Commit | Line | Data |
---|---|---|
25468daa FC |
1 | /* inline.h |
2 | * | |
3 | * Copyright (C) 2012 by Larry Wall and others | |
4 | * | |
5 | * You may distribute under the terms of either the GNU General Public | |
6 | * License or the Artistic License, as specified in the README file. | |
7 | * | |
8ed185f9 | 8 | * This file contains tables and code adapted from |
f6521f7c | 9 | * https://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which requires this |
8ed185f9 KW |
10 | * copyright notice: |
11 | ||
12 | Copyright (c) 2008-2009 Bjoern Hoehrmann <bjoern@hoehrmann.de> | |
13 | ||
14 | Permission is hereby granted, free of charge, to any person obtaining a copy of | |
15 | this software and associated documentation files (the "Software"), to deal in | |
16 | the Software without restriction, including without limitation the rights to | |
17 | use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies | |
18 | of the Software, and to permit persons to whom the Software is furnished to do | |
19 | so, subject to the following conditions: | |
20 | ||
21 | The above copyright notice and this permission notice shall be included in all | |
22 | copies or substantial portions of the Software. | |
23 | ||
24 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
25 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
26 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE | |
27 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
28 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
29 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE | |
30 | SOFTWARE. | |
31 | ||
32 | * | |
25468daa | 33 | * This file is a home for static inline functions that cannot go in other |
e15e54ff | 34 | * header files, because they depend on proto.h (included after most other |
25468daa FC |
35 | * headers) or struct definitions. |
36 | * | |
37 | * Each section names the header file that the functions "belong" to. | |
38 | */ | |
27669aa4 | 39 | |
be3a7a5d KW |
40 | /* ------------------------------- av.h ------------------------------- */ |
41 | ||
87306e06 | 42 | /* |
3f620621 | 43 | =for apidoc_section $AV |
87306e06 KW |
44 | =for apidoc av_count |
45 | Returns the number of elements in the array C<av>. This is the true length of | |
46 | the array, including any undefined elements. It is always the same as | |
47 | S<C<av_top_index(av) + 1>>. | |
48 | ||
49 | =cut | |
50 | */ | |
51 | PERL_STATIC_INLINE Size_t | |
52 | Perl_av_count(pTHX_ AV *av) | |
be3a7a5d | 53 | { |
87306e06 | 54 | PERL_ARGS_ASSERT_AV_COUNT; |
be3a7a5d KW |
55 | assert(SvTYPE(av) == SVt_PVAV); |
56 | ||
87306e06 | 57 | return AvFILL(av) + 1; |
be3a7a5d KW |
58 | } |
59 | ||
84c75204 RL |
60 | /* ------------------------------- av.c ------------------------------- */ |
61 | ||
62 | /* | |
63 | =for apidoc av_store_simple | |
64 | ||
65 | This is a cut-down version of av_store that assumes that the array is | |
66 | very straightforward - no magic, not readonly, and AvREAL - and that | |
67 | C<key> is not negative. This function MUST NOT be used in situations | |
68 | where any of those assumptions may not hold. | |
69 | ||
70 | Stores an SV in an array. The array index is specified as C<key>. It | |
71 | can be dereferenced to get the C<SV*> that was stored there (= C<val>)). | |
72 | ||
73 | Note that the caller is responsible for suitably incrementing the reference | |
74 | count of C<val> before the call. | |
75 | ||
76 | Approximate Perl equivalent: C<splice(@myarray, $key, 1, $val)>. | |
77 | ||
78 | =cut | |
79 | */ | |
80 | ||
81 | PERL_STATIC_INLINE SV** | |
82 | Perl_av_store_simple(pTHX_ AV *av, SSize_t key, SV *val) | |
83 | { | |
56077d95 LT |
84 | SV** ary; |
85 | ||
84c75204 RL |
86 | PERL_ARGS_ASSERT_AV_STORE_SIMPLE; |
87 | assert(SvTYPE(av) == SVt_PVAV); | |
88 | assert(!SvMAGICAL(av)); | |
89 | assert(!SvREADONLY(av)); | |
90 | assert(AvREAL(av)); | |
91 | assert(key > -1); | |
92 | ||
56077d95 | 93 | ary = AvARRAY(av); |
84c75204 RL |
94 | |
95 | if (AvFILLp(av) < key) { | |
96 | if (key > AvMAX(av)) { | |
97 | av_extend(av,key); | |
98 | ary = AvARRAY(av); | |
99 | } | |
100 | AvFILLp(av) = key; | |
101 | } else | |
102 | SvREFCNT_dec(ary[key]); | |
103 | ||
104 | ary[key] = val; | |
105 | return &ary[key]; | |
106 | } | |
107 | ||
108 | /* | |
109 | =for apidoc av_fetch_simple | |
110 | ||
111 | This is a cut-down version of av_fetch that assumes that the array is | |
112 | very straightforward - no magic, not readonly, and AvREAL - and that | |
113 | C<key> is not negative. This function MUST NOT be used in situations | |
114 | where any of those assumptions may not hold. | |
115 | ||
116 | Returns the SV at the specified index in the array. The C<key> is the | |
117 | index. If lval is true, you are guaranteed to get a real SV back (in case | |
118 | it wasn't real before), which you can then modify. Check that the return | |
119 | value is non-null before dereferencing it to a C<SV*>. | |
120 | ||
121 | The rough perl equivalent is C<$myarray[$key]>. | |
122 | ||
123 | =cut | |
124 | */ | |
125 | ||
126 | PERL_STATIC_INLINE SV** | |
127 | Perl_av_fetch_simple(pTHX_ AV *av, SSize_t key, I32 lval) | |
128 | { | |
129 | PERL_ARGS_ASSERT_AV_FETCH_SIMPLE; | |
130 | assert(SvTYPE(av) == SVt_PVAV); | |
131 | assert(!SvMAGICAL(av)); | |
132 | assert(!SvREADONLY(av)); | |
133 | assert(AvREAL(av)); | |
134 | assert(key > -1); | |
135 | ||
136 | if ( (key > AvFILLp(av)) || !AvARRAY(av)[key]) { | |
8fcb2425 | 137 | return lval ? av_store_simple(av,key,newSV_type(SVt_NULL)) : NULL; |
84c75204 RL |
138 | } else { |
139 | return &AvARRAY(av)[key]; | |
140 | } | |
141 | } | |
142 | ||
1afe1db1 FC |
143 | /* ------------------------------- cv.h ------------------------------- */ |
144 | ||
fa3e44c0 | 145 | /* |
3f620621 | 146 | =for apidoc_section $CV |
fa3e44c0 KW |
147 | =for apidoc CvGV |
148 | Returns the GV associated with the CV C<sv>, reifying it if necessary. | |
149 | ||
150 | =cut | |
151 | */ | |
ae77754a | 152 | PERL_STATIC_INLINE GV * |
c9182d9c | 153 | Perl_CvGV(pTHX_ CV *sv) |
ae77754a | 154 | { |
74804ad1 KW |
155 | PERL_ARGS_ASSERT_CVGV; |
156 | ||
ae77754a | 157 | return CvNAMED(sv) |
1604cfb0 MS |
158 | ? Perl_cvgv_from_hek(aTHX_ sv) |
159 | : ((XPVCV*)MUTABLE_PTR(SvANY(sv)))->xcv_gv_u.xcv_gv; | |
ae77754a FC |
160 | } |
161 | ||
1afe1db1 | 162 | PERL_STATIC_INLINE I32 * |
74804ad1 | 163 | Perl_CvDEPTH(const CV * const sv) |
1afe1db1 | 164 | { |
74804ad1 | 165 | PERL_ARGS_ASSERT_CVDEPTH; |
1afe1db1 | 166 | assert(SvTYPE(sv) == SVt_PVCV || SvTYPE(sv) == SVt_PVFM); |
74804ad1 | 167 | |
8de47657 | 168 | return &((XPVCV*)SvANY(sv))->xcv_depth; |
1afe1db1 FC |
169 | } |
170 | ||
d16269d8 PM |
171 | /* |
172 | CvPROTO returns the prototype as stored, which is not necessarily what | |
173 | the interpreter should be using. Specifically, the interpreter assumes | |
174 | that spaces have been stripped, which has been the case if the prototype | |
175 | was added by toke.c, but is generally not the case if it was added elsewhere. | |
176 | Since we can't enforce the spacelessness at assignment time, this routine | |
177 | provides a temporary copy at parse time with spaces removed. | |
178 | I<orig> is the start of the original buffer, I<len> is the length of the | |
179 | prototype and will be updated when this returns. | |
180 | */ | |
181 | ||
5b67adb8 | 182 | #ifdef PERL_CORE |
d16269d8 PM |
183 | PERL_STATIC_INLINE char * |
184 | S_strip_spaces(pTHX_ const char * orig, STRLEN * const len) | |
185 | { | |
186 | SV * tmpsv; | |
187 | char * tmps; | |
188 | tmpsv = newSVpvn_flags(orig, *len, SVs_TEMP); | |
189 | tmps = SvPVX(tmpsv); | |
190 | while ((*len)--) { | |
1604cfb0 MS |
191 | if (!isSPACE(*orig)) |
192 | *tmps++ = *orig; | |
193 | orig++; | |
d16269d8 PM |
194 | } |
195 | *tmps = '\0'; | |
196 | *len = tmps - SvPVX(tmpsv); | |
1604cfb0 | 197 | return SvPVX(tmpsv); |
d16269d8 | 198 | } |
5b67adb8 | 199 | #endif |
d16269d8 | 200 | |
25fdce4a FC |
201 | /* ------------------------------- mg.h ------------------------------- */ |
202 | ||
203 | #if defined(PERL_CORE) || defined(PERL_EXT) | |
204 | /* assumes get-magic and stringification have already occurred */ | |
205 | PERL_STATIC_INLINE STRLEN | |
206 | S_MgBYTEPOS(pTHX_ MAGIC *mg, SV *sv, const char *s, STRLEN len) | |
207 | { | |
208 | assert(mg->mg_type == PERL_MAGIC_regex_global); | |
209 | assert(mg->mg_len != -1); | |
210 | if (mg->mg_flags & MGf_BYTES || !DO_UTF8(sv)) | |
1604cfb0 | 211 | return (STRLEN)mg->mg_len; |
25fdce4a | 212 | else { |
1604cfb0 MS |
213 | const STRLEN pos = (STRLEN)mg->mg_len; |
214 | /* Without this check, we may read past the end of the buffer: */ | |
215 | if (pos > sv_or_pv_len_utf8(sv, s, len)) return len+1; | |
216 | return sv_or_pv_pos_u2b(sv, s, pos, NULL); | |
25fdce4a FC |
217 | } |
218 | } | |
219 | #endif | |
220 | ||
03414f05 FC |
221 | /* ------------------------------- pad.h ------------------------------ */ |
222 | ||
223 | #if defined(PERL_IN_PAD_C) || defined(PERL_IN_OP_C) | |
224 | PERL_STATIC_INLINE bool | |
b9d5702c | 225 | S_PadnameIN_SCOPE(const PADNAME * const pn, const U32 seq) |
03414f05 | 226 | { |
b9d5702c KW |
227 | PERL_ARGS_ASSERT_PADNAMEIN_SCOPE; |
228 | ||
03414f05 FC |
229 | /* is seq within the range _LOW to _HIGH ? |
230 | * This is complicated by the fact that PL_cop_seqmax | |
231 | * may have wrapped around at some point */ | |
232 | if (COP_SEQ_RANGE_LOW(pn) == PERL_PADSEQ_INTRO) | |
1604cfb0 | 233 | return FALSE; /* not yet introduced */ |
03414f05 FC |
234 | |
235 | if (COP_SEQ_RANGE_HIGH(pn) == PERL_PADSEQ_INTRO) { | |
236 | /* in compiling scope */ | |
1604cfb0 MS |
237 | if ( |
238 | (seq > COP_SEQ_RANGE_LOW(pn)) | |
239 | ? (seq - COP_SEQ_RANGE_LOW(pn) < (U32_MAX >> 1)) | |
240 | : (COP_SEQ_RANGE_LOW(pn) - seq > (U32_MAX >> 1)) | |
241 | ) | |
242 | return TRUE; | |
03414f05 FC |
243 | } |
244 | else if ( | |
1604cfb0 MS |
245 | (COP_SEQ_RANGE_LOW(pn) > COP_SEQ_RANGE_HIGH(pn)) |
246 | ? | |
247 | ( seq > COP_SEQ_RANGE_LOW(pn) | |
248 | || seq <= COP_SEQ_RANGE_HIGH(pn)) | |
03414f05 | 249 | |
1604cfb0 MS |
250 | : ( seq > COP_SEQ_RANGE_LOW(pn) |
251 | && seq <= COP_SEQ_RANGE_HIGH(pn)) | |
03414f05 | 252 | ) |
1604cfb0 | 253 | return TRUE; |
03414f05 FC |
254 | return FALSE; |
255 | } | |
256 | #endif | |
257 | ||
33a4312b FC |
258 | /* ------------------------------- pp.h ------------------------------- */ |
259 | ||
260 | PERL_STATIC_INLINE I32 | |
c9182d9c | 261 | Perl_TOPMARK(pTHX) |
33a4312b FC |
262 | { |
263 | DEBUG_s(DEBUG_v(PerlIO_printf(Perl_debug_log, | |
1604cfb0 MS |
264 | "MARK top %p %" IVdf "\n", |
265 | PL_markstack_ptr, | |
266 | (IV)*PL_markstack_ptr))); | |
33a4312b FC |
267 | return *PL_markstack_ptr; |
268 | } | |
269 | ||
270 | PERL_STATIC_INLINE I32 | |
c9182d9c | 271 | Perl_POPMARK(pTHX) |
33a4312b FC |
272 | { |
273 | DEBUG_s(DEBUG_v(PerlIO_printf(Perl_debug_log, | |
1604cfb0 MS |
274 | "MARK pop %p %" IVdf "\n", |
275 | (PL_markstack_ptr-1), | |
276 | (IV)*(PL_markstack_ptr-1)))); | |
33a4312b FC |
277 | assert((PL_markstack_ptr > PL_markstack) || !"MARK underflow"); |
278 | return *PL_markstack_ptr--; | |
279 | } | |
280 | ||
8d919b0a FC |
281 | /* ----------------------------- regexp.h ----------------------------- */ |
282 | ||
9d0d3060 NC |
283 | /* PVLVs need to act as a superset of all scalar types - they are basically |
284 | * PVMGs with a few extra fields. | |
285 | * REGEXPs are first class scalars, but have many fields that can't be copied | |
286 | * into a PVLV body. | |
287 | * | |
288 | * Hence we take a different approach - instead of a copy, PVLVs store a pointer | |
289 | * back to the original body. To avoid increasing the size of PVLVs just for the | |
290 | * rare case of REGEXP assignment, this pointer is stored in the memory usually | |
291 | * used for SvLEN(). Hence the check for SVt_PVLV below, and the ? : ternary to | |
292 | * read the pointer from the two possible locations. The macro SvLEN() wraps the | |
293 | * access to the union's member xpvlenu_len, but there is no equivalent macro | |
294 | * for wrapping the union's member xpvlenu_rx, hence the direct reference here. | |
295 | * | |
296 | * See commit df6b4bd56551f2d3 for more details. */ | |
297 | ||
8d919b0a | 298 | PERL_STATIC_INLINE struct regexp * |
c9182d9c | 299 | Perl_ReANY(const REGEXP * const re) |
8d919b0a | 300 | { |
df6b4bd5 | 301 | XPV* const p = (XPV*)SvANY(re); |
bdef45de KW |
302 | |
303 | PERL_ARGS_ASSERT_REANY; | |
8d919b0a | 304 | assert(isREGEXP(re)); |
bdef45de | 305 | |
df6b4bd5 DM |
306 | return SvTYPE(re) == SVt_PVLV ? p->xpv_len_u.xpvlenu_rx |
307 | : (struct regexp *)p; | |
8d919b0a FC |
308 | } |
309 | ||
27669aa4 FC |
310 | /* ------------------------------- sv.h ------------------------------- */ |
311 | ||
a887b094 | 312 | PERL_STATIC_INLINE bool |
4eff5eb8 KW |
313 | Perl_SvTRUE(pTHX_ SV *sv) |
314 | { | |
315 | PERL_ARGS_ASSERT_SVTRUE; | |
316 | ||
6aa30f5e | 317 | if (UNLIKELY(sv == NULL)) |
9d0469db | 318 | return FALSE; |
7b9b9e32 PE |
319 | SvGETMAGIC(sv); |
320 | return SvTRUE_nomg_NN(sv); | |
a887b094 PE |
321 | } |
322 | ||
4eff5eb8 KW |
323 | PERL_STATIC_INLINE bool |
324 | Perl_SvTRUE_nomg(pTHX_ SV *sv) | |
325 | { | |
326 | PERL_ARGS_ASSERT_SVTRUE_NOMG; | |
327 | ||
328 | if (UNLIKELY(sv == NULL)) | |
329 | return FALSE; | |
330 | return SvTRUE_nomg_NN(sv); | |
331 | } | |
332 | ||
333 | PERL_STATIC_INLINE bool | |
334 | Perl_SvTRUE_NN(pTHX_ SV *sv) | |
335 | { | |
336 | PERL_ARGS_ASSERT_SVTRUE_NN; | |
337 | ||
338 | SvGETMAGIC(sv); | |
339 | return SvTRUE_nomg_NN(sv); | |
340 | } | |
341 | ||
342 | PERL_STATIC_INLINE bool | |
343 | Perl_SvTRUE_common(pTHX_ SV * sv, const bool sv_2bool_is_fallback) | |
344 | { | |
345 | PERL_ARGS_ASSERT_SVTRUE_COMMON; | |
346 | ||
347 | if (UNLIKELY(SvIMMORTAL_INTERP(sv))) | |
348 | return SvIMMORTAL_TRUE(sv); | |
349 | ||
350 | if (! SvOK(sv)) | |
351 | return FALSE; | |
352 | ||
353 | if (SvPOK(sv)) | |
354 | return SvPVXtrue(sv); | |
355 | ||
356 | if (SvIOK(sv)) | |
357 | return SvIVX(sv) != 0; /* casts to bool */ | |
358 | ||
359 | if (SvROK(sv) && !(SvOBJECT(SvRV(sv)) && HvAMAGIC(SvSTASH(SvRV(sv))))) | |
360 | return TRUE; | |
361 | ||
362 | if (sv_2bool_is_fallback) | |
363 | return sv_2bool_nomg(sv); | |
364 | ||
365 | return isGV_with_GP(sv); | |
366 | } | |
367 | ||
368 | ||
27669aa4 | 369 | PERL_STATIC_INLINE SV * |
c9182d9c | 370 | Perl_SvREFCNT_inc(SV *sv) |
27669aa4 | 371 | { |
2439e033 | 372 | if (LIKELY(sv != NULL)) |
1604cfb0 | 373 | SvREFCNT(sv)++; |
27669aa4 FC |
374 | return sv; |
375 | } | |
376 | PERL_STATIC_INLINE SV * | |
c9182d9c | 377 | Perl_SvREFCNT_inc_NN(SV *sv) |
27669aa4 | 378 | { |
3f2f854a KW |
379 | PERL_ARGS_ASSERT_SVREFCNT_INC_NN; |
380 | ||
27669aa4 FC |
381 | SvREFCNT(sv)++; |
382 | return sv; | |
383 | } | |
384 | PERL_STATIC_INLINE void | |
c9182d9c | 385 | Perl_SvREFCNT_inc_void(SV *sv) |
27669aa4 | 386 | { |
2439e033 | 387 | if (LIKELY(sv != NULL)) |
1604cfb0 | 388 | SvREFCNT(sv)++; |
27669aa4 | 389 | } |
75e16a44 | 390 | PERL_STATIC_INLINE void |
c9182d9c | 391 | Perl_SvREFCNT_dec(pTHX_ SV *sv) |
75e16a44 | 392 | { |
2439e033 | 393 | if (LIKELY(sv != NULL)) { |
1604cfb0 MS |
394 | U32 rc = SvREFCNT(sv); |
395 | if (LIKELY(rc > 1)) | |
396 | SvREFCNT(sv) = rc - 1; | |
397 | else | |
398 | Perl_sv_free2(aTHX_ sv, rc); | |
75e16a44 FC |
399 | } |
400 | } | |
541377b1 FC |
401 | |
402 | PERL_STATIC_INLINE void | |
c9182d9c | 403 | Perl_SvREFCNT_dec_NN(pTHX_ SV *sv) |
4a9a56a7 DM |
404 | { |
405 | U32 rc = SvREFCNT(sv); | |
3f2f854a KW |
406 | |
407 | PERL_ARGS_ASSERT_SVREFCNT_DEC_NN; | |
408 | ||
79e2a32a | 409 | if (LIKELY(rc > 1)) |
1604cfb0 | 410 | SvREFCNT(sv) = rc - 1; |
4a9a56a7 | 411 | else |
1604cfb0 | 412 | Perl_sv_free2(aTHX_ sv, rc); |
4a9a56a7 DM |
413 | } |
414 | ||
415 | PERL_STATIC_INLINE void | |
1bd041dc | 416 | Perl_SvAMAGIC_on(SV *sv) |
541377b1 | 417 | { |
1bd041dc | 418 | PERL_ARGS_ASSERT_SVAMAGIC_ON; |
541377b1 | 419 | assert(SvROK(sv)); |
1bd041dc | 420 | |
541377b1 FC |
421 | if (SvOBJECT(SvRV(sv))) HvAMAGIC_on(SvSTASH(SvRV(sv))); |
422 | } | |
423 | PERL_STATIC_INLINE void | |
1bd041dc | 424 | Perl_SvAMAGIC_off(SV *sv) |
541377b1 | 425 | { |
1bd041dc KW |
426 | PERL_ARGS_ASSERT_SVAMAGIC_OFF; |
427 | ||
541377b1 | 428 | if (SvROK(sv) && SvOBJECT(SvRV(sv))) |
1604cfb0 | 429 | HvAMAGIC_off(SvSTASH(SvRV(sv))); |
541377b1 FC |
430 | } |
431 | ||
432 | PERL_STATIC_INLINE U32 | |
c9182d9c | 433 | Perl_SvPADSTALE_on(SV *sv) |
541377b1 | 434 | { |
c0683843 | 435 | assert(!(SvFLAGS(sv) & SVs_PADTMP)); |
541377b1 FC |
436 | return SvFLAGS(sv) |= SVs_PADSTALE; |
437 | } | |
438 | PERL_STATIC_INLINE U32 | |
c9182d9c | 439 | Perl_SvPADSTALE_off(SV *sv) |
541377b1 | 440 | { |
c0683843 | 441 | assert(!(SvFLAGS(sv) & SVs_PADTMP)); |
541377b1 FC |
442 | return SvFLAGS(sv) &= ~SVs_PADSTALE; |
443 | } | |
25fdce4a | 444 | #if defined(PERL_CORE) || defined (PERL_EXT) |
4ddea69a | 445 | PERL_STATIC_INLINE STRLEN |
6964422a | 446 | S_sv_or_pv_pos_u2b(pTHX_ SV *sv, const char *pv, STRLEN pos, STRLEN *lenp) |
4ddea69a | 447 | { |
