected_return_flags, 0,
1686                         "    Got all expected return flags"))
1687                     {
1688                         diag "The expected flags not gotten were: "
1689                            . (flags_to_text(eval join("|",
1690                                                 @scratch_expected_return_flags),
1691                                             \@utf8n_flags_to_text)
1692                                 # We strip off any prefixes from the flag
1693                                 # names
1694                              =~ s/ \b [A-Z] _ //xgr);
1695                         diag "Call was: " . utf8n_display_call($eval_text);
1696                     }
1697
1698                     if ($which_func) {
1699                         my @returned_warnings;
1700                         for my $element_ref (@{$ret_ref->[3]}) {
1701                             push @returned_warnings, $element_ref->{'text'};
1702                             my $text = $element_ref->{'text'};
1703                             my $flag = $element_ref->{'flag_bit'};
1704                             my $category = $element_ref->{'warning_category'};
1705
1706                             if (! ok(($flag & ($flag-1)) == 0,
1707                                       "flag for returned msg is a single bit"))
1708                             {
1709                               diag sprintf("flags are %x; msg=%s", $flag, $text);
1710                             }
1711                             else {
1712                               if (grep { $_ == $flag } @expected_return_flags) {
1713                                   pass("flag for returned msg is expected");
1714                               }
1715                               else {
1716                                   fail("flag for returned msg is expected: "
1717                                  . flags_to_text($flag, \@utf8n_flags_to_text));
1718                               }
1719                             }
1720
1721                             # In perl space, don't know the category numbers
1722                             isnt($category, 0,
1723                                           "returned category for msg isn't 0");
1724                         }
1725
1726                         ok(@warnings_gotten == 0, "$func raised no warnings;"
1727                               . " the next tests are for ones in the returned"
1728                               . " variable")
1729                             or diag join "\n", "The unexpected warnings were:",
1730                                                               @warnings_gotten;
1731                         @warnings_gotten = @returned_warnings;
1732                     }
1733
1734                     do_warnings_test(@expected_warnings)
1735                       or diag "Call was: " . utf8n_display_call($eval_text);
1736                     undef @warnings_gotten;
1737
1738                     # Check CHECK_ONLY results when the input is
1739                     # disallowed.  Do this when actually disallowed,
1740                     # not just when the $this_disallow_flags is set.  We only
1741                     # test once utf8n_to_uvchr_msgs() with this.
1742                     if (   $disallowed
1743                         && ($which_func == 0 || ! $tested_CHECK_ONLY))
1744                     {
1745                         $tested_CHECK_ONLY = 1;
1746                         my $this_flags = $this_disallow_flags|$::UTF8_CHECK_ONLY;
1747                         my $eval_text = "use warnings; \$ret_ref ="
1748                                       . " test_$func('"
1749                                       . "$this_bytes', $this_length,"
1750                                       . " $this_flags)";
1751                         eval $eval_text;
1752                         if (! ok ($@ eq "",
1753                             "    And eval succeeded with CHECK_ONLY"))
1754                         {
1755                             diag "\$@='$@'; Call was: "
1756                                . utf8n_display_call($eval_text);
1757                             next;
1758                         }
1759                         is($ret_ref->[0], 0, "    CHECK_ONLY: Returns 0")
1760                           or diag "Call was: " . utf8n_display_call($eval_text);
1761                         is($ret_ref->[1], -1,
1762                                        "    CHECK_ONLY: returns -1 for length")
1763                           or diag "Call was: " . utf8n_display_call($eval_text);
1764                         if (! is(scalar @warnings_gotten, 0,
1765                                       "    CHECK_ONLY: no warnings generated"))
1766                         {
1767                             diag "Call was: " . utf8n_display_call($eval_text);
1768                             output_warnings(@warnings_gotten);
1769                         }
1770                     }
1771
1772                     # Now repeat some of the above, but for
1773                     # uvchr_to_utf8_flags().  Since this comes from an
1774                     # existing code point, it hasn't overflowed, and isn't
1775                     # malformed.
1776                     next if @malformation_names;
1777                     next if $which_func;
1778
1779                     $this_warning_flags = ($use_warn_flag)
1780                                           ? $this_uvchr_flag_to_warn
1781                                           : 0;
1782                     $this_disallow_flags = ($do_disallow)
1783                                            ? $this_uvchr_flag_to_disallow
1784                                            : 0;
1785
1786                     $disallowed = $this_disallow_flags
1787                                 & $this_uvchr_flag_to_disallow;
1788                     $this_name .= ", " . ((  $this_warning_flags
1789                                            & $this_utf8n_flag_to_warn)
1790                                           ? 'with flag for raising warnings'
1791                                           : 'no flag for raising warnings');
1792
1793                     $this_name = "uvchr_to_utf8_flags() $testname: "
1794                                             . (($disallowed)
1795                                                 ? 'disallowed'
1796                                                 : 'allowed');
1797                     $this_name .= ", $eval_warn";
1798                     $this_name .= ", " . ((  $this_warning_flags
1799                                            & $this_uvchr_flag_to_warn)
1800                                         ? 'with warning flag'
1801                                         : 'no warning flag');
1802
1803                     undef @warnings_gotten;
1804                     my $ret;
1805                     $this_flags = $this_warning_flags|$this_disallow_flags;
1806                     $eval_text = "$eval_warn; \$ret ="
1807                             . " test_uvchr_to_utf8_flags("
1808                             . "$allowed_uv, $this_flags)";
1809                     eval "$eval_text";
1810                     if (! ok ($@ eq "", "$this_name: eval succeeded"))
1811                     {
1812                         diag "\$@='$@'; call was: "
1813                            . uvchr_display_call($eval_text);
1814                         next;
1815                     }
1816                     if ($disallowed) {
1817                         is($ret, undef, "    And returns undef")
1818                           or diag "Call was: " . uvchr_display_call($eval_text);
1819                     }
1820                     else {
1821                         is($ret, $this_bytes, "    And returns expected string")
1822                           or diag "Call was: " . uvchr_display_call($eval_text);
1823                     }
1824
1825                     do_warnings_test(@expected_warnings)
1826                       or diag "Call was: " . uvchr_display_call($eval_text);
1827                 }
1828               }
1829             }
1830           }
1831         }
1832       }
1833     }
1834   }
1835 }
1836
1837 done_testing;