25fdce4a | 448 | PERL_ARGS_ASSERT_SV_OR_PV_POS_U2B; |
4ddea69a | 449 | if (SvGAMAGIC(sv)) { |
1604cfb0 MS |
450 | U8 *hopped = utf8_hop((U8 *)pv, pos); |
451 | if (lenp) *lenp = (STRLEN)(utf8_hop(hopped, *lenp) - hopped); | |
452 | return (STRLEN)(hopped - (U8 *)pv); | |
4ddea69a FC |
453 | } |
454 | return sv_pos_u2b_flags(sv,pos,lenp,SV_CONST_RETURN); | |
455 | } | |
456 | #endif | |
f019c49e | 457 | |
a8a2ceaa KW |
458 | /* ------------------------------- utf8.h ------------------------------- */ |
459 | ||
2fe720e2 | 460 | /* |
3f620621 | 461 | =for apidoc_section $unicode |
2fe720e2 KW |
462 | */ |
463 | ||
55d09dc8 | 464 | PERL_STATIC_INLINE void |
c9182d9c | 465 | Perl_append_utf8_from_native_byte(const U8 byte, U8** dest) |
55d09dc8 KW |
466 | { |
467 | /* Takes an input 'byte' (Latin1 or EBCDIC) and appends it to the UTF-8 | |
468 | * encoded string at '*dest', updating '*dest' to include it */ | |
469 | ||
55d09dc8 KW |
470 | PERL_ARGS_ASSERT_APPEND_UTF8_FROM_NATIVE_BYTE; |
471 | ||
6f2d5cbc | 472 | if (NATIVE_BYTE_IS_INVARIANT(byte)) |
a09ec51a | 473 | *((*dest)++) = byte; |
55d09dc8 | 474 | else { |
a09ec51a KW |
475 | *((*dest)++) = UTF8_EIGHT_BIT_HI(byte); |
476 | *((*dest)++) = UTF8_EIGHT_BIT_LO(byte); | |
55d09dc8 KW |
477 | } |
478 | } | |
479 | ||
e123187a | 480 | /* |
2fe720e2 | 481 | =for apidoc valid_utf8_to_uvchr |
09232555 KW |
482 | Like C<L<perlapi/utf8_to_uvchr_buf>>, but should only be called when it is |
483 | known that the next character in the input UTF-8 string C<s> is well-formed | |
484 | (I<e.g.>, it passes C<L<perlapi/isUTF8_CHAR>>. Surrogates, non-character code | |
485 | points, and non-Unicode code points are allowed. | |
2fe720e2 KW |
486 | |
487 | =cut | |
488 | ||
489 | */ | |
490 | ||
491 | PERL_STATIC_INLINE UV | |
492 | Perl_valid_utf8_to_uvchr(const U8 *s, STRLEN *retlen) | |
493 | { | |
c41b2540 | 494 | const UV expectlen = UTF8SKIP(s); |
2fe720e2 KW |
495 | const U8* send = s + expectlen; |
496 | UV uv = *s; | |
497 | ||
498 | PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR; | |
499 | ||
500 | if (retlen) { | |
501 | *retlen = expectlen; | |
502 | } | |
503 | ||
504 | /* An invariant is trivially returned */ | |
505 | if (expectlen == 1) { | |
1604cfb0 | 506 | return uv; |
2fe720e2 KW |
507 | } |
508 | ||
509 | /* Remove the leading bits that indicate the number of bytes, leaving just | |
510 | * the bits that are part of the value */ | |
511 | uv = NATIVE_UTF8_TO_I8(uv) & UTF_START_MASK(expectlen); | |
512 | ||
513 | /* Now, loop through the remaining bytes, accumulating each into the | |
514 | * working total as we go. (I khw tried unrolling the loop for up to 4 | |
515 | * bytes, but there was no performance improvement) */ | |
516 | for (++s; s < send; s++) { | |
517 | uv = UTF8_ACCUMULATE(uv, *s); | |
518 | } | |
519 | ||
520 | return UNI_TO_NATIVE(uv); | |
521 | ||
522 | } | |
523 | ||
1e599354 KW |
524 | /* |
525 | =for apidoc is_utf8_invariant_string | |
526 | ||
82c5d941 | 527 | Returns TRUE if the first C<len> bytes of the string C<s> are the same |
1e599354 | 528 | regardless of the UTF-8 encoding of the string (or UTF-EBCDIC encoding on |
82c5d941 KW |
529 | EBCDIC machines); otherwise it returns FALSE. That is, it returns TRUE if they |
530 | are UTF-8 invariant. On ASCII-ish machines, all the ASCII characters and only | |
531 | the ASCII characters fit this definition. On EBCDIC machines, the ASCII-range | |
532 | characters are invariant, but so also are the C1 controls. | |
1e599354 KW |
533 | |
534 | If C<len> is 0, it will be calculated using C<strlen(s)>, (which means if you | |
535 | use this option, that C<s> can't have embedded C<NUL> characters and has to | |
536 | have a terminating C<NUL> byte). | |
537 | ||
9f2abfde KW |
538 | See also |
539 | C<L</is_utf8_string>>, | |
540 | C<L</is_utf8_string_flags>>, | |
541 | C<L</is_utf8_string_loc>>, | |
542 | C<L</is_utf8_string_loc_flags>>, | |
543 | C<L</is_utf8_string_loclen>>, | |
544 | C<L</is_utf8_string_loclen_flags>>, | |
8bc127bf KW |
545 | C<L</is_utf8_fixed_width_buf_flags>>, |
546 | C<L</is_utf8_fixed_width_buf_loc_flags>>, | |
547 | C<L</is_utf8_fixed_width_buf_loclen_flags>>, | |
9f2abfde KW |
548 | C<L</is_strict_utf8_string>>, |
549 | C<L</is_strict_utf8_string_loc>>, | |
550 | C<L</is_strict_utf8_string_loclen>>, | |
551 | C<L</is_c9strict_utf8_string>>, | |
552 | C<L</is_c9strict_utf8_string_loc>>, | |
553 | and | |
554 | C<L</is_c9strict_utf8_string_loclen>>. | |
1e599354 KW |
555 | |
556 | =cut | |
0cbf5865 KW |
557 | |
558 | */ | |
559 | ||
560 | #define is_utf8_invariant_string(s, len) \ | |
561 | is_utf8_invariant_string_loc(s, len, NULL) | |
562 | ||
563 | /* | |
564 | =for apidoc is_utf8_invariant_string_loc | |
565 | ||
566 | Like C<L</is_utf8_invariant_string>> but upon failure, stores the location of | |
567 | the first UTF-8 variant character in the C<ep> pointer; if all characters are | |
568 | UTF-8 invariant, this function does not change the contents of C<*ep>. | |
569 | ||
570 | =cut | |
571 | ||
1e599354 KW |
572 | */ |
573 | ||
574 | PERL_STATIC_INLINE bool | |
c9182d9c | 575 | Perl_is_utf8_invariant_string_loc(const U8* const s, STRLEN len, const U8 ** ep) |
1e599354 | 576 | { |
e17544a6 | 577 | const U8* send; |
1e599354 KW |
578 | const U8* x = s; |
579 | ||
0cbf5865 KW |
580 | PERL_ARGS_ASSERT_IS_UTF8_INVARIANT_STRING_LOC; |
581 | ||
e17544a6 KW |
582 | if (len == 0) { |
583 | len = strlen((const char *)s); | |
584 | } | |
585 | ||
586 | send = s + len; | |
587 | ||
4ab2fd9b | 588 | /* This looks like 0x010101... */ |
2c5c8af5 | 589 | # define PERL_COUNT_MULTIPLIER (~ (UINTMAX_C(0)) / 0xFF) |
4ab2fd9b KW |
590 | |
591 | /* This looks like 0x808080... */ | |
2c5c8af5 | 592 | # define PERL_VARIANTS_WORD_MASK (PERL_COUNT_MULTIPLIER * 0x80) |
e099ea69 | 593 | # define PERL_WORDSIZE sizeof(PERL_UINTMAX_T) |
2c5c8af5 | 594 | # define PERL_WORD_BOUNDARY_MASK (PERL_WORDSIZE - 1) |
e17544a6 | 595 | |
099e59a4 KW |
596 | /* Evaluates to 0 if 'x' is at a word boundary; otherwise evaluates to 1, by |
597 | * or'ing together the lowest bits of 'x'. Hopefully the final term gets | |
598 | * optimized out completely on a 32-bit system, and its mask gets optimized out | |
599 | * on a 64-bit system */ | |
2c5c8af5 | 600 | # define PERL_IS_SUBWORD_ADDR(x) (1 & ( PTR2nat(x) \ |
5eabe374 KW |
601 | | ( PTR2nat(x) >> 1) \ |
602 | | ( ( (PTR2nat(x) \ | |
603 | & PERL_WORD_BOUNDARY_MASK) >> 2)))) | |
099e59a4 | 604 | |
3f515a2e KW |
605 | #ifndef EBCDIC |
606 | ||
099e59a4 KW |
607 | /* Do the word-at-a-time iff there is at least one usable full word. That |
608 | * means that after advancing to a word boundary, there still is at least a | |
609 | * full word left. The number of bytes needed to advance is 'wordsize - | |
610 | * offset' unless offset is 0. */ | |
611 | if ((STRLEN) (send - x) >= PERL_WORDSIZE | |
612 | ||
613 | /* This term is wordsize if subword; 0 if not */ | |
614 | + PERL_WORDSIZE * PERL_IS_SUBWORD_ADDR(x) | |
615 | ||
616 | /* 'offset' */ | |
617 | - (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK)) | |
618 | { | |
b40579ff | 619 | |
46bb68f6 KW |
620 | /* Process per-byte until reach word boundary. XXX This loop could be |
621 | * eliminated if we knew that this platform had fast unaligned reads */ | |
b40579ff | 622 | while (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK) { |
46bb68f6 KW |
623 | if (! UTF8_IS_INVARIANT(*x)) { |
624 | if (ep) { | |
625 | *ep = x; | |
626 | } | |
e17544a6 | 627 | |
46bb68f6 KW |
628 | return FALSE; |
629 | } | |
630 | x++; | |
e17544a6 | 631 | } |
e17544a6 | 632 | |
099e59a4 KW |
633 | /* Here, we know we have at least one full word to process. Process |
634 | * per-word as long as we have at least a full word left */ | |
635 | do { | |
4ab2fd9b | 636 | if ((* (PERL_UINTMAX_T *) x) & PERL_VARIANTS_WORD_MASK) { |
e17544a6 | 637 | |
46bb68f6 KW |
638 | /* Found a variant. Just return if caller doesn't want its |
639 | * exact position */ | |
640 | if (! ep) { | |
641 | return FALSE; | |
642 | } | |
e17544a6 | 643 | |
2c5c8af5 KW |
644 | # if BYTEORDER == 0x1234 || BYTEORDER == 0x12345678 \ |
645 | || BYTEORDER == 0x4321 || BYTEORDER == 0x87654321 | |
1d2af574 | 646 | |
73f0a2eb | 647 | *ep = x + variant_byte_number(* (PERL_UINTMAX_T *) x); |
1d2af574 KW |
648 | assert(*ep >= s && *ep < send); |
649 | ||
650 | return FALSE; | |
651 | ||
2c5c8af5 | 652 | # else /* If weird byte order, drop into next loop to do byte-at-a-time |
1d2af574 KW |
653 | checks. */ |
654 | ||
46bb68f6 | 655 | break; |
2c5c8af5 | 656 | # endif |
46bb68f6 | 657 | } |
1d2af574 | 658 | |
46bb68f6 | 659 | x += PERL_WORDSIZE; |
1d2af574 | 660 | |
099e59a4 | 661 | } while (x + PERL_WORDSIZE <= send); |
b40579ff | 662 | } |
e17544a6 | 663 | |
0b08cab0 | 664 | #endif /* End of ! EBCDIC */ |
e17544a6 KW |
665 | |
666 | /* Process per-byte */ | |
667 | while (x < send) { | |
1604cfb0 | 668 | if (! UTF8_IS_INVARIANT(*x)) { |
e17544a6 KW |
669 | if (ep) { |
670 | *ep = x; | |
671 | } | |
0cbf5865 | 672 | |
e17544a6 | 673 | return FALSE; |
0cbf5865 | 674 | } |
1e599354 | 675 | |
e17544a6 | 676 | x++; |
1e599354 KW |
677 | } |
678 | ||
679 | return TRUE; | |
680 | } | |
bf874180 | 681 | |
fc1bb663 KW |
682 | /* See if the platform has builtins for finding the most/least significant bit, |
683 | * and which one is right for using on 32 and 64 bit operands */ | |
684 | #if (__has_builtin(__builtin_clz) || PERL_GCC_VERSION_GE(3,4,0)) | |
685 | # if U32SIZE == INTSIZE | |
686 | # define PERL_CLZ_32 __builtin_clz | |
687 | # endif | |
688 | # if defined(U64TYPE) && U64SIZE == INTSIZE | |
689 | # define PERL_CLZ_64 __builtin_clz | |
690 | # endif | |
691 | #endif | |
692 | #if (__has_builtin(__builtin_ctz) || PERL_GCC_VERSION_GE(3,4,0)) | |
693 | # if U32SIZE == INTSIZE | |
694 | # define PERL_CTZ_32 __builtin_ctz | |
695 | # endif | |
696 | # if defined(U64TYPE) && U64SIZE == INTSIZE | |
697 | # define PERL_CTZ_64 __builtin_ctz | |
698 | # endif | |
699 | #endif | |
700 | ||
701 | #if (__has_builtin(__builtin_clzl) || PERL_GCC_VERSION_GE(3,4,0)) | |
702 | # if U32SIZE == LONGSIZE && ! defined(PERL_CLZ_32) | |
703 | # define PERL_CLZ_32 __builtin_clzl | |
704 | # endif | |
705 | # if defined(U64TYPE) && U64SIZE == LONGSIZE && ! defined(PERL_CLZ_64) | |
706 | # define PERL_CLZ_64 __builtin_clzl | |
707 | # endif | |
708 | #endif | |
709 | #if (__has_builtin(__builtin_ctzl) || PERL_GCC_VERSION_GE(3,4,0)) | |
710 | # if U32SIZE == LONGSIZE && ! defined(PERL_CTZ_32) | |
711 | # define PERL_CTZ_32 __builtin_ctzl | |
712 | # endif | |
713 | # if defined(U64TYPE) && U64SIZE == LONGSIZE && ! defined(PERL_CTZ_64) | |
714 | # define PERL_CTZ_64 __builtin_ctzl | |
715 | # endif | |
716 | #endif | |
717 | ||
718 | #if (__has_builtin(__builtin_clzll) || PERL_GCC_VERSION_GE(3,4,0)) | |
719 | # if U32SIZE == LONGLONGSIZE && ! defined(PERL_CLZ_32) | |
720 | # define PERL_CLZ_32 __builtin_clzll | |
721 | # endif | |
722 | # if defined(U64TYPE) && U64SIZE == LONGLONGSIZE && ! defined(PERL_CLZ_64) | |
723 | # define PERL_CLZ_64 __builtin_clzll | |
724 | # endif | |
725 | #endif | |
726 | #if (__has_builtin(__builtin_ctzll) || PERL_GCC_VERSION_GE(3,4,0)) | |
727 | # if U32SIZE == LONGLONGSIZE && ! defined(PERL_CTZ_32) | |
728 | # define PERL_CTZ_32 __builtin_ctzll | |
729 | # endif | |
730 | # if defined(U64TYPE) && U64SIZE == LONGLONGSIZE && ! defined(PERL_CTZ_64) | |
731 | # define PERL_CTZ_64 __builtin_ctzll | |
732 | # endif | |
733 | #endif | |
734 | ||
e1f6bdff | 735 | #if defined(_MSC_VER) |
e88dde50 KW |
736 | # include <intrin.h> |
737 | # pragma intrinsic(_BitScanForward) | |
738 | # pragma intrinsic(_BitScanReverse) | |
739 | # ifdef _WIN64 | |
740 | # pragma intrinsic(_BitScanForward64) | |
741 | # pragma intrinsic(_BitScanReverse64) | |
742 | # endif | |
743 | #endif | |
744 | ||
250e5324 KW |
745 | /* The reason there are not checks to see if ffs() and ffsl() are available for |
746 | * determining the lsb, is because these don't improve on the deBruijn method | |
747 | * fallback, which is just a branchless integer multiply, array element | |
748 | * retrieval, and shift. The others, even if the function call overhead is | |
749 | * optimized out, have to cope with the possibility of the input being all | |
750 | * zeroes, and almost certainly will have conditionals for this eventuality. | |
751 | * khw, at the time of this commit, looked at the source for both gcc and clang | |
752 | * to verify this. (gcc used a method inferior to deBruijn.) */ | |
753 | ||
330cd0ce | 754 | /* Below are functions to find the first, last, or only set bit in a word. On |
19d2c525 KW |
755 | * platforms with 64-bit capability, there is a pair for each operation; the |
756 | * first taking a 64 bit operand, and the second a 32 bit one. The logic is | |
757 | * the same in each pair, so the second is stripped of most comments. */ | |
758 | ||
759 | #ifdef U64TYPE /* HAS_QUAD not usable outside the core */ | |
760 | ||
761 | PERL_STATIC_INLINE unsigned | |
762 | Perl_lsbit_pos64(U64 word) | |
763 | { | |
764 | /* Find the position (0..63) of the least significant set bit in the input | |
765 | * word */ | |
766 | ||
767 | ASSUME(word != 0); | |
768 | ||
fc1bb663 KW |
769 | /* If we can determine that the platform has a usable fast method to get |
770 | * this info, use that */ | |
771 | ||
772 | # if defined(PERL_CTZ_64) | |
2e0bc9ce | 773 | # define PERL_HAS_FAST_GET_LSB_POS64 |
fc1bb663 KW |
774 | |
775 | return (unsigned) PERL_CTZ_64(word); | |
776 | ||
a333292f | 777 | # elif U64SIZE == 8 && defined(_WIN64) && defined(_MSC_VER) |
2e0bc9ce | 778 | # define PERL_HAS_FAST_GET_LSB_POS64 |
e88dde50 KW |
779 | |
780 | { | |
781 | unsigned long index; | |
782 | _BitScanForward64(&index, word); | |
783 | return (unsigned)index; | |
784 | } | |
785 | ||
fc1bb663 KW |
786 | # else |
787 | ||
788 | /* Here, we didn't find a fast method for finding the lsb. Fall back to | |
789 | * making the lsb the only set bit in the word, and use our function that | |
790 | * works on words with a single bit set. | |
791 | * | |
792 | * Isolate the lsb; | |
19d2c525 KW |
793 | * https://stackoverflow.com/questions/757059/position-of-least-significant-bit-that-is-set |
794 | * | |
795 | * The word will look like this, with a rightmost set bit in position 's': | |
796 | * ('x's are don't cares, and 'y's are their complements) | |
797 | * s | |
798 | * x..x100..00 | |
799 | * y..y011..11 Complement | |
800 | * y..y100..00 Add 1 | |
801 | * 0..0100..00 And with the original | |
802 | * | |
803 | * (Yes, complementing and adding 1 is just taking the negative on 2's | |
804 | * complement machines, but not on 1's complement ones, and some compilers | |
805 | * complain about negating an unsigned.) | |
806 | */ | |
807 | return single_1bit_pos64(word & (~word + 1)); | |
fc1bb663 KW |
808 | |
809 | # endif | |
810 | ||
19d2c525 KW |
811 | } |
812 | ||
813 | # define lsbit_pos_uintmax_(word) lsbit_pos64(word) | |
814 | #else /* ! QUAD */ | |
815 | # define lsbit_pos_uintmax_(word) lsbit_pos32(word) | |
816 | #endif | |
817 | ||
818 | PERL_STATIC_INLINE unsigned /* Like above for 32 bit word */ | |
819 | Perl_lsbit_pos32(U32 word) | |
820 | { | |
821 | /* Find the position (0..31) of the least significant set bit in the input | |
822 | * word */ | |
823 | ||
824 | ASSUME(word != 0); | |
825 | ||
fc1bb663 | 826 | #if defined(PERL_CTZ_32) |
2e0bc9ce | 827 | # define PERL_HAS_FAST_GET_LSB_POS32 |
fc1bb663 KW |
828 | |
829 | return (unsigned) PERL_CTZ_32(word); | |
830 | ||
e1f6bdff | 831 | #elif U32SIZE == 4 && defined(_MSC_VER) |
2e0bc9ce | 832 | # define PERL_HAS_FAST_GET_LSB_POS32 |
e88dde50 KW |
833 | |
834 | { | |
835 | unsigned long index; | |
836 | _BitScanForward(&index, word); | |
837 | return (unsigned)index; | |
838 | } | |
839 | ||
fc1bb663 KW |
840 | #else |
841 | ||
19d2c525 | 842 | return single_1bit_pos32(word & (~word + 1)); |
fc1bb663 KW |
843 | |
844 | #endif | |
845 | ||
19d2c525 KW |
846 | } |
847 | ||
4a1b7bb2 | 848 | |
fc1bb663 | 849 | /* Convert the leading zeros count to the bit position of the first set bit. |
4a1b7bb2 KW |
850 | * This just subtracts from the highest position, 31 or 63. But some compilers |
851 | * don't optimize this optimally, and so a bit of bit twiddling encourages them | |
852 | * to do the right thing. It turns out that subtracting a smaller non-negative | |
853 | * number 'x' from 2**n-1 for any n is the same as taking the exclusive-or of | |
854 | * the two numbers. To see why, first note that the sum of any number, x, and | |
855 | * its complement, x', is all ones. So all ones minus x is x'. Then note that | |
856 | * the xor of x and all ones is x'. */ | |
857 | #define LZC_TO_MSBIT_POS_(size, lzc) ((size##SIZE * CHARBITS - 1) ^ (lzc)) | |
fc1bb663 | 858 | |
995a4954 KW |
859 | #ifdef U64TYPE /* HAS_QUAD not usable outside the core */ |
860 | ||
861 | PERL_STATIC_INLINE unsigned | |
330cd0ce KW |
862 | Perl_msbit_pos64(U64 word) |
863 | { | |
864 | /* Find the position (0..63) of the most significant set bit in the input | |
865 | * word */ | |
866 | ||
867 | ASSUME(word != 0); | |
868 | ||
fc1bb663 KW |
869 | /* If we can determine that the platform has a usable fast method to get |
870 | * this, use that */ | |
871 | ||
872 | # if defined(PERL_CLZ_64) | |
2e0bc9ce | 873 | # define PERL_HAS_FAST_GET_MSB_POS64 |
fc1bb663 KW |
874 | |
875 | return (unsigned) LZC_TO_MSBIT_POS_(U64, PERL_CLZ_64(word)); | |
876 | ||
e1f6bdff | 877 | # elif U64SIZE == 8 && defined(_WIN64) && defined(_MSC_VER) |
2e0bc9ce | 878 | # define PERL_HAS_FAST_GET_MSB_POS64 |
e88dde50 KW |
879 | |
880 | { | |
881 | unsigned long index; | |
882 | _BitScanReverse64(&index, word); | |
883 | return (unsigned)index; | |
884 | } | |
885 | ||
fc1bb663 KW |
886 | # else |
887 | ||
888 | /* Here, we didn't find a fast method for finding the msb. Fall back to | |
889 | * making the msb the only set bit in the word, and use our function that | |
890 | * works on words with a single bit set. | |
891 | * | |
892 | * Isolate the msb; http://codeforces.com/blog/entry/10330 | |
330cd0ce KW |
893 | * |
894 | * Only the most significant set bit matters. Or'ing word with its right | |
895 | * shift of 1 makes that bit and the next one to its right both 1. | |
896 | * Repeating that with the right shift of 2 makes for 4 1-bits in a row. | |
897 | * ... We end with the msb and all to the right being 1. */ | |
898 | word |= (word >> 1); | |
899 | word |= (word >> 2); | |
900 | word |= (word >> 4); | |
901 | word |= (word >> 8); | |
902 | word |= (word >> 16); | |
903 | word |= (word >> 32); | |
904 | ||
905 | /* Then subtracting the right shift by 1 clears all but the left-most of | |
906 | * the 1 bits, which is our desired result */ | |
907 | word -= (word >> 1); | |
908 | ||
909 | /* Now we have a single bit set */ | |
910 | return single_1bit_pos64(word); | |
fc1bb663 KW |
911 | |
912 | # endif | |
913 | ||
330cd0ce KW |
914 | } |
915 | ||
916 | # define msbit_pos_uintmax_(word) msbit_pos64(word) | |
917 | #else /* ! QUAD */ | |
918 | # define msbit_pos_uintmax_(word) msbit_pos32(word) | |
919 | #endif | |
920 | ||
921 | PERL_STATIC_INLINE unsigned | |
922 | Perl_msbit_pos32(U32 word) | |
923 | { | |
924 | /* Find the position (0..31) of the most significant set bit in the input | |
925 | * word */ | |
926 | ||
927 | ASSUME(word != 0); | |
928 | ||
fc1bb663 | 929 | #if defined(PERL_CLZ_32) |
2e0bc9ce | 930 | # define PERL_HAS_FAST_GET_MSB_POS32 |
fc1bb663 KW |
931 | |
932 | return (unsigned) LZC_TO_MSBIT_POS_(U32, PERL_CLZ_32(word)); | |
933 | ||
e1f6bdff | 934 | #elif U32SIZE == 4 && defined(_MSC_VER) |
2e0bc9ce | 935 | # define PERL_HAS_FAST_GET_MSB_POS32 |
e88dde50 KW |
936 | |
937 | { | |
938 | unsigned long index; | |
939 | _BitScanReverse(&index, word); | |
940 | return (unsigned)index; | |
941 | } | |
942 | ||
fc1bb663 KW |
943 | #else |
944 | ||
330cd0ce KW |
945 | word |= (word >> 1); |
946 | word |= (word >> 2); | |
947 | word |= (word >> 4); | |
948 | word |= (word >> 8); | |
949 | word |= (word >> 16); | |
950 | word -= (word >> 1); | |
951 | return single_1bit_pos32(word); | |
fc1bb663 KW |
952 | |
953 | #endif | |
954 | ||
330cd0ce KW |
955 | } |
956 | ||
787e8384 KW |
957 | #if UVSIZE == U64SIZE |
958 | # define msbit_pos(word) msbit_pos64(word) | |
959 | # define lsbit_pos(word) lsbit_pos64(word) | |
960 | #elif UVSIZE == U32SIZE | |
961 | # define msbit_pos(word) msbit_pos32(word) | |
962 | # define lsbit_pos(word) lsbit_pos32(word) | |
963 | #endif | |
964 | ||
330cd0ce KW |
965 | #ifdef U64TYPE /* HAS_QUAD not usable outside the core */ |
966 | ||
967 | PERL_STATIC_INLINE unsigned | |
995a4954 KW |
968 | Perl_single_1bit_pos64(U64 word) |
969 | { | |
970 | /* Given a 64-bit word known to contain all zero bits except one 1 bit, | |
971 | * find and return the 1's position: 0..63 */ | |
972 | ||
973 | # ifdef PERL_CORE /* macro not exported */ | |
974 | ASSUME(isPOWER_OF_2(word)); | |
975 | # else | |
976 | ASSUME(word && (word & (word-1)) == 0); | |
977 | # endif | |
978 | ||
2e0bc9ce KW |
979 | /* The only set bit is both the most and least significant bit. If we have |
980 | * a fast way of finding either one, use that. | |
981 | * | |
982 | * It may appear at first glance that those functions call this one, but | |
983 | * they don't if the corresponding #define is set */ | |
984 | ||
985 | # ifdef PERL_HAS_FAST_GET_MSB_POS64 | |
986 | ||
987 | return msbit_pos64(word); | |
988 | ||
989 | # elif defined(PERL_HAS_FAST_GET_LSB_POS64) | |
990 | ||
991 | return lsbit_pos64(word); | |
992 | ||
993 | # else | |
994 | ||
995a4954 KW |
995 | /* The position of the only set bit in a word can be quickly calculated |
996 | * using deBruijn sequences. See for example | |
997 | * https://en.wikipedia.org/wiki/De_Bruijn_sequence */ | |
998 | return PL_deBruijn_bitpos_tab64[(word * PERL_deBruijnMagic64_) | |
999 | >> PERL_deBruijnShift64_]; | |
2e0bc9ce KW |
1000 | # endif |
1001 | ||
995a4954 KW |
1002 | } |
1003 | ||
1004 | #endif | |
1005 | ||
bf874180 KW |
1006 | PERL_STATIC_INLINE unsigned |
1007 | Perl_single_1bit_pos32(U32 word) | |
1008 | { | |
1009 | /* Given a 32-bit word known to contain all zero bits except one 1 bit, | |
1010 | * find and return the 1's position: 0..31 */ | |
1011 | ||
1012 | #ifdef PERL_CORE /* macro not exported */ | |
1013 | ASSUME(isPOWER_OF_2(word)); | |
1014 | #else | |
1015 | ASSUME(word && (word & (word-1)) == 0); | |
1016 | #endif | |
2e0bc9ce KW |
1017 | #ifdef PERL_HAS_FAST_GET_MSB_POS32 |
1018 | ||
1019 | return msbit_pos32(word); | |
1020 | ||
1021 | #elif defined(PERL_HAS_FAST_GET_LSB_POS32) | |
1022 | ||
1023 | return lsbit_pos32(word); | |
1024 | ||
1025 | /* Unlikely, but possible for the platform to have a wider fast operation but | |
1026 | * not a narrower one. But easy enough to handle the case by widening the | |
1027 | * parameter size. (Going the other way, emulating 64 bit by two 32 bit ops | |
1028 | * would be slower than the deBruijn method.) */ | |
1029 | #elif defined(PERL_HAS_FAST_GET_MSB_POS64) | |
1030 | ||
1031 | return msbit_pos64(word); | |
1032 | ||
1033 | #elif defined(PERL_HAS_FAST_GET_LSB_POS64) | |
1034 | ||
1035 | return lsbit_pos64(word); | |
1036 | ||
1037 | #else | |
bf874180 | 1038 | |
bf874180 KW |
1039 | return PL_deBruijn_bitpos_tab32[(word * PERL_deBruijnMagic32_) |
1040 | >> PERL_deBruijnShift32_]; | |
2e0bc9ce KW |
1041 | #endif |
1042 | ||
bf874180 | 1043 | } |
1e599354 | 1044 | |
23a7ee81 KW |
1045 | #ifndef EBCDIC |
1046 | ||
1d2af574 | 1047 | PERL_STATIC_INLINE unsigned int |
73f0a2eb | 1048 | Perl_variant_byte_number(PERL_UINTMAX_T word) |
1d2af574 | 1049 | { |
1d2af574 KW |
1050 | /* This returns the position in a word (0..7) of the first variant byte in |
1051 | * it. This is a helper function. Note that there are no branches */ | |
1052 | ||
1d2af574 KW |
1053 | /* Get just the msb bits of each byte */ |
1054 | word &= PERL_VARIANTS_WORD_MASK; | |
1055 | ||
58ddb8c5 KW |
1056 | /* This should only be called if we know there is a variant byte in the |
1057 | * word */ | |
1058 | assert(word); | |
1059 | ||
7adf2470 | 1060 | # if BYTEORDER == 0x1234 || BYTEORDER == 0x12345678 |
1d2af574 KW |
1061 | |
1062 | /* Bytes are stored like | |
1063 | * Byte8 ... Byte2 Byte1 | |
1064 | * 63..56...15...8 7...0 | |
19d2c525 KW |
1065 | * so getting the lsb of the whole modified word is getting the msb of the |
1066 | * first byte that has its msb set */ | |
1067 | word = lsbit_pos_uintmax_(word); | |
1068 | ||
1069 | /* Here, word contains the position 7,15,23,...55,63 of that bit. Convert | |
1070 | * to 0..7 */ | |
1071 | return (unsigned int) ((word + 1) >> 3) - 1; | |
1d2af574 KW |
1072 | |
1073 | # elif BYTEORDER == 0x4321 || BYTEORDER == 0x87654321 | |
1074 | ||
1075 | /* Bytes are stored like | |
1076 | * Byte1 Byte2 ... Byte8 | |
1077 | * 63..56 55..47 ... 7...0 | |
330cd0ce KW |
1078 | * so getting the msb of the whole modified word is getting the msb of the |
1079 | * first byte that has its msb set */ | |
1080 | word = msbit_pos_uintmax_(word); | |
1d2af574 | 1081 | |
330cd0ce KW |
1082 | /* Here, word contains the position 63,55,...,23,15,7 of that bit. Convert |
1083 | * to 0..7 */ | |
1d2af574 KW |
1084 | word = ((word + 1) >> 3) - 1; |
1085 | ||
330cd0ce KW |
1086 | /* And invert the result because of the reversed byte order on this |
1087 | * platform */ | |
1d2af574 KW |
1088 | word = CHARBITS - word - 1; |
1089 | ||
330cd0ce KW |
1090 | return (unsigned int) word; |
1091 | ||
1092 | # else | |
1093 | # error Unexpected byte order | |
1d2af574 KW |
1094 | # endif |
1095 | ||
1d2af574 KW |
1096 | } |
1097 | ||
23a7ee81 | 1098 | #endif |
03c1e4ab KW |
1099 | #if defined(PERL_CORE) || defined(PERL_EXT) |
1100 | ||
1101 | /* | |
1102 | =for apidoc variant_under_utf8_count | |
1103 | ||
1104 | This function looks at the sequence of bytes between C<s> and C<e>, which are | |
1105 | assumed to be encoded in ASCII/Latin1, and returns how many of them would | |
1106 | change should the string be translated into UTF-8. Due to the nature of UTF-8, | |
1107 | each of these would occupy two bytes instead of the single one in the input | |
1108 | string. Thus, this function returns the precise number of bytes the string | |
1109 | would expand by when translated to UTF-8. | |
1110 | ||
1111 | Unlike most of the other functions that have C<utf8> in their name, the input | |
1112 | to this function is NOT a UTF-8-encoded string. The function name is slightly | |
1113 | I<odd> to emphasize this. | |
1114 | ||
1115 | This function is internal to Perl because khw thinks that any XS code that | |
1116 | would want this is probably operating too close to the internals. Presenting a | |
1117 | valid use case could change that. | |
1118 | ||
1119 | See also | |
1120 | C<L<perlapi/is_utf8_invariant_string>> | |
1121 | and | |
1122 | C<L<perlapi/is_utf8_invariant_string_loc>>, | |
1123 | ||
1124 | =cut | |
1125 | ||
1126 | */ | |
1127 | ||
1128 | PERL_STATIC_INLINE Size_t | |
1129 | S_variant_under_utf8_count(const U8* const s, const U8* const e) | |
1130 | { | |
1131 | const U8* x = s; | |
1132 | Size_t count = 0; | |
1133 | ||
1134 | PERL_ARGS_ASSERT_VARIANT_UNDER_UTF8_COUNT; | |
1135 | ||
1136 | # ifndef EBCDIC | |
1137 | ||
5d0379de KW |
1138 | /* Test if the string is long enough to use word-at-a-time. (Logic is the |
1139 | * same as for is_utf8_invariant_string()) */ | |
03c1e4ab KW |
1140 | if ((STRLEN) (e - x) >= PERL_WORDSIZE |
1141 | + PERL_WORDSIZE * PERL_IS_SUBWORD_ADDR(x) | |
1142 | - (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK)) | |
1143 | { | |
1144 | ||
1145 | /* Process per-byte until reach word boundary. XXX This loop could be | |
1146 | * eliminated if we knew that this platform had fast unaligned reads */ | |
1147 | while (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK) { | |
1148 | count += ! UTF8_IS_INVARIANT(*x++); | |
1149 | } | |
1150 | ||
1151 | /* Process per-word as long as we have at least a full word left */ | |
74472cc2 KW |
1152 | do { /* Commit 03c1e4ab1d6ee9062fb3f94b0ba31db6698724b1 contains an |
1153 | explanation of how this works */ | |
e5863284 KW |
1154 | PERL_UINTMAX_T increment |
1155 | = ((((* (PERL_UINTMAX_T *) x) & PERL_VARIANTS_WORD_MASK) >> 7) | |
03c1e4ab KW |
1156 | * PERL_COUNT_MULTIPLIER) |
1157 | >> ((PERL_WORDSIZE - 1) * CHARBITS); | |
e5863284 | 1158 | count += (Size_t) increment; |
03c1e4ab KW |
1159 | x += PERL_WORDSIZE; |
1160 | } while (x + PERL_WORDSIZE <= e); | |
1161 | } | |
1162 | ||
1163 | # endif | |
1164 | ||
1165 | /* Process per-byte */ | |
1166 | while (x < e) { | |
1604cfb0 | 1167 | if (! UTF8_IS_INVARIANT(*x)) { |
03c1e4ab KW |
1168 | count++; |
1169 | } | |
1170 | ||
1171 | x++; | |
1172 | } | |
1173 | ||
1174 | return count; | |
1175 | } | |
1176 | ||
1177 | #endif | |
1178 | ||
aff4cafe KW |
1179 | #ifndef PERL_IN_REGEXEC_C /* Keep these around for that file */ |
1180 | # undef PERL_WORDSIZE | |
1181 | # undef PERL_COUNT_MULTIPLIER | |
1182 | # undef PERL_WORD_BOUNDARY_MASK | |
1183 | # undef PERL_VARIANTS_WORD_MASK | |
1184 | #endif | |
03c1e4ab | 1185 | |
7c93d8f0 | 1186 | /* |
5ff889fb KW |
1187 | =for apidoc is_utf8_string |
1188 | ||
82c5d941 KW |
1189 | Returns TRUE if the first C<len> bytes of string C<s> form a valid |
1190 | Perl-extended-UTF-8 string; returns FALSE otherwise. If C<len> is 0, it will | |
1191 | be calculated using C<strlen(s)> (which means if you use this option, that C<s> | |
1192 | can't have embedded C<NUL> characters and has to have a terminating C<NUL> | |
1193 | byte). Note that all characters being ASCII constitute 'a valid UTF-8 string'. | |
1194 | ||
2717076a KW |
1195 | This function considers Perl's extended UTF-8 to be valid. That means that |
1196 | code points above Unicode, surrogates, and non-character code points are | |
9f2abfde KW |
1197 | considered valid by this function. Use C<L</is_strict_utf8_string>>, |
1198 | C<L</is_c9strict_utf8_string>>, or C<L</is_utf8_string_flags>> to restrict what | |
1199 | code points are considered valid. | |
5ff889fb | 1200 | |
9f2abfde KW |
1201 | See also |
1202 | C<L</is_utf8_invariant_string>>, | |
0cbf5865 | 1203 | C<L</is_utf8_invariant_string_loc>>, |
9f2abfde KW |
1204 | C<L</is_utf8_string_loc>>, |
1205 | C<L</is_utf8_string_loclen>>, | |
8bc127bf KW |
1206 | C<L</is_utf8_fixed_width_buf_flags>>, |
1207 | C<L</is_utf8_fixed_width_buf_loc_flags>>, | |
1208 | C<L</is_utf8_fixed_width_buf_loclen_flags>>, | |
5ff889fb KW |
1209 | |
1210 | =cut | |
1211 | */ | |
1212 | ||
dd237e82 | 1213 | #define is_utf8_string(s, len) is_utf8_string_loclen(s, len, NULL, NULL) |
5ff889fb | 1214 | |
c9cd936b KW |
1215 | #if defined(PERL_CORE) || defined (PERL_EXT) |
1216 | ||
1217 | /* | |
1218 | =for apidoc is_utf8_non_invariant_string | |
1219 | ||
1220 | Returns TRUE if L<perlapi/is_utf8_invariant_string> returns FALSE for the first | |
1221 | C<len> bytes of the string C<s>, but they are, nonetheless, legal Perl-extended | |
1222 | UTF-8; otherwise returns FALSE. | |
1223 | ||
1224 | A TRUE return means that at least one code point represented by the sequence | |
1225 | either is a wide character not representable as a single byte, or the | |
1226 | representation differs depending on whether the sequence is encoded in UTF-8 or | |
1227 | not. | |
1228 | ||
1229 | See also | |
1230 | C<L<perlapi/is_utf8_invariant_string>>, | |
1231 | C<L<perlapi/is_utf8_string>> | |
1232 | ||
1233 | =cut | |
1234 | ||
1235 | This is commonly used to determine if a SV's UTF-8 flag should be turned on. | |
b3b93dfe KW |
1236 | It generally needn't be if its string is entirely UTF-8 invariant, and it |
1237 | shouldn't be if it otherwise contains invalid UTF-8. | |
c9cd936b KW |
1238 | |
1239 | It is an internal function because khw thinks that XS code shouldn't be working | |
1240 | at this low a level. A valid use case could change that. | |
1241 | ||
1242 | */ | |
1243 | ||
1244 | PERL_STATIC_INLINE bool | |
86a87e17 | 1245 | Perl_is_utf8_non_invariant_string(const U8* const s, STRLEN len) |
c9cd936b KW |
1246 | { |
1247 | const U8 * first_variant; | |
1248 | ||
1249 | PERL_ARGS_ASSERT_IS_UTF8_NON_INVARIANT_STRING; | |
1250 | ||
1251 | if (is_utf8_invariant_string_loc(s, len, &first_variant)) { | |
1252 | return FALSE; | |
1253 | } | |
1254 | ||
1255 | return is_utf8_string(first_variant, len - (first_variant - s)); | |
1256 | } | |
1257 | ||
1258 | #endif | |
1259 | ||
5ff889fb | 1260 | /* |
9f2abfde KW |
1261 | =for apidoc is_strict_utf8_string |
1262 | ||
1263 | Returns TRUE if the first C<len> bytes of string C<s> form a valid | |
1264 | UTF-8-encoded string that is fully interchangeable by any application using | |
1265 | Unicode rules; otherwise it returns FALSE. If C<len> is 0, it will be | |
1266 | calculated using C<strlen(s)> (which means if you use this option, that C<s> | |
1267 | can't have embedded C<NUL> characters and has to have a terminating C<NUL> | |
1268 | byte). Note that all characters being ASCII constitute 'a valid UTF-8 string'. | |
1269 | ||
1270 | This function returns FALSE for strings containing any | |
1271 | code points above the Unicode max of 0x10FFFF, surrogate code points, or | |
1272 | non-character code points. | |
1273 | ||
1274 | See also | |
1275 | C<L</is_utf8_invariant_string>>, | |
0cbf5865 | 1276 | C<L</is_utf8_invariant_string_loc>>, |
9f2abfde KW |
1277 | C<L</is_utf8_string>>, |
1278 | C<L</is_utf8_string_flags>>, | |
1279 | C<L</is_utf8_string_loc>>, | |
1280 | C<L</is_utf8_string_loc_flags>>, | |
1281 | C<L</is_utf8_string_loclen>>, | |
1282 | C<L</is_utf8_string_loclen_flags>>, | |
8bc127bf KW |
1283 | C<L</is_utf8_fixed_width_buf_flags>>, |
1284 | C<L</is_utf8_fixed_width_buf_loc_flags>>, | |
1285 | C<L</is_utf8_fixed_width_buf_loclen_flags>>, | |
9f2abfde KW |
1286 | C<L</is_strict_utf8_string_loc>>, |
1287 | C<L</is_strict_utf8_string_loclen>>, | |
1288 | C<L</is_c9strict_utf8_string>>, | |
1289 | C<L</is_c9strict_utf8_string_loc>>, | |
1290 | and | |
1291 | C<L</is_c9strict_utf8_string_loclen>>. | |
1292 | ||
1293 | =cut | |
1294 | */ | |
1295 | ||
dd237e82 | 1296 | #define is_strict_utf8_string(s, len) is_strict_utf8_string_loclen(s, len, NULL, NULL) |
9f2abfde KW |
1297 | |
1298 | /* | |
1299 | =for apidoc is_c9strict_utf8_string | |
1300 | ||
1301 | Returns TRUE if the first C<len> bytes of string C<s> form a valid | |
1302 | UTF-8-encoded string that conforms to | |
1303 | L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>; | |
1304 | otherwise it returns FALSE. If C<len> is 0, it will be calculated using | |
1305 | C<strlen(s)> (which means if you use this option, that C<s> can't have embedded | |
1306 | C<NUL> characters and has to have a terminating C<NUL> byte). Note that all | |
1307 | characters being ASCII constitute 'a valid UTF-8 string'. | |
1308 | ||
1309 | This function returns FALSE for strings containing any code points above the | |
1310 | Unicode max of 0x10FFFF or surrogate code points, but accepts non-character | |
1311 | code points per | |
1312 | L<Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>. | |
1313 | ||
1314 | See also | |
1315 | C<L</is_utf8_invariant_string>>, | |
0cbf5865 | 1316 | C<L</is_utf8_invariant_string_loc>>, |
9f2abfde KW |
1317 | C<L</is_utf8_string>>, |
1318 | C<L</is_utf8_string_flags>>, | |
1319 | C<L</is_utf8_string_loc>>, | |
1320 | C<L</is_utf8_string_loc_flags>>, | |
1321 | C<L</is_utf8_string_loclen>>, | |
1322 | C<L</is_utf8_string_loclen_flags>>, | |
8bc127bf KW |
1323 | C<L</is_utf8_fixed_width_buf_flags>>, |
1324 | C<L</is_utf8_fixed_width_buf_loc_flags>>, | |
1325 | C<L</is_utf8_fixed_width_buf_loclen_flags>>, | |
9f2abfde KW |
1326 | C<L</is_strict_utf8_string>>, |
1327 | C<L</is_strict_utf8_string_loc>>, | |
1328 | C<L</is_strict_utf8_string_loclen>>, | |
1329 | C<L</is_c9strict_utf8_string_loc>>, | |
1330 | and | |
1331 | C<L</is_c9strict_utf8_string_loclen>>. | |
1332 | ||
1333 | =cut | |
1334 | */ | |
1335 | ||
dd237e82 | 1336 | #define is_c9strict_utf8_string(s, len) is_c9strict_utf8_string_loclen(s, len, NULL, 0) |
9f2abfde KW |
1337 | |
1338 | /* | |
1339 | =for apidoc is_utf8_string_flags | |
1340 | ||
1341 | Returns TRUE if the first C<len> bytes of string C<s> form a valid | |
1342 | UTF-8 string, subject to the restrictions imposed by C<flags>; | |
1343 | returns FALSE otherwise. If C<len> is 0, it will be calculated | |
1344 | using C<strlen(s)> (which means if you use this option, that C<s> can't have | |
1345 | embedded C<NUL> characters and has to have a terminating C<NUL> byte). Note | |
1346 | that all characters being ASCII constitute 'a valid UTF-8 string'. | |
1347 | ||
1348 | If C<flags> is 0, this gives the same results as C<L</is_utf8_string>>; if | |
1349 | C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results | |
1350 | as C<L</is_strict_utf8_string>>; and if C<flags> is | |
1351 | C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives the same results as | |
1352 | C<L</is_c9strict_utf8_string>>. Otherwise C<flags> may be any | |
1353 | combination of the C<UTF8_DISALLOW_I<foo>> flags understood by | |
1354 | C<L</utf8n_to_uvchr>>, with the same meanings. | |
1355 | ||
1356 | See also | |
1357 | C<L</is_utf8_invariant_string>>, | |
0cbf5865 | 1358 | C<L</is_utf8_invariant_string_loc>>, |
9f2abfde KW |
1359 | C<L</is_utf8_string>>, |
1360 | C<L</is_utf8_string_loc>>, | |
1361 | C<L</is_utf8_string_loc_flags>>, | |
1362 | C<L</is_utf8_string_loclen>>, | |
1363 | C<L</is_utf8_string_loclen_flags>>, | |
8bc127bf KW |
1364 | C<L</is_utf8_fixed_width_buf_flags>>, |
1365 | C<L</is_utf8_fixed_width_buf_loc_flags>>, | |
1366 | C<L</is_utf8_fixed_width_buf_loclen_flags>>, | |
9f2abfde KW |
1367 | C<L</is_strict_utf8_string>>, |
1368 | C<L</is_strict_utf8_string_loc>>, | |
1369 | C<L</is_strict_utf8_string_loclen>>, | |
1370 | C<L</is_c9strict_utf8_string>>, | |
1371 | C<L</is_c9strict_utf8_string_loc>>, | |
1372 | and | |
1373 | C<L</is_c9strict_utf8_string_loclen>>. | |
1374 | ||
1375 | =cut | |
1376 | */ | |
1377 | ||
1378 | PERL_STATIC_INLINE bool | |
c9182d9c | 1379 | Perl_is_utf8_string_flags(const U8 *s, STRLEN len, const U32 flags) |
9f2abfde | 1380 | { |
33756530 | 1381 | const U8 * first_variant; |
9f2abfde KW |
1382 | |
1383 | PERL_ARGS_ASSERT_IS_UTF8_STRING_FLAGS; | |
1384 | assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE | |
d044b7a7 | 1385 | |UTF8_DISALLOW_PERL_EXTENDED))); |
9f2abfde | 1386 | |
f60f61fd KW |
1387 | if (len == 0) { |
1388 | len = strlen((const char *)s); | |
1389 | } | |
1390 | ||
9f2abfde KW |
1391 | if (flags == 0) { |
1392 | return is_utf8_string(s, len); | |
1393 | } | |
1394 | ||
d044b7a7 | 1395 | if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED) |
9f2abfde KW |
1396 | == UTF8_DISALLOW_ILLEGAL_INTERCHANGE) |
1397 | { | |
1398 | return is_strict_utf8_string(s, len); | |
1399 | } | |
1400 | ||
d044b7a7 | 1401 | if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED) |
9f2abfde KW |
1402 | == UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE) |
1403 | { | |
1404 | return is_c9strict_utf8_string(s, len); | |
1405 | } | |
1406 | ||
33756530 KW |
1407 | if (! is_utf8_invariant_string_loc(s, len, &first_variant)) { |
1408 | const U8* const send = s + len; | |
1409 | const U8* x = first_variant; | |
1410 | ||
a0d7f935 KW |
1411 | while (x < send) { |
1412 | STRLEN cur_len = isUTF8_CHAR_flags(x, send, flags); | |
1413 | if (UNLIKELY(! cur_len)) { | |
1414 | return FALSE; | |
1415 | } | |
1416 | x += cur_len; | |
9f2abfde | 1417 | } |
33756530 | 1418 | } |
9f2abfde KW |
1419 | |
1420 | return TRUE; | |
1421 | } | |
1422 | ||
1423 | /* | |
5ff889fb KW |
1424 | |
1425 | =for apidoc is_utf8_string_loc | |
1426 | ||
2717076a | 1427 | Like C<L</is_utf8_string>> but stores the location of the failure (in the |
5ff889fb | 1428 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of |
82c5d941 | 1429 | "utf8ness success") in the C<ep> pointer. |
5ff889fb | 1430 | |
2717076a | 1431 | See also C<L</is_utf8_string_loclen>>. |
5ff889fb | 1432 | |
3964c812 KW |
1433 | =cut |
1434 | */ | |
1435 | ||
1436 | #define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0) | |
1437 | ||
1438 | /* | |
1439 | ||
5ff889fb KW |
1440 | =for apidoc is_utf8_string_loclen |
1441 | ||
2717076a | 1442 | Like C<L</is_utf8_string>> but stores the location of the failure (in the |
5ff889fb | 1443 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of |
9f2abfde | 1444 | "utf8ness success") in the C<ep> pointer, and the number of UTF-8 |
82c5d941 | 1445 | encoded characters in the C<el> pointer. |
5ff889fb | 1446 | |
2717076a | 1447 | See also C<L</is_utf8_string_loc>>. |
5ff889fb KW |
1448 | |
1449 | =cut | |
1450 | */ | |
1451 | ||
56e4cf64 | 1452 | PERL_STATIC_INLINE bool |
33756530 | 1453 | Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el) |
5ff889fb | 1454 | { |
33756530 | 1455 | const U8 * first_variant; |
5ff889fb KW |
1456 | |
1457 | PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN; | |
1458 | ||
33756530 KW |
1459 | if (len == 0) { |
1460 | len = strlen((const char *) s); | |
1461 | } | |
1462 | ||
1463 | if (is_utf8_invariant_string_loc(s, len, &first_variant)) { | |
1464 | if (el) | |
1465 | *el = len; | |
1466 | ||
1467 | if (ep) { | |
1468 | *ep = s + len; | |
1469 | } | |
1470 | ||
1471 | return TRUE; | |
1472 | } | |
1473 | ||
1474 | { | |
1475 | const U8* const send = s + len; | |
1476 | const U8* x = first_variant; | |
1477 | STRLEN outlen = first_variant - s; | |
1478 | ||
a0d7f935 KW |
1479 | while (x < send) { |
1480 | const STRLEN cur_len = isUTF8_CHAR(x, send); | |
1481 | if (UNLIKELY(! cur_len)) { | |
1482 | break; | |
1483 | } | |
1484 | x += cur_len; | |
1485 | outlen++; | |
5ff889fb | 1486 | } |
5ff889fb | 1487 | |
a0d7f935 KW |
1488 | if (el) |
1489 | *el = outlen; | |
5ff889fb | 1490 | |
a0d7f935 KW |
1491 | if (ep) { |
1492 | *ep = x; | |
1493 | } | |
5ff889fb | 1494 | |
a0d7f935 | 1495 | return (x == send); |
33756530 | 1496 | } |
5ff889fb KW |
1497 | } |
1498 | ||
213dc9d1 KW |
1499 | /* The perl core arranges to never call the DFA below without there being at |
1500 | * least one byte available to look at. This allows the DFA to use a do {} | |
1501 | * while loop which means that calling it with a UTF-8 invariant has a single | |
1502 | * conditional, same as the calling code checking for invariance ahead of time. | |
1503 | * And having the calling code remove that conditional speeds up by that | |
1504 | * conditional, the case where it wasn't invariant. So there's no reason to | |
1505 | * check before caling this. | |
1506 | * | |
1507 | * But we don't know this for non-core calls, so have to retain the check for | |
1508 | * them. */ | |
1509 | #ifdef PERL_CORE | |
1510 | # define PERL_NON_CORE_CHECK_EMPTY(s,e) assert((e) > (s)) | |
1511 | #else | |
1512 | # define PERL_NON_CORE_CHECK_EMPTY(s,e) if ((e) <= (s)) return FALSE | |
1513 | #endif | |
1514 | ||
5ff889fb | 1515 | /* |
50f7a4ce KW |
1516 | * DFA for checking input is valid UTF-8 syntax. |
1517 | * | |
1518 | * This uses adaptations of the table and algorithm given in | |
1519 | * https://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides comprehensive | |
1520 | * documentation of the original version. A copyright notice for the original | |
1521 | * version is given at the beginning of this file. The Perl adapations are | |
1522 | * documented at the definition of PL_extended_utf8_dfa_tab[]. | |
1523 | * | |
1524 | * This dfa is fast. There are three exit conditions: | |
1525 | * 1) a well-formed code point, acceptable to the table | |
1526 | * 2) the beginning bytes of an incomplete character, whose completion might | |
1527 | * or might not be acceptable | |
1528 | * 3) unacceptable to the table. Some of the adaptations have certain, | |
1529 | * hopefully less likely to occur, legal inputs be unacceptable to the | |
1530 | * table, so these must be sorted out afterwards. | |
1531 | * | |
1532 | * This macro is a complete implementation of the code executing the DFA. It | |
1533 | * is passed the input sequence bounds and the table to use, and what to do | |
1534 | * for each of the exit conditions. There are three canned actions, likely to | |
1535 | * be the ones you want: | |
1536 | * DFA_RETURN_SUCCESS_ | |
1537 | * DFA_RETURN_FAILURE_ | |
1538 | * DFA_GOTO_TEASE_APART_FF_ | |
1539 | * | |
1540 | * You pass a parameter giving the action to take for each of the three | |
1541 | * possible exit conditions: | |
1542 | * | |
1543 | * 'accept_action' This is executed when the DFA accepts the input. | |
1544 | * DFA_RETURN_SUCCESS_ is the most likely candidate. | |
1545 | * 'reject_action' This is executed when the DFA rejects the input. | |
1546 | * DFA_RETURN_FAILURE_ is a candidate, or 'goto label' where | |
1547 | * you have written code to distinguish the rejecting state | |
1548 | * results. Because it happens in several places, and | |
1549 | * involves #ifdefs, the special action | |
1550 | * DFA_GOTO_TEASE_APART_FF_ is what you want with | |
1551 | * PL_extended_utf8_dfa_tab. On platforms without | |
1552 | * EXTRA_LONG_UTF8, there is no need to tease anything apart, | |
1553 | * so this evaluates to DFA_RETURN_FAILURE_; otherwise you | |
1554 | * need to have a label 'tease_apart_FF' that it will transfer | |
1555 | * to. | |
1556 | * 'incomplete_char_action' This is executed when the DFA ran off the end | |
1557 | * before accepting or rejecting the input. | |
1558 | * DFA_RETURN_FAILURE_ is the likely action, but you could | |
1559 | * have a 'goto', or NOOP. In the latter case the DFA drops | |
1560 | * off the end, and you place your code to handle this case | |
1561 | * immediately after it. | |
1562 | */ | |
1563 | ||
1564 | #define DFA_RETURN_SUCCESS_ return s - s0 | |
1565 | #define DFA_RETURN_FAILURE_ return 0 | |
1566 | #ifdef HAS_EXTRA_LONG_UTF8 | |
1567 | # define DFA_TEASE_APART_FF_ goto tease_apart_FF | |
1568 | #else | |
1569 | # define DFA_TEASE_APART_FF_ DFA_RETURN_FAILURE_ | |
1570 | #endif | |
1571 | ||
1572 | #define PERL_IS_UTF8_CHAR_DFA(s0, e, dfa_tab, \ | |
1573 | accept_action, \ | |
1574 | reject_action, \ | |
1575 | incomplete_char_action) \ | |
1576 | STMT_START { \ | |
1577 | const U8 * s = s0; \ | |
1578 | UV state = 0; \ | |
1579 | \ | |
213dc9d1 KW |
1580 | PERL_NON_CORE_CHECK_EMPTY(s,e); \ |
1581 | \ | |
1582 | do { \ | |
50f7a4ce KW |
1583 | state = dfa_tab[256 + state + dfa_tab[*s]]; \ |
1584 | s++; \ | |
1585 | \ | |
1586 | if (state == 0) { /* Accepting state */ \ | |
1587 | accept_action; \ | |
1588 | } \ | |
1589 | \ | |
1590 | if (UNLIKELY(state == 1)) { /* Rejecting state */ \ | |
1591 | reject_action; \ | |
1592 | } \ | |
213dc9d1 | 1593 | } while (s < e); \ |
50f7a4ce KW |
1594 | \ |
1595 | /* Here, dropped out of loop before end-of-char */ \ | |
1596 | incomplete_char_action; \ | |
1597 | } STMT_END | |
1598 | ||
1599 | ||
1600 | /* | |
9f2abfde | 1601 | |
44170c9a | 1602 | =for apidoc isUTF8_CHAR |
8ed185f9 KW |
1603 | |
1604 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
1605 | looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl, | |
1606 | that represents some code point; otherwise it evaluates to 0. If non-zero, the | |
1607 | value gives how many bytes starting at C<s> comprise the code point's | |
1608 | representation. Any bytes remaining before C<e>, but beyond the ones needed to | |
1609 | form the first code point in C<s>, are not examined. | |
1610 | ||
13aab5dd | 1611 | The code point can be any that will fit in an IV on this machine, using Perl's |
8ed185f9 KW |
1612 | extension to official UTF-8 to represent those higher than the Unicode maximum |
1613 | of 0x10FFFF. That means that this macro is used to efficiently decide if the | |
1614 | next few bytes in C<s> is legal UTF-8 for a single character. | |
1615 | ||
1616 | Use C<L</isSTRICT_UTF8_CHAR>> to restrict the acceptable code points to those | |
1617 | defined by Unicode to be fully interchangeable across applications; | |
1618 | C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum | |
1619 | #9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable | |
1620 | code points; and C<L</isUTF8_CHAR_flags>> for a more customized definition. | |
1621 | ||
1622 | Use C<L</is_utf8_string>>, C<L</is_utf8_string_loc>>, and | |
1623 | C<L</is_utf8_string_loclen>> to check entire strings. | |
1624 | ||
13aab5dd KW |
1625 | Note also that a UTF-8 "invariant" character (i.e. ASCII on non-EBCDIC |
1626 | machines) is a valid UTF-8 character. | |
8ed185f9 KW |
1627 | |
1628 | =cut | |
1629 | ||
1630 | This uses an adaptation of the table and algorithm given in | |
f6521f7c | 1631 | https://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides comprehensive |
8ed185f9 KW |
1632 | documentation of the original version. A copyright notice for the original |
1633 | version is given at the beginning of this file. The Perl adapation is | |
71525f77 | 1634 | documented at the definition of PL_extended_utf8_dfa_tab[]. |
8ed185f9 KW |
1635 | */ |
1636 | ||
1637 | PERL_STATIC_INLINE Size_t | |
c9182d9c | 1638 | Perl_isUTF8_CHAR(const U8 * const s0, const U8 * const e) |
8ed185f9 | 1639 | { |
8ed185f9 KW |
1640 | PERL_ARGS_ASSERT_ISUTF8_CHAR; |
1641 | ||
50f7a4ce KW |
1642 | PERL_IS_UTF8_CHAR_DFA(s0, e, PL_extended_utf8_dfa_tab, |
1643 | DFA_RETURN_SUCCESS_, | |
1644 | DFA_TEASE_APART_FF_, | |
1645 | DFA_RETURN_FAILURE_); | |
8ed185f9 | 1646 | |
50f7a4ce KW |
1647 | /* Here, we didn't return success, but dropped out of the loop. In the |
1648 | * case of PL_extended_utf8_dfa_tab, this means the input is either | |
1649 | * malformed, or the start byte was FF on a platform that the dfa doesn't | |
1650 | * handle FF's. Call a helper function. */ | |
ffea7477 | 1651 | |
50f7a4ce | 1652 | #ifdef HAS_EXTRA_LONG_UTF8 |
8ed185f9 | 1653 | |
50f7a4ce | 1654 | tease_apart_FF: |
8ed185f9 | 1655 | |
50f7a4ce KW |
1656 | /* In the case of PL_extended_utf8_dfa_tab, getting here means the input is |
1657 | * either malformed, or was for the largest possible start byte, which we | |
1658 | * now check, not inline */ | |
1659 | if (*s0 != I8_TO_NATIVE_UTF8(0xFF)) { | |
1660 | return 0; | |
8ed185f9 KW |
1661 | } |
1662 | ||
50f7a4ce KW |
1663 | return is_utf8_FF_helper_(s0, e, |
1664 | FALSE /* require full, not partial char */ | |
1665 | ); | |
8ed185f9 KW |
1666 | #endif |
1667 | ||
8ed185f9 KW |
1668 | } |
1669 | ||
1670 | /* | |
1671 | ||
67049a5f KW |
1672 | =for apidoc isSTRICT_UTF8_CHAR |
1673 | ||
1674 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
1675 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some | |
1676 | Unicode code point completely acceptable for open interchange between all | |
1677 | applications; otherwise it evaluates to 0. If non-zero, the value gives how | |
1678 | many bytes starting at C<s> comprise the code point's representation. Any | |
1679 | bytes remaining before C<e>, but beyond the ones needed to form the first code | |
1680 | point in C<s>, are not examined. | |
1681 | ||
1682 | The largest acceptable code point is the Unicode maximum 0x10FFFF, and must not | |
1683 | be a surrogate nor a non-character code point. Thus this excludes any code | |
1684 | point from Perl's extended UTF-8. | |
1685 | ||
1686 | This is used to efficiently decide if the next few bytes in C<s> is | |
1687 | legal Unicode-acceptable UTF-8 for a single character. | |
1688 | ||
1689 | Use C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum | |
1690 | #9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable | |
1691 | code points; C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8; | |
1692 | and C<L</isUTF8_CHAR_flags>> for a more customized definition. | |
1693 | ||
1694 | Use C<L</is_strict_utf8_string>>, C<L</is_strict_utf8_string_loc>>, and | |
1695 | C<L</is_strict_utf8_string_loclen>> to check entire strings. | |
1696 | ||
1697 | =cut | |
1698 | ||
1699 | This uses an adaptation of the tables and algorithm given in | |
f6521f7c | 1700 | https://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides comprehensive |
67049a5f KW |
1701 | documentation of the original version. A copyright notice for the original |
1702 | version is given at the beginning of this file. The Perl adapation is | |
1703 | documented at the definition of strict_extended_utf8_dfa_tab[]. | |
1704 | ||
1705 | */ | |
1706 | ||
1707 | PERL_STATIC_INLINE Size_t | |
c9182d9c | 1708 | Perl_isSTRICT_UTF8_CHAR(const U8 * const s0, const U8 * const e) |
67049a5f | 1709 | { |
67049a5f KW |
1710 | PERL_ARGS_ASSERT_ISSTRICT_UTF8_CHAR; |
1711 | ||
50f7a4ce KW |
1712 | PERL_IS_UTF8_CHAR_DFA(s0, e, PL_strict_utf8_dfa_tab, |
1713 | DFA_RETURN_SUCCESS_, | |
1714 | goto check_hanguls, | |
1715 | DFA_RETURN_FAILURE_); | |
1716 | check_hanguls: | |
67049a5f | 1717 | |
50f7a4ce KW |
1718 | /* Here, we didn't return success, but dropped out of the loop. In the |
1719 | * case of PL_strict_utf8_dfa_tab, this means the input is either | |
1720 | * malformed, or was for certain Hanguls; handle them specially */ | |
67049a5f | 1721 | |
67260a96 KW |
1722 | /* The dfa above drops out for incomplete or illegal inputs, and certain |
1723 | * legal Hanguls; check and return accordingly */ | |
1724 | return is_HANGUL_ED_utf8_safe(s0, e); | |
67049a5f KW |
1725 | } |
1726 | ||
1727 | /* | |
1728 | ||
44170c9a | 1729 | =for apidoc isC9_STRICT_UTF8_CHAR |
c5bfbb64 KW |
1730 | |
1731 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
1732 | looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some | |
1733 | Unicode non-surrogate code point; otherwise it evaluates to 0. If non-zero, | |
1734 | the value gives how many bytes starting at C<s> comprise the code point's | |
1735 | representation. Any bytes remaining before C<e>, but beyond the ones needed to | |
1736 | form the first code point in C<s>, are not examined. | |
1737 | ||
1738 | The largest acceptable code point is the Unicode maximum 0x10FFFF. This | |
1739 | differs from C<L</isSTRICT_UTF8_CHAR>> only in that it accepts non-character | |
1740 | code points. This corresponds to | |
1741 | L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>. | |
1742 | which said that non-character code points are merely discouraged rather than | |
1743 | completely forbidden in open interchange. See | |
1744 | L<perlunicode/Noncharacter code points>. | |
1745 | ||
1746 | Use C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8; and | |
1747 | C<L</isUTF8_CHAR_flags>> for a more customized definition. | |
1748 | ||
1749 | Use C<L</is_c9strict_utf8_string>>, C<L</is_c9strict_utf8_string_loc>>, and | |
1750 | C<L</is_c9strict_utf8_string_loclen>> to check entire strings. | |
1751 | ||
1752 | =cut | |
1753 | ||
1754 | This uses an adaptation of the tables and algorithm given in | |
f6521f7c | 1755 | https://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides comprehensive |
c5bfbb64 KW |
1756 | documentation of the original version. A copyright notice for the original |
1757 | version is given at the beginning of this file. The Perl adapation is | |
71525f77 | 1758 | documented at the definition of PL_c9_utf8_dfa_tab[]. |
c5bfbb64 KW |
1759 | |
1760 | */ | |
1761 | ||
1762 | PERL_STATIC_INLINE Size_t | |
c9182d9c | 1763 | Perl_isC9_STRICT_UTF8_CHAR(const U8 * const s0, const U8 * const e) |
c5bfbb64 | 1764 | { |
c5bfbb64 KW |
1765 | PERL_ARGS_ASSERT_ISC9_STRICT_UTF8_CHAR; |
1766 | ||
50f7a4ce KW |
1767 | PERL_IS_UTF8_CHAR_DFA(s0, e, PL_c9_utf8_dfa_tab, |
1768 | DFA_RETURN_SUCCESS_, | |
1769 | DFA_RETURN_FAILURE_, | |
1770 | DFA_RETURN_FAILURE_); | |
c5bfbb64 KW |
1771 | } |
1772 | ||
1773 | /* | |
1774 | ||
9f2abfde KW |
1775 | =for apidoc is_strict_utf8_string_loc |
1776 | ||
1777 | Like C<L</is_strict_utf8_string>> but stores the location of the failure (in the | |
1778 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of | |
1779 | "utf8ness success") in the C<ep> pointer. | |
1780 | ||
1781 | See also C<L</is_strict_utf8_string_loclen>>. | |
1782 | ||
1783 | =cut | |
1784 | */ | |
1785 | ||
1786 | #define is_strict_utf8_string_loc(s, len, ep) \ | |
1787 | is_strict_utf8_string_loclen(s, len, ep, 0) | |
1788 | ||
1789 | /* | |
1790 | ||
1791 | =for apidoc is_strict_utf8_string_loclen | |
1792 | ||
1793 | Like C<L</is_strict_utf8_string>> but stores the location of the failure (in the | |
1794 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of | |
1795 | "utf8ness success") in the C<ep> pointer, and the number of UTF-8 | |
1796 | encoded characters in the C<el> pointer. | |
1797 | ||
1798 | See also C<L</is_strict_utf8_string_loc>>. | |
1799 | ||
1800 | =cut | |
1801 | */ | |
1802 | ||
1803 | PERL_STATIC_INLINE bool | |
c9182d9c | 1804 | Perl_is_strict_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el) |
9f2abfde | 1805 | { |
33756530 | 1806 | const U8 * first_variant; |
9f2abfde KW |
1807 | |
1808 | PERL_ARGS_ASSERT_IS_STRICT_UTF8_STRING_LOCLEN; | |
1809 | ||
33756530 KW |
1810 | if (len == 0) { |
1811 | len = strlen((const char *) s); | |
1812 | } | |
1813 | ||
1814 | if (is_utf8_invariant_string_loc(s, len, &first_variant)) { | |
1815 | if (el) | |
1816 | *el = len; | |
1817 | ||
1818 | if (ep) { | |
1819 | *ep = s + len; | |
1820 | } | |
1821 | ||
1822 | return TRUE; | |
1823 | } | |
1824 | ||
1825 | { | |
1826 | const U8* const send = s + len; | |
1827 | const U8* x = first_variant; | |
1828 | STRLEN outlen = first_variant - s; | |
1829 | ||
a0d7f935 KW |
1830 | while (x < send) { |
1831 | const STRLEN cur_len = isSTRICT_UTF8_CHAR(x, send); | |
1832 | if (UNLIKELY(! cur_len)) { | |
1833 | break; | |
1834 | } | |
1835 | x += cur_len; | |
1836 | outlen++; | |
9f2abfde | 1837 | } |
9f2abfde | 1838 | |
a0d7f935 KW |
1839 | if (el) |
1840 | *el = outlen; | |
9f2abfde | 1841 | |
a0d7f935 KW |
1842 | if (ep) { |
1843 | *ep = x; | |
1844 | } | |
9f2abfde | 1845 | |
a0d7f935 | 1846 | return (x == send); |
33756530 | 1847 | } |
9f2abfde KW |
1848 | } |
1849 | ||
1850 | /* | |
1851 | ||
1852 | =for apidoc is_c9strict_utf8_string_loc | |
1853 | ||
1854 | Like C<L</is_c9strict_utf8_string>> but stores the location of the failure (in | |
1855 | the case of "utf8ness failure") or the location C<s>+C<len> (in the case of | |
1856 | "utf8ness success") in the C<ep> pointer. | |
1857 | ||
1858 | See also C<L</is_c9strict_utf8_string_loclen>>. | |
1859 | ||
1860 | =cut | |
1861 | */ | |
1862 | ||
1863 | #define is_c9strict_utf8_string_loc(s, len, ep) \ | |
1864 | is_c9strict_utf8_string_loclen(s, len, ep, 0) | |
1865 | ||
1866 | /* | |
1867 | ||
1868 | =for apidoc is_c9strict_utf8_string_loclen | |
1869 | ||
1870 | Like C<L</is_c9strict_utf8_string>> but stores the location of the failure (in | |
1871 | the case of "utf8ness failure") or the location C<s>+C<len> (in the case of | |
1872 | "utf8ness success") in the C<ep> pointer, and the number of UTF-8 encoded | |
1873 | characters in the C<el> pointer. | |
1874 | ||
1875 | See also C<L</is_c9strict_utf8_string_loc>>. | |
1876 | ||
1877 | =cut | |
1878 | */ | |
1879 | ||
1880 | PERL_STATIC_INLINE bool | |
c9182d9c | 1881 | Perl_is_c9strict_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el) |
9f2abfde | 1882 | { |
33756530 | 1883 | const U8 * first_variant; |
9f2abfde KW |
1884 | |
1885 | PERL_ARGS_ASSERT_IS_C9STRICT_UTF8_STRING_LOCLEN; | |
1886 | ||
33756530 KW |
1887 | if (len == 0) { |
1888 | len = strlen((const char *) s); | |
1889 | } | |
1890 | ||
1891 | if (is_utf8_invariant_string_loc(s, len, &first_variant)) { | |
1892 | if (el) | |
1893 | *el = len; | |
1894 | ||
1895 | if (ep) { | |
1896 | *ep = s + len; | |
1897 | } | |
1898 | ||
1899 | return TRUE; | |
1900 | } | |
1901 | ||
1902 | { | |
1903 | const U8* const send = s + len; | |
1904 | const U8* x = first_variant; | |
1905 | STRLEN outlen = first_variant - s; | |
1906 | ||
a0d7f935 KW |
1907 | while (x < send) { |
1908 | const STRLEN cur_len = isC9_STRICT_UTF8_CHAR(x, send); | |
1909 | if (UNLIKELY(! cur_len)) { | |
1910 | break; | |
1911 | } | |
1912 | x += cur_len; | |
1913 | outlen++; | |
9f2abfde | 1914 | } |
9f2abfde | 1915 | |
a0d7f935 KW |
1916 | if (el) |
1917 | *el = outlen; | |
9f2abfde | 1918 | |
a0d7f935 KW |
1919 | if (ep) { |
1920 | *ep = x; | |
1921 | } | |
9f2abfde | 1922 | |
a0d7f935 | 1923 | return (x == send); |
33756530 | 1924 | } |
9f2abfde KW |
1925 | } |
1926 | ||
1927 | /* | |
1928 | ||
1929 | =for apidoc is_utf8_string_loc_flags | |
1930 | ||
1931 | Like C<L</is_utf8_string_flags>> but stores the location of the failure (in the | |
1932 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of | |
1933 | "utf8ness success") in the C<ep> pointer. | |
1934 | ||
1935 | See also C<L</is_utf8_string_loclen_flags>>. | |
1936 | ||
1937 | =cut | |
1938 | */ | |
1939 | ||
1940 | #define is_utf8_string_loc_flags(s, len, ep, flags) \ | |
1941 | is_utf8_string_loclen_flags(s, len, ep, 0, flags) | |
1942 | ||
1943 | ||
1944 | /* The above 3 actual functions could have been moved into the more general one | |
1945 | * just below, and made #defines that call it with the right 'flags'. They are | |
1946 | * currently kept separate to increase their chances of getting inlined */ | |
1947 | ||
1948 | /* | |
1949 | ||
1950 | =for apidoc is_utf8_string_loclen_flags | |
1951 | ||
1952 | Like C<L</is_utf8_string_flags>> but stores the location of the failure (in the | |
1953 | case of "utf8ness failure") or the location C<s>+C<len> (in the case of | |
1954 | "utf8ness success") in the C<ep> pointer, and the number of UTF-8 | |
1955 | encoded characters in the C<el> pointer. | |
1956 | ||
1957 | See also C<L</is_utf8_string_loc_flags>>. | |
1958 | ||
1959 | =cut | |
1960 | */ | |
1961 | ||
1962 | PERL_STATIC_INLINE bool | |
c9182d9c | 1963 | Perl_is_utf8_string_loclen_flags(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el, const U32 flags) |
9f2abfde | 1964 | { |
33756530 | 1965 | const U8 * first_variant; |
9f2abfde KW |
1966 | |
1967 | PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN_FLAGS; | |
1968 | assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE | |
d044b7a7 | 1969 | |UTF8_DISALLOW_PERL_EXTENDED))); |
9f2abfde | 1970 | |
f60f61fd | 1971 | if (len == 0) { |
a0d7f935 | 1972 | len = strlen((const char *) s); |
f60f61fd KW |
1973 | } |
1974 | ||
9f2abfde KW |
1975 | if (flags == 0) { |
1976 | return is_utf8_string_loclen(s, len, ep, el); | |
1977 | } | |
1978 | ||
d044b7a7 | 1979 | if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED) |
9f2abfde KW |
1980 | == UTF8_DISALLOW_ILLEGAL_INTERCHANGE) |
1981 | { | |
1982 | return is_strict_utf8_string_loclen(s, len, ep, el); | |
1983 | } | |
1984 | ||
d044b7a7 | 1985 | if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED) |
9f2abfde KW |
1986 | == UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE) |
1987 | { | |
1988 | return is_c9strict_utf8_string_loclen(s, len, ep, el); | |
1989 | } | |
1990 | ||
33756530 KW |
1991 | if (is_utf8_invariant_string_loc(s, len, &first_variant)) { |
1992 | if (el) | |
1993 | *el = len; | |
1994 | ||
1995 | if (ep) { | |
1996 | *ep = s + len; | |
1997 | } | |
1998 | ||
1999 | return TRUE; | |
2000 | } | |
2001 | ||
2002 | { | |
2003 | const U8* send = s + len; | |
2004 | const U8* x = first_variant; | |
2005 | STRLEN outlen = first_variant - s; | |
2006 | ||
a0d7f935 KW |
2007 | while (x < send) { |
2008 | const STRLEN cur_len = isUTF8_CHAR_flags(x, send, flags); | |
2009 | if (UNLIKELY(! cur_len)) { | |
2010 | break; | |
2011 | } | |
2012 | x += cur_len; | |
2013 | outlen++; | |
9f2abfde | 2014 | } |
9f2abfde | 2015 | |
a0d7f935 KW |
2016 | if (el) |
2017 | *el = outlen; | |
9f2abfde | 2018 | |
a0d7f935 KW |
2019 | if (ep) { |
2020 | *ep = x; | |
2021 | } | |
9f2abfde | 2022 | |
a0d7f935 | 2023 | return (x == send); |
33756530 | 2024 | } |
9f2abfde KW |
2025 | } |
2026 | ||
2027 | /* | |
7c93d8f0 KW |
2028 | =for apidoc utf8_distance |
2029 | ||
2030 | Returns the number of UTF-8 characters between the UTF-8 pointers C<a> | |
2031 | and C<b>. | |
2032 | ||
2033 | WARNING: use only if you *know* that the pointers point inside the | |
2034 | same UTF-8 buffer. | |
2035 | ||
2036 | =cut | |
2037 | */ | |
2038 | ||
2039 | PERL_STATIC_INLINE IV | |
2040 | Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b) | |
2041 | { | |
2042 | PERL_ARGS_ASSERT_UTF8_DISTANCE; | |
2043 | ||
2044 | return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a); | |
2045 | } | |
2046 | ||
2047 | /* | |
2048 | =for apidoc utf8_hop | |
2049 | ||
2050 | Return the UTF-8 pointer C<s> displaced by C<off> characters, either | |
2051 | forward or backward. | |
2052 | ||
2053 | WARNING: do not use the following unless you *know* C<off> is within | |
2054 | the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned | |
2055 | on the first byte of character or just after the last byte of a character. | |
2056 | ||
2057 | =cut | |
2058 | */ | |
2059 | ||
2060 | PERL_STATIC_INLINE U8 * | |
2061 | Perl_utf8_hop(const U8 *s, SSize_t off) | |
2062 | { | |
2063 | PERL_ARGS_ASSERT_UTF8_HOP; | |
2064 | ||
2065 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g | |
2066 | * the bitops (especially ~) can create illegal UTF-8. | |
2067 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
2068 | ||
2069 | if (off >= 0) { | |
1604cfb0 MS |
2070 | while (off--) |
2071 | s += UTF8SKIP(s); | |
7c93d8f0 KW |
2072 | } |
2073 | else { | |
1604cfb0 MS |
2074 | while (off++) { |
2075 | s--; | |
2076 | while (UTF8_IS_CONTINUATION(*s)) | |
2077 | s--; | |
2078 | } | |
7c93d8f0 | 2079 | } |
e099ea69 | 2080 | GCC_DIAG_IGNORE(-Wcast-qual) |
7c93d8f0 | 2081 | return (U8 *)s; |
e099ea69 | 2082 | GCC_DIAG_RESTORE |
7c93d8f0 KW |
2083 | } |
2084 | ||
4dab108f | 2085 | /* |
65df57a8 TC |
2086 | =for apidoc utf8_hop_forward |
2087 | ||
2088 | Return the UTF-8 pointer C<s> displaced by up to C<off> characters, | |
2089 | forward. | |
2090 | ||
2091 | C<off> must be non-negative. | |
2092 | ||
2093 | C<s> must be before or equal to C<end>. | |
2094 | ||
2095 | When moving forward it will not move beyond C<end>. | |
2096 | ||
2097 | Will not exceed this limit even if the string is not valid "UTF-8". | |
2098 | ||
2099 | =cut | |
2100 | */ | |
2101 | ||
2102 | PERL_STATIC_INLINE U8 * | |
2103 | Perl_utf8_hop_forward(const U8 *s, SSize_t off, const U8 *end) | |
2104 | { | |
2105 | PERL_ARGS_ASSERT_UTF8_HOP_FORWARD; | |
2106 | ||
2107 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g | |
2108 | * the bitops (especially ~) can create illegal UTF-8. | |
2109 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
2110 | ||
2111 | assert(s <= end); | |
2112 | assert(off >= 0); | |
2113 | ||
2114 | while (off--) { | |
2115 | STRLEN skip = UTF8SKIP(s); | |
de979548 | 2116 | if ((STRLEN)(end - s) <= skip) { |
e099ea69 | 2117 | GCC_DIAG_IGNORE(-Wcast-qual) |
65df57a8 | 2118 | return (U8 *)end; |
e099ea69 | 2119 | GCC_DIAG_RESTORE |
de979548 | 2120 | } |
65df57a8 TC |
2121 | s += skip; |
2122 | } | |
2123 | ||
e099ea69 | 2124 | GCC_DIAG_IGNORE(-Wcast-qual) |
65df57a8 | 2125 | return (U8 *)s; |
e099ea69 | 2126 | GCC_DIAG_RESTORE |
65df57a8 TC |
2127 | } |
2128 | ||
2129 | /* | |
2130 | =for apidoc utf8_hop_back | |
2131 | ||
2132 | Return the UTF-8 pointer C<s> displaced by up to C<off> characters, | |
2133 | backward. | |
2134 | ||
2135 | C<off> must be non-positive. | |
2136 | ||
2137 | C<s> must be after or equal to C<start>. | |
2138 | ||
2139 | When moving backward it will not move before C<start>. | |
2140 | ||
2141 | Will not exceed this limit even if the string is not valid "UTF-8". | |
2142 | ||
2143 | =cut | |
2144 | */ | |
2145 | ||
2146 | PERL_STATIC_INLINE U8 * | |
2147 | Perl_utf8_hop_back(const U8 *s, SSize_t off, const U8 *start) | |
2148 | { | |
2149 | PERL_ARGS_ASSERT_UTF8_HOP_BACK; | |
2150 | ||
2151 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g | |
2152 | * the bitops (especially ~) can create illegal UTF-8. | |
2153 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
2154 | ||
2155 | assert(start <= s); | |
2156 | assert(off <= 0); | |
2157 | ||
2158 | while (off++ && s > start) { | |
e7185695 | 2159 | do { |
65df57a8 | 2160 | s--; |
e7185695 | 2161 | } while (UTF8_IS_CONTINUATION(*s) && s > start); |
65df57a8 | 2162 | } |
f6521f7c | 2163 | |
e099ea69 | 2164 | GCC_DIAG_IGNORE(-Wcast-qual) |
65df57a8 | 2165 | return (U8 *)s; |
e099ea69 | 2166 | GCC_DIAG_RESTORE |
65df57a8 TC |
2167 | } |
2168 | ||
2169 | /* | |
2170 | =for apidoc utf8_hop_safe | |
2171 | ||
2172 | Return the UTF-8 pointer C<s> displaced by up to C<off> characters, | |
2173 | either forward or backward. | |
2174 | ||
2175 | When moving backward it will not move before C<start>. | |
2176 | ||
2177 | When moving forward it will not move beyond C<end>. | |
2178 | ||
2179 | Will not exceed those limits even if the string is not valid "UTF-8". | |
2180 | ||
2181 | =cut | |
2182 | */ | |
2183 | ||
2184 | PERL_STATIC_INLINE U8 * | |
2185 | Perl_utf8_hop_safe(const U8 *s, SSize_t off, const U8 *start, const U8 *end) | |
2186 | { | |
2187 | PERL_ARGS_ASSERT_UTF8_HOP_SAFE; | |
2188 | ||
2189 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g | |
2190 | * the bitops (especially ~) can create illegal UTF-8. | |
2191 | * In other words: in Perl UTF-8 is not just for Unicode. */ | |
2192 | ||
2193 | assert(start <= s && s <= end); | |
2194 | ||
2195 | if (off >= 0) { | |
2196 | return utf8_hop_forward(s, off, end); | |
2197 | } | |
2198 | else { | |
2199 | return utf8_hop_back(s, off, start); | |
2200 | } | |
2201 | } | |
2202 | ||
2203 | /* | |
4dab108f | 2204 | |
247cc51e | 2205 | =for apidoc isUTF8_CHAR_flags |
22f363ff KW |
2206 | |
2207 | Evaluates to non-zero if the first few bytes of the string starting at C<s> and | |
2208 | looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl, | |
2209 | that represents some code point, subject to the restrictions given by C<flags>; | |
2210 | otherwise it evaluates to 0. If non-zero, the value gives how many bytes | |
2211 | starting at C<s> comprise the code point's representation. Any bytes remaining | |
2212 | before C<e>, but beyond the ones needed to form the first code point in C<s>, | |
2213 | are not examined. | |
2214 | ||
2215 | If C<flags> is 0, this gives the same results as C<L</isUTF8_CHAR>>; | |
2216 | if C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results | |
2217 | as C<L</isSTRICT_UTF8_CHAR>>; | |
2218 | and if C<flags> is C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives | |
2219 | the same results as C<L</isC9_STRICT_UTF8_CHAR>>. | |
2220 | Otherwise C<flags> may be any combination of the C<UTF8_DISALLOW_I<foo>> flags | |
2221 | understood by C<L</utf8n_to_uvchr>>, with the same meanings. | |
2222 | ||
2223 | The three alternative macros are for the most commonly needed validations; they | |
2224 | are likely to run somewhat faster than this more general one, as they can be | |
2225 | inlined into your code. | |
2226 | ||
2227 | Use L</is_utf8_string_flags>, L</is_utf8_string_loc_flags>, and | |
2228 | L</is_utf8_string_loclen_flags> to check entire strings. | |
2229 | ||
2230 | =cut | |
2231 | */ | |
2232 | ||
2233 | PERL_STATIC_INLINE STRLEN | |
2234 | Perl_isUTF8_CHAR_flags(const U8 * const s0, const U8 * const e, const U32 flags) | |
2235 | { | |
2236 | PERL_ARGS_ASSERT_ISUTF8_CHAR_FLAGS; | |
2237 | assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE | |
2238 | |UTF8_DISALLOW_PERL_EXTENDED))); | |
2239 | ||
2240 | PERL_IS_UTF8_CHAR_DFA(s0, e, PL_extended_utf8_dfa_tab, | |
2241 | goto check_success, | |
2242 | DFA_TEASE_APART_FF_, | |
2243 | DFA_RETURN_FAILURE_); | |
2244 | ||
2245 | check_success: | |
2246 | ||
1aa501c2 | 2247 | return is_utf8_char_helper_(s0, e, flags); |
22f363ff KW |
2248 | |
2249 | #ifdef HAS_EXTRA_LONG_UTF8 | |
2250 | ||
2251 | tease_apart_FF: | |
2252 | ||
2253 | /* In the case of PL_extended_utf8_dfa_tab, getting here means the input is | |
2254 | * either malformed, or was for the largest possible start byte, which | |
2255 | * indicates perl extended UTF-8, well above the Unicode maximum */ | |
2256 | if ( *s0 != I8_TO_NATIVE_UTF8(0xFF) | |
2257 | || (flags & (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_PERL_EXTENDED))) | |
2258 | { | |
2259 | return 0; | |
2260 | } | |
2261 | ||
2262 | /* Otherwise examine the sequence not inline */ | |
2263 | return is_utf8_FF_helper_(s0, e, | |
2264 | FALSE /* require full, not partial char */ | |
2265 | ); | |
2266 | #endif | |
2267 | ||
2268 | } | |
2269 | ||
2270 | /* | |
2271 | ||
4dab108f KW |
2272 | =for apidoc is_utf8_valid_partial_char |
2273 | ||
6cbb9248 KW |
2274 | Returns 0 if the sequence of bytes starting at C<s> and looking no further than |
2275 | S<C<e - 1>> is the UTF-8 encoding, as extended by Perl, for one or more code | |
2276 | points. Otherwise, it returns 1 if there exists at least one non-empty | |
2277 | sequence of bytes that when appended to sequence C<s>, starting at position | |
2278 | C<e> causes the entire sequence to be the well-formed UTF-8 of some code point; | |
2279 | otherwise returns 0. | |
2280 | ||
2281 | In other words this returns TRUE if C<s> points to a partial UTF-8-encoded code | |
2282 | point. | |
2283 | ||
2284 | This is useful when a fixed-length buffer is being tested for being well-formed | |
2285 | UTF-8, but the final few bytes in it don't comprise a full character; that is, | |
2286 | it is split somewhere in the middle of the final code point's UTF-8 | |
2287 | representation. (Presumably when the buffer is refreshed with the next chunk | |
2288 | of data, the new first bytes will complete the partial code point.) This | |
2289 | function is used to verify that the final bytes in the current buffer are in | |
2290 | fact the legal beginning of some code point, so that if they aren't, the | |
2291 | failure can be signalled without having to wait for the next read. | |
4dab108f KW |
2292 | |
2293 | =cut | |
2294 | */ | |
2717076a KW |
2295 | #define is_utf8_valid_partial_char(s, e) \ |
2296 | is_utf8_valid_partial_char_flags(s, e, 0) | |
f1c999a7 KW |
2297 | |
2298 | /* | |
2299 | ||
2300 | =for apidoc is_utf8_valid_partial_char_flags | |
2301 | ||
2302 | Like C<L</is_utf8_valid_partial_char>>, it returns a boolean giving whether | |
2303 | or not the input is a valid UTF-8 encoded partial character, but it takes an | |
2304 | extra parameter, C<flags>, which can further restrict which code points are | |
2305 | considered valid. | |
2306 | ||
2307 | If C<flags> is 0, this behaves identically to | |
2308 | C<L</is_utf8_valid_partial_char>>. Otherwise C<flags> can be any combination | |
2309 | of the C<UTF8_DISALLOW_I<foo>> flags accepted by C<L</utf8n_to_uvchr>>. If | |
2310 | there is any sequence of bytes that can complete the input partial character in | |
2311 | such a way that a non-prohibited character is formed, the function returns | |
2717076a KW |
2312 | TRUE; otherwise FALSE. Non character code points cannot be determined based on |
2313 | partial character input. But many of the other possible excluded types can be | |
f1c999a7 KW |
2314 | determined from just the first one or two bytes. |
2315 | ||
2316 | =cut | |
2317 | */ | |
2318 | ||
56e4cf64 | 2319 | PERL_STATIC_INLINE bool |
22afef87 | 2320 | Perl_is_utf8_valid_partial_char_flags(const U8 * const s0, const U8 * const e, const U32 flags) |
4dab108f | 2321 | { |
f1c999a7 | 2322 | PERL_ARGS_ASSERT_IS_UTF8_VALID_PARTIAL_CHAR_FLAGS; |
f1c999a7 | 2323 | assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE |
d044b7a7 | 2324 | |UTF8_DISALLOW_PERL_EXTENDED))); |
4dab108f | 2325 | |
22afef87 KW |
2326 | PERL_IS_UTF8_CHAR_DFA(s0, e, PL_extended_utf8_dfa_tab, |
2327 | DFA_RETURN_FAILURE_, | |
2328 | DFA_TEASE_APART_FF_, | |
2329 | NOOP); | |
2330 | ||
2331 | /* The NOOP above causes the DFA to drop down here iff the input was a | |
2332 | * partial character. flags=0 => can return TRUE immediately; otherwise we | |
2333 | * need to check (not inline) if the partial character is the beginning of | |
2334 | * a disallowed one */ | |
2335 | if (flags == 0) { | |
2336 | return TRUE; | |
2337 | } | |
2338 | ||
1aa501c2 | 2339 | return cBOOL(is_utf8_char_helper_(s0, e, flags)); |
22afef87 KW |
2340 | |
2341 | #ifdef HAS_EXTRA_LONG_UTF8 | |
2342 | ||
2343 | tease_apart_FF: | |
2344 | ||
2345 | /* Getting here means the input is either malformed, or, in the case of | |
2346 | * PL_extended_utf8_dfa_tab, was for the largest possible start byte. The | |
2347 | * latter case has to be extended UTF-8, so can fail immediately if that is | |
2348 | * forbidden */ | |
2349 | ||
2350 | if ( *s0 != I8_TO_NATIVE_UTF8(0xFF) | |
2351 | || (flags & (UTF8_DISALLOW_SUPER|UTF8_DISALLOW_PERL_EXTENDED))) | |
2352 | { | |
2353 | return 0; | |
4dab108f KW |
2354 | } |
2355 | ||
22afef87 KW |
2356 | return is_utf8_FF_helper_(s0, e, |
2357 | TRUE /* Require to be a partial character */ | |
2358 | ); | |
2359 | #endif | |
2360 | ||
4dab108f KW |
2361 | } |
2362 | ||
8bc127bf KW |
2363 | /* |
2364 | ||
2365 | =for apidoc is_utf8_fixed_width_buf_flags | |
2366 | ||
2367 | Returns TRUE if the fixed-width buffer starting at C<s> with length C<len> | |
2368 | is entirely valid UTF-8, subject to the restrictions given by C<flags>; | |
2369 | otherwise it returns FALSE. | |
2370 | ||
2371 | If C<flags> is 0, any well-formed UTF-8, as extended by Perl, is accepted | |
2372 | without restriction. If the final few bytes of the buffer do not form a | |
2373 | complete code point, this will return TRUE anyway, provided that | |
2374 | C<L</is_utf8_valid_partial_char_flags>> returns TRUE for them. | |
2375 | ||
2376 | If C<flags> in non-zero, it can be any combination of the | |
2377 | C<UTF8_DISALLOW_I<foo>> flags accepted by C<L</utf8n_to_uvchr>>, and with the | |
2378 | same meanings. | |
2379 | ||
2380 | This function differs from C<L</is_utf8_string_flags>> only in that the latter | |
2381 | returns FALSE if the final few bytes of the string don't form a complete code | |
2382 | point. | |
2383 | ||
2384 | =cut | |
2385 | */ | |
2386 | #define is_utf8_fixed_width_buf_flags(s, len, flags) \ | |
2387 | is_utf8_fixed_width_buf_loclen_flags(s, len, 0, 0, flags) | |
2388 | ||
2389 | /* | |
2390 | ||
2391 | =for apidoc is_utf8_fixed_width_buf_loc_flags | |
2392 | ||
2393 | Like C<L</is_utf8_fixed_width_buf_flags>> but stores the location of the | |
2394 | failure in the C<ep> pointer. If the function returns TRUE, C<*ep> will point | |
2395 | to the beginning of any partial character at the end of the buffer; if there is | |
2396 | no partial character C<*ep> will contain C<s>+C<len>. | |
2397 | ||
2398 | See also C<L</is_utf8_fixed_width_buf_loclen_flags>>. | |
2399 | ||
2400 | =cut | |
2401 | */ | |
2402 | ||
2403 | #define is_utf8_fixed_width_buf_loc_flags(s, len, loc, flags) \ | |
2404 | is_utf8_fixed_width_buf_loclen_flags(s, len, loc, 0, flags) | |
2405 | ||
2406 | /* | |
2407 | ||
2408 | =for apidoc is_utf8_fixed_width_buf_loclen_flags | |
2409 | ||
2410 | Like C<L</is_utf8_fixed_width_buf_loc_flags>> but stores the number of | |
2411 | complete, valid characters found in the C<el> pointer. | |
2412 | ||
2413 | =cut | |
2414 | */ | |
2415 | ||
2416 | PERL_STATIC_INLINE bool | |
c9182d9c | 2417 | Perl_is_utf8_fixed_width_buf_loclen_flags(const U8 * const s, |
33756530 | 2418 | STRLEN len, |
8bc127bf KW |
2419 | const U8 **ep, |
2420 | STRLEN *el, | |
2421 | const U32 flags) | |
2422 | { | |
2423 | const U8 * maybe_partial; | |
2424 | ||
2425 | PERL_ARGS_ASSERT_IS_UTF8_FIXED_WIDTH_BUF_LOCLEN_FLAGS; | |
2426 | ||
2427 | if (! ep) { | |
2428 | ep = &maybe_partial; | |
2429 | } | |
2430 | ||
2431 | /* If it's entirely valid, return that; otherwise see if the only error is | |
2432 | * that the final few bytes are for a partial character */ | |
2433 | return is_utf8_string_loclen_flags(s, len, ep, el, flags) | |
2434 | || is_utf8_valid_partial_char_flags(*ep, s + len, flags); | |
2435 | } | |
2436 | ||
e6a4ffc3 | 2437 | PERL_STATIC_INLINE UV |
c9182d9c | 2438 | Perl_utf8n_to_uvchr_msgs(const U8 *s, |
e6a4ffc3 KW |
2439 | STRLEN curlen, |
2440 | STRLEN *retlen, | |
2441 | const U32 flags, | |
2442 | U32 * errors, | |
2443 | AV ** msgs) | |
2444 | { | |
2445 | /* This is the inlined portion of utf8n_to_uvchr_msgs. It handles the | |
2446 | * simple cases, and, if necessary calls a helper function to deal with the | |
2447 | * more complex ones. Almost all well-formed non-problematic code points | |
2448 | * are considered simple, so that it's unlikely that the helper function | |
2449 | * will need to be called. | |
2450 | * | |
2451 | * This is an adaptation of the tables and algorithm given in | |
f6521f7c | 2452 | * https://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides |
e6a4ffc3 KW |
2453 | * comprehensive documentation of the original version. A copyright notice |
2454 | * for the original version is given at the beginning of this file. The | |
71525f77 | 2455 | * Perl adapation is documented at the definition of PL_strict_utf8_dfa_tab[]. |
e6a4ffc3 KW |
2456 | */ |
2457 | ||
2458 | const U8 * const s0 = s; | |
2459 | const U8 * send = s0 + curlen; | |
a4609251 KW |
2460 | UV type; |
2461 | UV uv; | |
e6a4ffc3 KW |
2462 | |
2463 | PERL_ARGS_ASSERT_UTF8N_TO_UVCHR_MSGS; | |
2464 | ||
2465 | /* This dfa is fast. If it accepts the input, it was for a well-formed, | |
2466 | * non-problematic code point, which can be returned immediately. | |
2467 | * Otherwise we call a helper function to figure out the more complicated | |
2468 | * cases. */ | |
2469 | ||
a4609251 | 2470 | /* No calls from core pass in an empty string; non-core need a check */ |
d1e771d8 KW |
2471 | #ifdef PERL_CORE |
2472 | assert(curlen > 0); | |
2473 | #else | |
2474 | if (curlen == 0) return _utf8n_to_uvchr_msgs_helper(s0, 0, retlen, | |
2475 | flags, errors, msgs); | |
2476 | #endif | |
e6a4ffc3 | 2477 | |
a4609251 | 2478 | type = PL_strict_utf8_dfa_tab[*s]; |
e6a4ffc3 | 2479 | |
a4609251 KW |
2480 | /* The table is structured so that 'type' is 0 iff the input byte is |
2481 | * represented identically regardless of the UTF-8ness of the string */ | |
2482 | if (type == 0) { /* UTF-8 invariants are returned unchanged */ | |
2483 | uv = *s; | |
2484 | } | |
2485 | else { | |
2486 | UV state = PL_strict_utf8_dfa_tab[256 + type]; | |
2487 | uv = (0xff >> type) & NATIVE_UTF8_TO_I8(*s); | |
e6a4ffc3 | 2488 | |
a4609251 KW |
2489 | while (++s < send) { |
2490 | type = PL_strict_utf8_dfa_tab[*s]; | |
2491 | state = PL_strict_utf8_dfa_tab[256 + state + type]; | |
2492 | ||
2493 | uv = UTF8_ACCUMULATE(uv, *s); | |
2494 | ||
2495 | if (state == 0) { | |
2496 | goto success; | |
2497 | } | |
2498 | ||
2499 | if (UNLIKELY(state == 1)) { | |
2500 | break; | |
2501 | } | |
e6a4ffc3 KW |
2502 | } |
2503 | ||
a4609251 KW |
2504 | /* Here is potentially problematic. Use the full mechanism */ |
2505 | return _utf8n_to_uvchr_msgs_helper(s0, curlen, retlen, flags, | |
2506 | errors, msgs); | |
2507 | } | |
2508 | ||
2509 | success: | |
2510 | if (retlen) { | |
2511 | *retlen = s - s0 + 1; | |
2512 | } | |
2513 | if (errors) { | |
2514 | *errors = 0; | |
2515 | } | |
2516 | if (msgs) { | |
2517 | *msgs = NULL; | |
e6a4ffc3 KW |
2518 | } |
2519 | ||
a4609251 | 2520 | return UNI_TO_NATIVE(uv); |
e6a4ffc3 KW |
2521 | } |
2522 | ||
82651abe | 2523 | PERL_STATIC_INLINE UV |
9a9a6c98 | 2524 | Perl_utf8_to_uvchr_buf_helper(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen) |
82651abe | 2525 | { |
9a9a6c98 | 2526 | PERL_ARGS_ASSERT_UTF8_TO_UVCHR_BUF_HELPER; |
82651abe KW |
2527 | |
2528 | assert(s < send); | |
2529 | ||
2530 | if (! ckWARN_d(WARN_UTF8)) { | |
3eaa7592 KW |
2531 | |
2532 | /* EMPTY is not really allowed, and asserts on debugging builds. But | |
2533 | * on non-debugging we have to deal with it, and this causes it to | |
2534 | * return the REPLACEMENT CHARACTER, as the documentation indicates */ | |
82651abe | 2535 | return utf8n_to_uvchr(s, send - s, retlen, |
3eaa7592 | 2536 | (UTF8_ALLOW_ANY | UTF8_ALLOW_EMPTY)); |
82651abe KW |
2537 | } |
2538 | else { | |
2539 | UV ret = utf8n_to_uvchr(s, send - s, retlen, 0); | |
286a1bfd | 2540 | if (retlen && ret == 0 && (send <= s || *s != '\0')) { |
82651abe KW |
2541 | *retlen = (STRLEN) -1; |
2542 | } | |
2543 | ||
2544 | return ret; | |
2545 | } | |
2546 | } | |
2547 | ||
c8028aa6 TC |
2548 | /* ------------------------------- perl.h ----------------------------- */ |
2549 | ||
2550 | /* | |
3f620621 | 2551 | =for apidoc_section $utility |
dcccc8ff | 2552 | |
44170c9a | 2553 | =for apidoc is_safe_syscall |
c8028aa6 | 2554 | |
1a0efc9a KW |
2555 | Test that the given C<pv> (with length C<len>) doesn't contain any internal |
2556 | C<NUL> characters. | |
2557 | If it does, set C<errno> to C<ENOENT>, optionally warn using the C<syscalls> | |
2558 | category, and return FALSE. | |
c8028aa6 TC |
2559 | |
2560 | Return TRUE if the name is safe. | |
2561 | ||
1a0efc9a KW |
2562 | C<what> and C<op_name> are used in any warning. |
2563 | ||
796b6530 | 2564 | Used by the C<IS_SAFE_SYSCALL()> macro. |
c8028aa6 TC |
2565 | |
2566 | =cut | |
2567 | */ | |
2568 | ||
2569 | PERL_STATIC_INLINE bool | |
ffd62fc2 KW |
2570 | Perl_is_safe_syscall(pTHX_ const char *pv, STRLEN len, const char *what, const char *op_name) |
2571 | { | |
c8028aa6 TC |
2572 | /* While the Windows CE API provides only UCS-16 (or UTF-16) APIs |
2573 | * perl itself uses xce*() functions which accept 8-bit strings. | |
2574 | */ | |
2575 | ||
2576 | PERL_ARGS_ASSERT_IS_SAFE_SYSCALL; | |
2577 | ||
6c4650b3 | 2578 | if (len > 1) { |
c8028aa6 | 2579 | char *null_at; |
41188aa0 | 2580 | if (UNLIKELY((null_at = (char *)memchr(pv, 0, len-1)) != NULL)) { |
c8028aa6 | 2581 | SETERRNO(ENOENT, LIB_INVARG); |
1d505182 | 2582 | Perl_ck_warner(aTHX_ packWARN(WARN_SYSCALLS), |
c8028aa6 | 2583 | "Invalid \\0 character in %s for %s: %s\\0%s", |
41188aa0 | 2584 | what, op_name, pv, null_at+1); |
c8028aa6 TC |
2585 | return FALSE; |
2586 | } | |
2587 | } | |
2588 | ||
2589 | return TRUE; | |
2590 | } | |
2591 | ||
2592 | /* | |
7cb3f959 TC |
2593 | |
2594 | Return true if the supplied filename has a newline character | |
fa6c7d00 | 2595 | immediately before the first (hopefully only) NUL. |
7cb3f959 TC |
2596 | |
2597 | My original look at this incorrectly used the len from SvPV(), but | |
2598 | that's incorrect, since we allow for a NUL in pv[len-1]. | |
2599 | ||
2600 | So instead, strlen() and work from there. | |
2601 | ||
2602 | This allow for the user reading a filename, forgetting to chomp it, | |
2603 | then calling: | |
2604 | ||
2605 | open my $foo, "$file\0"; | |
2606 | ||
2607 | */ | |
2608 | ||
2609 | #ifdef PERL_CORE | |
2610 | ||
2611 | PERL_STATIC_INLINE bool | |
ffd62fc2 KW |
2612 | S_should_warn_nl(const char *pv) |
2613 | { | |
7cb3f959 TC |
2614 | STRLEN len; |
2615 | ||
2616 | PERL_ARGS_ASSERT_SHOULD_WARN_NL; | |
2617 | ||
2618 | len = strlen(pv); | |
2619 | ||
2620 | return len > 0 && pv[len-1] == '\n'; | |
2621 | } | |
2622 | ||
2623 | #endif | |
2624 | ||
3a019afd KW |
2625 | #if defined(PERL_IN_PP_C) || defined(PERL_IN_PP_HOT_C) |
2626 | ||
2627 | PERL_STATIC_INLINE bool | |
2628 | S_lossless_NV_to_IV(const NV nv, IV *ivp) | |
2629 | { | |
2630 | /* This function determines if the input NV 'nv' may be converted without | |
2631 | * loss of data to an IV. If not, it returns FALSE taking no other action. | |
2632 | * But if it is possible, it does the conversion, returning TRUE, and | |
2633 | * storing the converted result in '*ivp' */ | |
2634 | ||
2635 | PERL_ARGS_ASSERT_LOSSLESS_NV_TO_IV; | |
2636 | ||
cd304e76 DM |
2637 | # if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan) |
2638 | /* Normally any comparison with a NaN returns false; if we can't rely | |
2639 | * on that behaviour, check explicitly */ | |
3a019afd KW |
2640 | if (UNLIKELY(Perl_isnan(nv))) { |
2641 | return FALSE; | |
2642 | } | |
3a019afd KW |
2643 | # endif |
2644 | ||
cd304e76 DM |
2645 | /* Written this way so that with an always-false NaN comparison we |
2646 | * return false */ | |
ef0a8475 | 2647 | if (!(LIKELY(nv >= (NV) IV_MIN) && LIKELY(nv < IV_MAX_P1))) { |
3a019afd KW |
2648 | return FALSE; |
2649 | } | |
2650 | ||
2651 | if ((IV) nv != nv) { | |
2652 | return FALSE; | |
2653 | } | |
2654 | ||
2655 | *ivp = (IV) nv; | |
2656 | return TRUE; | |
2657 | } | |
2658 | ||
2659 | #endif | |
2660 | ||
81d52ecd JH |
2661 | /* ------------------ pp.c, regcomp.c, toke.c, universal.c ------------ */ |
2662 | ||
94b0cb42 KW |
2663 | #if defined(PERL_IN_PP_C) || defined(PERL_IN_REGCOMP_C) || defined(PERL_IN_TOKE_C) || defined(PERL_IN_UNIVERSAL_C) |
2664 | ||
81d52ecd JH |
2665 | #define MAX_CHARSET_NAME_LENGTH 2 |
2666 | ||
2667 | PERL_STATIC_INLINE const char * | |
94b0cb42 | 2668 | S_get_regex_charset_name(const U32 flags, STRLEN* const lenp) |
81d52ecd | 2669 | { |
94b0cb42 KW |
2670 | PERL_ARGS_ASSERT_GET_REGEX_CHARSET_NAME; |
2671 | ||
81d52ecd JH |
2672 | /* Returns a string that corresponds to the name of the regex character set |
2673 | * given by 'flags', and *lenp is set the length of that string, which | |
2674 | * cannot exceed MAX_CHARSET_NAME_LENGTH characters */ | |
2675 | ||
2676 | *lenp = 1; | |
2677 | switch (get_regex_charset(flags)) { | |
2678 | case REGEX_DEPENDS_CHARSET: return DEPENDS_PAT_MODS; | |
2679 | case REGEX_LOCALE_CHARSET: return LOCALE_PAT_MODS; | |
2680 | case REGEX_UNICODE_CHARSET: return UNICODE_PAT_MODS; | |
1604cfb0 MS |
2681 | case REGEX_ASCII_RESTRICTED_CHARSET: return ASCII_RESTRICT_PAT_MODS; |
2682 | case REGEX_ASCII_MORE_RESTRICTED_CHARSET: | |
2683 | *lenp = 2; | |
2684 | return ASCII_MORE_RESTRICT_PAT_MODS; | |
81d52ecd JH |
2685 | } |
2686 | /* The NOT_REACHED; hides an assert() which has a rather complex | |
2687 | * definition in perl.h. */ | |
2688 | NOT_REACHED; /* NOTREACHED */ | |
2689 | return "?"; /* Unknown */ | |
2690 | } | |
2691 | ||
94b0cb42 KW |
2692 | #endif |
2693 | ||
7cb3f959 | 2694 | /* |
ed382232 TC |
2695 | |
2696 | Return false if any get magic is on the SV other than taint magic. | |
2697 | ||
2698 | */ | |
2699 | ||
2700 | PERL_STATIC_INLINE bool | |
ffd62fc2 KW |
2701 | Perl_sv_only_taint_gmagic(SV *sv) |
2702 | { | |
ed382232 TC |
2703 | MAGIC *mg = SvMAGIC(sv); |
2704 | ||
2705 | PERL_ARGS_ASSERT_SV_ONLY_TAINT_GMAGIC; | |
2706 | ||
2707 | while (mg) { | |
2708 | if (mg->mg_type != PERL_MAGIC_taint | |
2709 | && !(mg->mg_flags & MGf_GSKIP) | |
2710 | && mg->mg_virtual->svt_get) { | |
2711 | return FALSE; | |
2712 | } | |
2713 | mg = mg->mg_moremagic; | |
2714 | } | |
2715 | ||
2716 | return TRUE; | |
2717 | } | |
2718 | ||
ed8ff0f3 DM |
2719 | /* ------------------ cop.h ------------------------------------------- */ |
2720 | ||
5b6f7443 DM |
2721 | /* implement GIMME_V() macro */ |
2722 | ||
2723 | PERL_STATIC_INLINE U8 | |
2724 | Perl_gimme_V(pTHX) | |
2725 | { | |
2726 | I32 cxix; | |
2727 | U8 gimme = (PL_op->op_flags & OPf_WANT); | |
2728 | ||
2729 | if (gimme) | |
2730 | return gimme; | |
2731 | cxix = PL_curstackinfo->si_cxsubix; | |
2732 | if (cxix < 0) | |
390fe0c0 | 2733 | return PL_curstackinfo->si_type == PERLSI_SORT ? G_SCALAR: G_VOID; |
5b6f7443 DM |
2734 | assert(cxstack[cxix].blk_gimme & G_WANT); |
2735 | return (cxstack[cxix].blk_gimme & G_WANT); | |
2736 | } | |
2737 | ||
ed8ff0f3 DM |
2738 | |
2739 | /* Enter a block. Push a new base context and return its address. */ | |
2740 | ||
2741 | PERL_STATIC_INLINE PERL_CONTEXT * | |
c9182d9c | 2742 | Perl_cx_pushblock(pTHX_ U8 type, U8 gimme, SV** sp, I32 saveix) |
ed8ff0f3 DM |
2743 | { |
2744 | PERL_CONTEXT * cx; | |
2745 | ||
2746 | PERL_ARGS_ASSERT_CX_PUSHBLOCK; | |
2747 | ||
2748 | CXINC; | |
2749 | cx = CX_CUR(); | |
2750 | cx->cx_type = type; | |
2751 | cx->blk_gimme = gimme; | |
2752 | cx->blk_oldsaveix = saveix; | |
4caf7d8c | 2753 | cx->blk_oldsp = (I32)(sp - PL_stack_base); |
ed8ff0f3 | 2754 | cx->blk_oldcop = PL_curcop; |
4caf7d8c | 2755 | cx->blk_oldmarksp = (I32)(PL_markstack_ptr - PL_markstack); |
ed8ff0f3 DM |
2756 | cx->blk_oldscopesp = PL_scopestack_ix; |
2757 | cx->blk_oldpm = PL_curpm; | |
ce8bb8d8 | 2758 | cx->blk_old_tmpsfloor = PL_tmps_floor; |
ed8ff0f3 DM |
2759 | |
2760 | PL_tmps_floor = PL_tmps_ix; | |
2761 | CX_DEBUG(cx, "PUSH"); | |
2762 | return cx; | |
2763 | } | |
2764 | ||
2765 | ||
2766 | /* Exit a block (RETURN and LAST). */ | |
2767 | ||
2768 | PERL_STATIC_INLINE void | |
c9182d9c | 2769 | Perl_cx_popblock(pTHX_ PERL_CONTEXT *cx) |
ed8ff0f3 DM |
2770 | { |
2771 | PERL_ARGS_ASSERT_CX_POPBLOCK; | |
2772 | ||
2773 | CX_DEBUG(cx, "POP"); | |
2774 | /* these 3 are common to cx_popblock and cx_topblock */ | |
2775 | PL_markstack_ptr = PL_markstack + cx->blk_oldmarksp; | |
2776 | PL_scopestack_ix = cx->blk_oldscopesp; | |
2777 | PL_curpm = cx->blk_oldpm; | |
2778 | ||
2779 | /* LEAVE_SCOPE() should have made this true. /(?{})/ cheats | |
2780 | * and leaves a CX entry lying around for repeated use, so | |
2781 | * skip for multicall */ \ | |
2782 | assert( (CxTYPE(cx) == CXt_SUB && CxMULTICALL(cx)) | |
2783 | || PL_savestack_ix == cx->blk_oldsaveix); | |
2784 | PL_curcop = cx->blk_oldcop; | |
ce8bb8d8 | 2785 | PL_tmps_floor = cx->blk_old_tmpsfloor; |
ed8ff0f3 DM |
2786 | } |
2787 | ||
2788 | /* Continue a block elsewhere (e.g. NEXT, REDO, GOTO). | |
2789 | * Whereas cx_popblock() restores the state to the point just before | |
2790 | * cx_pushblock() was called, cx_topblock() restores it to the point just | |
2791 | * *after* cx_pushblock() was called. */ | |
2792 | ||
2793 | PERL_STATIC_INLINE void | |
c9182d9c | 2794 | Perl_cx_topblock(pTHX_ PERL_CONTEXT *cx) |
ed8ff0f3 DM |
2795 | { |
2796 | PERL_ARGS_ASSERT_CX_TOPBLOCK; | |
2797 | ||
2798 | CX_DEBUG(cx, "TOP"); | |
2799 | /* these 3 are common to cx_popblock and cx_topblock */ | |
2800 | PL_markstack_ptr = PL_markstack + cx->blk_oldmarksp; | |
2801 | PL_scopestack_ix = cx->blk_oldscopesp; | |
2802 | PL_curpm = cx->blk_oldpm; | |
2803 | ||
2804 | PL_stack_sp = PL_stack_base + cx->blk_oldsp; | |
2805 | } | |
2806 | ||
2807 | ||
a73d8813 | 2808 | PERL_STATIC_INLINE void |
c9182d9c | 2809 | Perl_cx_pushsub(pTHX_ PERL_CONTEXT *cx, CV *cv, OP *retop, bool hasargs) |
a73d8813 DM |
2810 | { |
2811 | U8 phlags = CX_PUSHSUB_GET_LVALUE_MASK(Perl_was_lvalue_sub); | |
2812 | ||
2813 | PERL_ARGS_ASSERT_CX_PUSHSUB; | |
2814 | ||
3f6bd23a | 2815 | PERL_DTRACE_PROBE_ENTRY(cv); |
5b6f7443 DM |
2816 | cx->blk_sub.old_cxsubix = PL_curstackinfo->si_cxsubix; |
2817 | PL_curstackinfo->si_cxsubix = cx - PL_curstackinfo->si_cxstack; | |
a73d8813 DM |
2818 | cx->blk_sub.cv = cv; |
2819 | cx->blk_sub.olddepth = CvDEPTH(cv); | |
2820 | cx->blk_sub.prevcomppad = PL_comppad; | |
2821 | cx->cx_type |= (hasargs) ? CXp_HASARGS : 0; | |
2822 | cx->blk_sub.retop = retop; | |
2823 | SvREFCNT_inc_simple_void_NN(cv); | |
2824 | cx->blk_u16 = PL_op->op_private & (phlags|OPpDEREF); | |
2825 | } | |
2826 | ||
2827 | ||
2828 | /* subsets of cx_popsub() */ | |
2829 | ||
2830 | PERL_STATIC_INLINE void | |
c9182d9c | 2831 | Perl_cx_popsub_common(pTHX_ PERL_CONTEXT *cx) |
a73d8813 DM |
2832 | { |
2833 | CV *cv; | |
2834 | ||
2835 | PERL_ARGS_ASSERT_CX_POPSUB_COMMON; | |
2836 | assert(CxTYPE(cx) == CXt_SUB); | |
2837 | ||
2838 | PL_comppad = cx->blk_sub.prevcomppad; | |
2839 | PL_curpad = LIKELY(PL_comppad) ? AvARRAY(PL_comppad) : NULL; | |
2840 | cv = cx->blk_sub.cv; | |
2841 | CvDEPTH(cv) = cx->blk_sub.olddepth; | |
2842 | cx->blk_sub.cv = NULL; | |
2843 | SvREFCNT_dec(cv); | |
5b6f7443 | 2844 | PL_curstackinfo->si_cxsubix = cx->blk_sub.old_cxsubix; |
a73d8813 DM |
2845 | } |
2846 | ||
2847 | ||
2848 | /* handle the @_ part of leaving a sub */ | |
2849 | ||
2850 | PERL_STATIC_INLINE void | |
c9182d9c | 2851 | Perl_cx_popsub_args(pTHX_ PERL_CONTEXT *cx) |
a73d8813 DM |
2852 | { |
2853 | AV *av; | |
2854 | ||
2855 | PERL_ARGS_ASSERT_CX_POPSUB_ARGS; | |
2856 | assert(CxTYPE(cx) == CXt_SUB); | |
2857 | assert(AvARRAY(MUTABLE_AV( | |
2858 | PadlistARRAY(CvPADLIST(cx->blk_sub.cv))[ | |
2859 | CvDEPTH(cx->blk_sub.cv)])) == PL_curpad); | |
2860 | ||
2861 | CX_POP_SAVEARRAY(cx); | |
2862 | av = MUTABLE_AV(PAD_SVl(0)); | |
2863 | if (UNLIKELY(AvREAL(av))) | |
2864 | /* abandon @_ if it got reified */ | |
2865 | clear_defarray(av, 0); | |
2866 | else { | |
2867 | CLEAR_ARGARRAY(av); | |
2868 | } | |
2869 | } | |
2870 | ||
2871 | ||
2872 | PERL_STATIC_INLINE void | |
c9182d9c | 2873 | Perl_cx_popsub(pTHX_ PERL_CONTEXT *cx) |
a73d8813 DM |
2874 | { |
2875 | PERL_ARGS_ASSERT_CX_POPSUB; | |
2876 | assert(CxTYPE(cx) == CXt_SUB); | |
2877 | ||
3f6bd23a | 2878 | PERL_DTRACE_PROBE_RETURN(cx->blk_sub.cv); |
a73d8813 DM |
2879 | |
2880 | if (CxHASARGS(cx)) | |
2881 | cx_popsub_args(cx); | |
2882 | cx_popsub_common(cx); | |
2883 | } | |
2884 | ||
2885 | ||
6a7d52cc | 2886 | PERL_STATIC_INLINE void |
c9182d9c | 2887 | Perl_cx_pushformat(pTHX_ PERL_CONTEXT *cx, CV *cv, OP *retop, GV *gv) |
6a7d52cc DM |
2888 | { |
2889 | PERL_ARGS_ASSERT_CX_PUSHFORMAT; | |
2890 | ||
5b6f7443 DM |
2891 | cx->blk_format.old_cxsubix = PL_curstackinfo->si_cxsubix; |
2892 | PL_curstackinfo->si_cxsubix= cx - PL_curstackinfo->si_cxstack; | |
6a7d52cc DM |
2893 | cx->blk_format.cv = cv; |
2894 | cx->blk_format.retop = retop; | |
2895 | cx->blk_format.gv = gv; | |
2896 | cx->blk_format.dfoutgv = PL_defoutgv; | |
2897 | cx->blk_format.prevcomppad = PL_comppad; | |
2898 | cx->blk_u16 = 0; | |
2899 | ||
2900 | SvREFCNT_inc_simple_void_NN(cv); | |
2901 | CvDEPTH(cv)++; | |
2902 | SvREFCNT_inc_void(cx->blk_format.dfoutgv); | |
2903 | } | |
2904 | ||
2905 | ||
2906 | PERL_STATIC_INLINE void | |
c9182d9c | 2907 | Perl_cx_popformat(pTHX_ PERL_CONTEXT *cx) |
6a7d52cc DM |
2908 | { |
2909 | CV *cv; | |
2910 | GV *dfout; | |
2911 | ||
2912 | PERL_ARGS_ASSERT_CX_POPFORMAT; | |
2913 | assert(CxTYPE(cx) == CXt_FORMAT); | |
2914 | ||
2915 | dfout = cx->blk_format.dfoutgv; | |
2916 | setdefout(dfout); | |
2917 | cx->blk_format.dfoutgv = NULL; | |
2918 | SvREFCNT_dec_NN(dfout); | |
2919 | ||
2920 | PL_comppad = cx->blk_format.prevcomppad; | |
2921 | PL_curpad = LIKELY(PL_comppad) ? AvARRAY(PL_comppad) : NULL; | |
2922 | cv = cx->blk_format.cv; | |
2923 | cx->blk_format.cv = NULL; | |
2924 | --CvDEPTH(cv); | |
2925 | SvREFCNT_dec_NN(cv); | |
5b6f7443 | 2926 | PL_curstackinfo->si_cxsubix = cx->blk_format.old_cxsubix; |
6a7d52cc DM |
2927 | } |
2928 | ||
2929 | ||
13febba5 | 2930 | PERL_STATIC_INLINE void |
6b729d24 | 2931 | Perl_push_evalortry_common(pTHX_ PERL_CONTEXT *cx, OP *retop, SV *namesv) |
13febba5 | 2932 | { |
13febba5 DM |
2933 | cx->blk_eval.retop = retop; |
2934 | cx->blk_eval.old_namesv = namesv; | |
2935 | cx->blk_eval.old_eval_root = PL_eval_root; | |
2936 | cx->blk_eval.cur_text = PL_parser ? PL_parser->linestr : NULL; | |
2937 | cx->blk_eval.cv = NULL; /* later set by doeval_compile() */ | |
2938 | cx->blk_eval.cur_top_env = PL_top_env; | |
2939 | ||
4c57ced5 | 2940 | assert(!(PL_in_eval & ~ 0x3F)); |
13febba5 | 2941 | assert(!(PL_op->op_type & ~0x1FF)); |
4c57ced5 | 2942 | cx->blk_u16 = (PL_in_eval & 0x3F) | ((U16)PL_op->op_type << 7); |
13febba5 DM |
2943 | } |
2944 | ||
6b729d24 TC |
2945 | PERL_STATIC_INLINE void |
2946 | Perl_cx_pusheval(pTHX_ PERL_CONTEXT *cx, OP *retop, SV *namesv) | |
2947 | { | |
2948 | PERL_ARGS_ASSERT_CX_PUSHEVAL; | |
2949 | ||
2950 | Perl_push_evalortry_common(aTHX_ cx, retop, namesv); | |
2951 | ||
2952 | cx->blk_eval.old_cxsubix = PL_curstackinfo->si_cxsubix; | |
2953 | PL_curstackinfo->si_cxsubix = cx - PL_curstackinfo->si_cxstack; | |
2954 | } | |
2955 | ||
2956 | PERL_STATIC_INLINE void | |
2957 | Perl_cx_pushtry(pTHX_ PERL_CONTEXT *cx, OP *retop) | |
2958 | { | |
2959 | PERL_ARGS_ASSERT_CX_PUSHTRY; | |
2960 | ||
2961 | Perl_push_evalortry_common(aTHX_ cx, retop, NULL); | |
2962 | ||
2963 | /* Don't actually change it, just store the current value so it's restored | |
2964 | * by the common popeval */ | |
2965 | cx->blk_eval.old_cxsubix = PL_curstackinfo->si_cxsubix; | |
2966 | } | |
2967 | ||
13febba5 DM |
2968 | |
2969 | PERL_STATIC_INLINE void | |
c9182d9c | 2970 | Perl_cx_popeval(pTHX_ PERL_CONTEXT *cx) |
13febba5 DM |
2971 | { |
2972 | SV *sv; | |
2973 | ||
2974 | PERL_ARGS_ASSERT_CX_POPEVAL; | |
2975 | assert(CxTYPE(cx) == CXt_EVAL); | |
2976 | ||
2977 | PL_in_eval = CxOLD_IN_EVAL(cx); | |
4c57ced5 | 2978 | assert(!(PL_in_eval & 0xc0)); |
13febba5 DM |
2979 | PL_eval_root = cx->blk_eval.old_eval_root; |
2980 | sv = cx->blk_eval.cur_text; | |
4c57ced5 | 2981 | if (sv && CxEVAL_TXT_REFCNTED(cx)) { |
13febba5 DM |
2982 | cx->blk_eval.cur_text = NULL; |
2983 | SvREFCNT_dec_NN(sv); | |
2984 | } | |
2985 | ||
2986 | sv = cx->blk_eval.old_namesv; | |
2a1e0dfe DM |
2987 | if (sv) { |
2988 | cx->blk_eval.old_namesv = NULL; | |
2989 | SvREFCNT_dec_NN(sv); | |
2990 | } | |
5b6f7443 | 2991 | PL_curstackinfo->si_cxsubix = cx->blk_eval.old_cxsubix; |
13febba5 | 2992 | } |
6a7d52cc | 2993 | |
a73d8813 | 2994 | |
d1b6bf72 DM |
2995 | /* push a plain loop, i.e. |
2996 | * { block } | |
2997 | * while (cond) { block } | |
2998 | * for (init;cond;continue) { block } | |
2999 | * This loop can be last/redo'ed etc. | |
3000 | */ | |
3001 | ||
3002 | PERL_STATIC_INLINE void | |
c9182d9c | 3003 | Perl_cx_pushloop_plain(pTHX_ PERL_CONTEXT *cx) |
d1b6bf72 DM |
3004 | { |
3005 | PERL_ARGS_ASSERT_CX_PUSHLOOP_PLAIN; | |
3006 | cx->blk_loop.my_op = cLOOP; | |
3007 | } | |
3008 | ||
3009 | ||
3010 | /* push a true for loop, i.e. | |
3011 | * for var (list) { block } | |
3012 | */ | |
3013 | ||
3014 | PERL_STATIC_INLINE void | |
c9182d9c | 3015 | Perl_cx_pushloop_for(pTHX_ PERL_CONTEXT *cx, void *itervarp, SV* itersave) |
d1b6bf72 DM |
3016 | { |
3017 | PERL_ARGS_ASSERT_CX_PUSHLOOP_FOR; | |
3018 | ||
3019 | /* this one line is common with cx_pushloop_plain */ | |
3020 | cx->blk_loop.my_op = cLOOP; | |
3021 | ||
3022 | cx->blk_loop.itervar_u.svp = (SV**)itervarp; | |
3023 | cx->blk_loop.itersave = itersave; | |
3024 | #ifdef USE_ITHREADS | |
3025 | cx->blk_loop.oldcomppad = PL_comppad; | |
3026 | #endif | |
3027 | } | |
3028 | ||
3029 | ||
3030 | /* pop all loop types, including plain */ | |
3031 | ||
3032 | PERL_STATIC_INLINE void | |
c9182d9c | 3033 | Perl_cx_poploop(pTHX_ PERL_CONTEXT *cx) |
d1b6bf72 DM |
3034 | { |
3035 | PERL_ARGS_ASSERT_CX_POPLOOP; | |
3036 | ||
3037 | assert(CxTYPE_is_LOOP(cx)); | |
3038 | if ( CxTYPE(cx) == CXt_LOOP_ARY | |
3039 | || CxTYPE(cx) == CXt_LOOP_LAZYSV) | |
3040 | { | |
3041 | /* Free ary or cur. This assumes that state_u.ary.ary | |
3042 | * aligns with state_u.lazysv.cur. See cx_dup() */ | |
3043 | SV *sv = cx->blk_loop.state_u.lazysv.cur; | |
3044 | cx->blk_loop.state_u.lazysv.cur = NULL; | |
3045 | SvREFCNT_dec_NN(sv); | |
3046 | if (CxTYPE(cx) == CXt_LOOP_LAZYSV) { | |
3047 | sv = cx->blk_loop.state_u.lazysv.end; | |
3048 | cx->blk_loop.state_u.lazysv.end = NULL; | |
3049 | SvREFCNT_dec_NN(sv); | |
3050 | } | |
3051 | } | |
3052 | if (cx->cx_type & (CXp_FOR_PAD|CXp_FOR_GV)) { | |
3053 | SV *cursv; | |
3054 | SV **svp = (cx)->blk_loop.itervar_u.svp; | |
3055 | if ((cx->cx_type & CXp_FOR_GV)) | |
3056 | svp = &GvSV((GV*)svp); | |
3057 | cursv = *svp; | |
3058 | *svp = cx->blk_loop.itersave; | |
3059 | cx->blk_loop.itersave = NULL; | |
3060 | SvREFCNT_dec(cursv); | |
3061 | } | |
3062 | } | |
3063 | ||
2a7b7c61 DM |
3064 | |
3065 | PERL_STATIC_INLINE void | |
c9182d9c | 3066 | Perl_cx_pushwhen(pTHX_ PERL_CONTEXT *cx) |
2a7b7c61 | 3067 | { |
7896dde7 | 3068 | PERL_ARGS_ASSERT_CX_PUSHWHEN; |
2a7b7c61 | 3069 | |
7896dde7 | 3070 | cx->blk_givwhen.leave_op = cLOGOP->op_other; |
2a7b7c61 DM |
3071 | } |
3072 | ||
3073 | ||
3074 | PERL_STATIC_INLINE void | |
c9182d9c | 3075 | Perl_cx_popwhen(pTHX_ PERL_CONTEXT *cx) |
2a7b7c61 | 3076 | { |
7896dde7 Z |
3077 | PERL_ARGS_ASSERT_CX_POPWHEN; |
3078 | assert(CxTYPE(cx) == CXt_WHEN); | |
2a7b7c61 DM |
3079 | |
3080 | PERL_UNUSED_ARG(cx); | |
59a14f30 | 3081 | PERL_UNUSED_CONTEXT; |
2a7b7c61 DM |
3082 | /* currently NOOP */ |
3083 | } | |
3084 | ||
3085 | ||
7896dde7 | 3086 | PERL_STATIC_INLINE void |
c9182d9c | 3087 | Perl_cx_pushgiven(pTHX_ PERL_CONTEXT *cx, SV *orig_defsv) |
7896dde7 Z |
3088 | { |
3089 | PERL_ARGS_ASSERT_CX_PUSHGIVEN; | |
3090 | ||
3091 | cx->blk_givwhen.leave_op = cLOGOP->op_other; | |
3092 | cx->blk_givwhen.defsv_save = orig_defsv; | |
3093 | } | |
3094 | ||
3095 | ||
3096 | PERL_STATIC_INLINE void | |
c9182d9c | 3097 | Perl_cx_popgiven(pTHX_ PERL_CONTEXT *cx) |
7896dde7 Z |
3098 | { |
3099 | SV *sv; | |
3100 | ||
3101 | PERL_ARGS_ASSERT_CX_POPGIVEN; | |
3102 | assert(CxTYPE(cx) == CXt_GIVEN); | |
3103 | ||
3104 | sv = GvSV(PL_defgv); | |
3105 | GvSV(PL_defgv) = cx->blk_givwhen.defsv_save; | |
3106 | cx->blk_givwhen.defsv_save = NULL; | |
3107 | SvREFCNT_dec(sv); | |
3108 | } | |
3109 | ||
ec2c235b KW |
3110 | /* ------------------ util.h ------------------------------------------- */ |
3111 | ||
3112 | /* | |
3f620621 | 3113 | =for apidoc_section $string |
ec2c235b KW |
3114 | |
3115 | =for apidoc foldEQ | |
3116 | ||
3117 | Returns true if the leading C<len> bytes of the strings C<s1> and C<s2> are the | |
3118 | same | |
3119 | case-insensitively; false otherwise. Uppercase and lowercase ASCII range bytes | |
3120 | match themselves and their opposite case counterparts. Non-cased and non-ASCII | |
3121 | range bytes match only themselves. | |
3122 | ||
3123 | =cut | |
3124 | */ | |
3125 | ||
3126 | PERL_STATIC_INLINE I32 | |
3127 | Perl_foldEQ(const char *s1, const char *s2, I32 len) | |
3128 | { | |
3129 | const U8 *a = (const U8 *)s1; | |
3130 | const U8 *b = (const U8 *)s2; | |
3131 | ||
3132 | PERL_ARGS_ASSERT_FOLDEQ; | |
3133 | ||
3134 | assert(len >= 0); | |
3135 | ||
3136 | while (len--) { | |
1604cfb0 MS |
3137 | if (*a != *b && *a != PL_fold[*b]) |
3138 | return 0; | |
3139 | a++,b++; | |
ec2c235b KW |
3140 | } |
3141 | return 1; | |
3142 | } | |
3143 | ||
0f9cb40c | 3144 | PERL_STATIC_INLINE I32 |
ec2c235b KW |
3145 | Perl_foldEQ_latin1(const char *s1, const char *s2, I32 len) |
3146 | { | |
79a1fabd KW |
3147 | /* Compare non-UTF-8 using Unicode (Latin1) semantics. Works on all folds |
3148 | * representable without UTF-8, except for LATIN_SMALL_LETTER_SHARP_S, and | |
3149 | * does not check for this. Nor does it check that the strings each have | |
3150 | * at least 'len' characters. */ | |
ec2c235b KW |
3151 | |
3152 | const U8 *a = (const U8 *)s1; | |
3153 | const U8 *b = (const U8 *)s2; | |
3154 | ||
3155 | PERL_ARGS_ASSERT_FOLDEQ_LATIN1; | |
3156 | ||
3157 | assert(len >= 0); | |
3158 | ||
3159 | while (len--) { | |
1604cfb0 MS |
3160 | if (*a != *b && *a != PL_fold_latin1[*b]) { |
3161 | return 0; | |
3162 | } | |
3163 | a++, b++; | |
ec2c235b KW |
3164 | } |
3165 | return 1; | |
3166 | } | |
3167 | ||
3168 | /* | |
3f620621 | 3169 | =for apidoc_section $locale |
ec2c235b KW |
3170 | =for apidoc foldEQ_locale |
3171 | ||
3172 | Returns true if the leading C<len> bytes of the strings C<s1> and C<s2> are the | |
3173 | same case-insensitively in the current locale; false otherwise. | |
3174 | ||
3175 | =cut | |
3176 | */ | |
3177 | ||
0f9cb40c | 3178 | PERL_STATIC_INLINE I32 |
ec2c235b KW |
3179 | Perl_foldEQ_locale(const char *s1, const char *s2, I32 len) |
3180 | { | |
ec2c235b KW |
3181 | const U8 *a = (const U8 *)s1; |
3182 | const U8 *b = (const U8 *)s2; | |
3183 | ||
3184 | PERL_ARGS_ASSERT_FOLDEQ_LOCALE; | |
3185 | ||
3186 | assert(len >= 0); | |
3187 | ||
3188 | while (len--) { | |
1604cfb0 MS |
3189 | if (*a != *b && *a != PL_fold_locale[*b]) |
3190 | return 0; | |
3191 | a++,b++; | |
ec2c235b KW |
3192 | } |
3193 | return 1; | |
3194 | } | |
3195 | ||
1ab100a8 | 3196 | /* |
3f620621 | 3197 | =for apidoc_section $string |
1ab100a8 KW |
3198 | =for apidoc my_strnlen |
3199 | ||
3200 | The C library C<strnlen> if available, or a Perl implementation of it. | |
3201 | ||
3202 | C<my_strnlen()> computes the length of the string, up to C<maxlen> | |
a3815e44 | 3203 | characters. It will never attempt to address more than C<maxlen> |
1ab100a8 KW |
3204 | characters, making it suitable for use with strings that are not |
3205 | guaranteed to be NUL-terminated. | |
3206 | ||
3207 | =cut | |
3208 | ||
3209 | Description stolen from http://man.openbsd.org/strnlen.3, | |
3210 | implementation stolen from PostgreSQL. | |
3211 | */ | |
3212 | #ifndef HAS_STRNLEN | |
3213 | ||
3214 | PERL_STATIC_INLINE Size_t | |
3215 | Perl_my_strnlen(const char *str, Size_t maxlen) | |
3216 | { | |
3217 | const char *end = (char *) memchr(str, '\0', maxlen); | |
3218 | ||
3219 | PERL_ARGS_ASSERT_MY_STRNLEN; | |
3220 | ||
3221 | if (end == NULL) return maxlen; | |
3222 | return end - str; | |
3223 | } | |
3224 | ||
3225 | #endif | |
3226 | ||
6dba01e2 KW |
3227 | #if ! defined (HAS_MEMRCHR) && (defined(PERL_CORE) || defined(PERL_EXT)) |
3228 | ||
3229 | PERL_STATIC_INLINE void * | |
3230 | S_my_memrchr(const char * s, const char c, const STRLEN len) | |
3231 | { | |
3232 | /* memrchr(), since many platforms lack it */ | |
3233 | ||
3234 | const char * t = s + len - 1; | |
3235 | ||
3236 | PERL_ARGS_ASSERT_MY_MEMRCHR; | |
3237 | ||
3238 | while (t >= s) { | |
3239 | if (*t == c) { | |
3240 | return (void *) t; | |
3241 | } | |
3242 | t--; | |
3243 | } | |
3244 | ||
3245 | return NULL; | |
3246 | } | |
3247 | ||
3248 | #endif | |
3249 | ||
24f3e849 KW |
3250 | PERL_STATIC_INLINE char * |
3251 | Perl_mortal_getenv(const char * str) | |
3252 | { | |
3253 | /* This implements a (mostly) thread-safe, sequential-call-safe getenv(). | |
3254 | * | |
03694582 KW |
3255 | * It's (mostly) thread-safe because it uses a mutex to prevent other |
3256 | * threads (that look at this mutex) from destroying the result before this | |
3257 | * routine has a chance to copy the result to a place that won't be | |
3258 | * destroyed before the caller gets a chance to handle it. That place is a | |
3259 | * mortal SV. khw chose this over SAVEFREEPV because he is under the | |
3260 | * impression that the SV will hang around longer under more circumstances | |
24f3e849 | 3261 | * |
03694582 KW |
3262 | * The reason it isn't completely thread-safe is that other code could |
3263 | * simply not pay attention to the mutex. All of the Perl core uses the | |
3264 | * mutex, but it is possible for code from, say XS, to not use this mutex, | |
3265 | * defeating the safety. | |
24f3e849 | 3266 | * |
03694582 KW |
3267 | * getenv() returns, in some implementations, a pointer to a spot in the |
3268 | * **environ array, which could be invalidated at any time by this or | |
3269 | * another thread changing the environment. Other implementations copy the | |
3270 | * **environ value to a static buffer, returning a pointer to that. That | |
3271 | * buffer might or might not be invalidated by a getenv() call in another | |
3272 | * thread. If it does get zapped, we need an exclusive lock. Otherwise, | |
3273 | * many getenv() calls can safely be running simultaneously, so a | |
3274 | * many-reader (but no simultaneous writers) lock is ok. There is a | |
3275 | * Configure probe to see if another thread destroys the buffer, and the | |
3276 | * mutex is defined accordingly. | |
3277 | * | |
3278 | * But in all cases, using the mutex prevents these problems, as long as | |
3279 | * all code uses the same mutex.. | |
24f3e849 KW |
3280 | * |
3281 | * A complication is that this can be called during phases where the | |
3282 | * mortalization process isn't available. These are in interpreter | |
3283 | * destruction or early in construction. khw believes that at these times | |
3284 | * there shouldn't be anything else going on, so plain getenv is safe AS | |
3285 | * LONG AS the caller acts on the return before calling it again. */ | |
3286 | ||
3287 | char * ret; | |
3288 | dTHX; | |
3289 | ||
3290 | PERL_ARGS_ASSERT_MORTAL_GETENV; | |
3291 | ||
3292 | /* Can't mortalize without stacks. khw believes that no other threads | |
3293 | * should be running, so no need to lock things, and this may be during a | |
3294 | * phase when locking isn't even available */ | |
3295 | if (UNLIKELY(PL_scopestack_ix == 0)) { | |
3296 | return getenv(str); | |
3297 | } | |
3298 | ||
03694582 KW |
3299 | #ifdef PERL_MEM_LOG |
3300 | ||
3301 | /* A major complication arises under PERL_MEM_LOG. When that is active, | |
3302 | * every memory allocation may result in logging, depending on the value of | |
3303 | * ENV{PERL_MEM_LOG} at the moment. That means, as we create the SV for | |
3304 | * saving ENV{foo}'s value (but before saving it), the logging code will | |
3305 | * call us recursively to find out what ENV{PERL_MEM_LOG} is. Without some | |
3306 | * care that could lead to: 1) infinite recursion; or 2) deadlock (trying to | |
3307 | * lock a boolean mutex recursively); 3) destroying the getenv() static | |
3308 | * buffer; or 4) destroying the temporary created by this for the copy | |
3309 | * causes a log entry to be made which could cause a new temporary to be | |
3310 | * created, which will need to be destroyed at some point, leading to an | |
3311 | * infinite loop. | |
3312 | * | |
3313 | * The solution adopted here (after some gnashing of teeth) is to detect | |
3314 | * the recursive calls and calls from the logger, and treat them specially. | |
3315 | * Let's say we want to do getenv("foo"). We first find | |
3316 | * getenv(PERL_MEM_LOG) and save it to a fixed-length per-interpreter | |
3317 | * variable, so no temporary is required. Then we do getenv(foo}, and in | |
3318 | * the process of creating a temporary to save it, this function will be | |
3319 | * called recursively to do a getenv(PERL_MEM_LOG). On the recursed call, | |
3320 | * we detect that it is such a call and return our saved value instead of | |
3321 | * locking and doing a new getenv(). This solves all of problems 1), 2), | |
3322 | * and 3). Because all the getenv()s are done while the mutex is locked, | |
3323 | * the state cannot have changed. To solve 4), we don't create a temporary | |
3324 | * when this is called from the logging code. That code disposes of the | |
3325 | * return value while the mutex is still locked. | |
3326 | * | |
3327 | * The value of getenv(PERL_MEM_LOG) can be anything, but only initial | |
3328 | * digits and 3 particular letters are significant; the rest are ignored by | |
3329 | * the memory logging code. Thus the per-interpreter variable only needs | |
3330 | * to be large enough to save the significant information, the size of | |
3331 | * which is known at compile time. The first byte is extra, reserved for | |
3332 | * flags for our use. To protect against overflowing, only the reserved | |
3333 | * byte, as many digits as don't overflow, and the three letters are | |
3334 | * stored. | |
3335 | * | |
3336 | * The reserved byte has two bits: | |
3337 | * 0x1 if set indicates that if we get here, it is a recursive call of | |
3338 | * getenv() | |
3339 | * 0x2 if set indicates that the call is from the logging code. | |
3340 | * | |
3341 | * If the flag indicates this is a recursive call, just return the stored | |
3342 | * value of PL_mem_log; An empty value gets turned into NULL. */ | |
3343 | if (strEQ(str, "PERL_MEM_LOG") && PL_mem_log[0] & 0x1) { | |
3344 | if (PL_mem_log[1] == '\0') { | |
3345 | return NULL; | |
3346 | } else { | |
3347 | return PL_mem_log + 1; | |
3348 | } | |
3349 | } | |
3350 | ||
3351 | #endif | |
3352 | ||
35bcf7ff | 3353 | GETENV_LOCK; |
24f3e849 | 3354 | |
03694582 KW |
3355 | #ifdef PERL_MEM_LOG |
3356 | ||
3357 | /* Here we are in a critical section. As explained above, we do our own | |
3358 | * getenv(PERL_MEM_LOG), saving the result safely. */ | |
3359 | ret = getenv("PERL_MEM_LOG"); | |
3360 | if (ret == NULL) { /* No logging active */ | |
3361 | ||
3362 | /* Return that immediately if called from the logging code */ | |
3363 | if (PL_mem_log[0] & 0x2) { | |
3364 | GETENV_UNLOCK; | |
3365 | return NULL; | |
3366 | } | |
3367 | ||
3368 | PL_mem_log[1] = '\0'; | |
3369 | } | |
3370 | else { | |
3371 | char *mem_log_meat = PL_mem_log + 1; /* first byte reserved */ | |
3372 | ||
3373 | /* There is nothing to prevent the value of PERL_MEM_LOG from being an | |
3374 | * extremely long string. But we want only a few characters from it. | |
3375 | * PL_mem_log has been made large enough to hold just the ones we need. | |
3376 | * First the file descriptor. */ | |
3377 | if (isDIGIT(*ret)) { | |
3378 | const char * s = ret; | |
3379 | if (UNLIKELY(*s == '0')) { | |
3380 | ||
3381 | /* Reduce multiple leading zeros to a single one. This is to | |
3382 | * allow the caller to change what to do with leading zeros. */ | |
3383 | *mem_log_meat++ = '0'; | |
3384 | s++; | |
3385 | while (*s == '0') { | |
3386 | s++; | |
3387 | } | |
3388 | } | |
3389 | ||
3390 | /* If the input overflows, copy just enough for the result to also | |
3391 | * overflow, plus 1 to make sure */ | |
3392 | while (isDIGIT(*s) && s < ret + TYPE_DIGITS(UV) + 1) { | |
3393 | *mem_log_meat++ = *s++; | |
3394 | } | |
3395 | } | |
3396 | ||
3397 | /* Then each of the three significant characters */ | |
3398 | if (strchr(ret, 'm')) { | |
3399 | *mem_log_meat++ = 'm'; | |
3400 | } | |
3401 | if (strchr(ret, 's')) { | |
3402 | *mem_log_meat++ = 's'; | |
3403 | } | |
3404 | if (strchr(ret, 't')) { | |
3405 | *mem_log_meat++ = 't'; | |
3406 | } | |
3407 | *mem_log_meat = '\0'; | |
3408 | ||
3409 | assert(mem_log_meat < PL_mem_log + sizeof(PL_mem_log)); | |
3410 | } | |
3411 | ||
3412 | /* If we are being called from the logger, it only needs the significant | |
3413 | * portion of PERL_MEM_LOG, and doesn't need a safe copy */ | |
3414 | if (PL_mem_log[0] & 0x2) { | |
3415 | assert(strEQ(str, "PERL_MEM_LOG")); | |
3416 | GETENV_UNLOCK; | |
3417 | return PL_mem_log + 1; | |
3418 | } | |
3419 | ||
3420 | /* Here is a generic getenv(). This could be a getenv("PERL_MEM_LOG") that | |
3421 | * is coming from other than the logging code, so it should be treated the | |
3422 | * same as any other getenv(), returning the full value, not just the | |
3423 | * significant part, and having its value saved. Set the flag that | |
3424 | * indicates any call to this routine will be a recursion from here */ | |
3425 | PL_mem_log[0] = 0x1; | |
3426 | ||
3427 | #endif | |
3428 | ||
3429 | /* Now get the value of the real desired variable, and save a copy */ | |
24f3e849 KW |
3430 | ret = getenv(str); |
3431 | ||
3432 | if (ret != NULL) { | |
c80a8618 | 3433 | ret = SvPVX( newSVpvn_flags(ret, strlen(ret) ,SVs_TEMP) ); |
24f3e849 KW |
3434 | } |
3435 | ||
35bcf7ff KW |
3436 | GETENV_UNLOCK; |
3437 | ||
03694582 KW |
3438 | #ifdef PERL_MEM_LOG |
3439 | ||
3440 | /* Clear the buffer */ | |
3441 | Zero(PL_mem_log, sizeof(PL_mem_log), char); | |
3442 | ||
3443 | #endif | |
3444 | ||
24f3e849 KW |
3445 | return ret; |
3446 | } | |
3447 | ||
1d0d673f PE |
3448 | PERL_STATIC_INLINE bool |
3449 | Perl_sv_isbool(pTHX_ const SV *sv) | |
3450 | { | |
3451 | return SvIOK(sv) && SvPOK(sv) && SvIsCOW_static(sv) && | |
3452 | (SvPVX_const(sv) == PL_Yes || SvPVX_const(sv) == PL_No); | |
3453 | } | |
3454 | ||
9c913148 TC |
3455 | #ifdef USE_ITHREADS |
3456 | ||
3457 | PERL_STATIC_INLINE AV * | |
3458 | Perl_cop_file_avn(pTHX_ const COP *cop) { | |
3459 | ||
3460 | PERL_ARGS_ASSERT_COP_FILE_AVN; | |
3461 | ||
3462 | const char *file = CopFILE(cop); | |
3463 | if (file) { | |
3464 | GV *gv = gv_fetchfile_flags(file, strlen(file), GVF_NOADD); | |
3465 | if (gv) { | |
3466 | return GvAVn(gv); | |
3467 | } | |
3468 | else | |
3469 | return NULL; | |
3470 | } | |
3471 | else | |
3472 | return NULL; | |
3473 | } | |
3474 | ||
3475 | #endif | |
3476 | ||
ed382232 | 3477 | /* |
c8028aa6 TC |
3478 | * ex: set ts=8 sts=4 sw=4 et: |
3479 | */ |