This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
study_chunk: extract rck_elide_nothing
[perl5.git] / pp.c
CommitLineData
a0d0e21e 1/* pp.c
79072805 2 *
1129b882
NC
3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others
79072805 5 *
a0d0e21e
LW
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
79072805 8 *
a0d0e21e
LW
9 */
10
11/*
4ac71550
TC
12 * 'It's a big house this, and very peculiar. Always a bit more
13 * to discover, and no knowing what you'll find round a corner.
14 * And Elves, sir!' --Samwise Gamgee
15 *
16 * [p.225 of _The Lord of the Rings_, II/i: "Many Meetings"]
a0d0e21e 17 */
79072805 18
166f8a29
DM
19/* This file contains general pp ("push/pop") functions that execute the
20 * opcodes that make up a perl program. A typical pp function expects to
21 * find its arguments on the stack, and usually pushes its results onto
22 * the stack, hence the 'pp' terminology. Each OP structure contains
23 * a pointer to the relevant pp_foo() function.
24 */
25
79072805 26#include "EXTERN.h"
864dbfa3 27#define PERL_IN_PP_C
79072805 28#include "perl.h"
77bc9082 29#include "keywords.h"
79072805 30
dbb3849a 31#include "invlist_inline.h"
a4af207c 32#include "reentr.h"
685289b5 33#include "regcharclass.h"
a4af207c 34
13017935
SM
35/* variations on pp_null */
36
93a17b20
LW
37PP(pp_stub)
38{
39644a26 39 dSP;
54310121 40 if (GIMME_V == G_SCALAR)
3280af22 41 XPUSHs(&PL_sv_undef);
93a17b20
LW
42 RETURN;
43}
44
79072805
LW
45/* Pushy stuff. */
46
a46a7b6e 47
93a17b20 48
ac217057
FC
49PP(pp_padcv)
50{
20b7effb 51 dSP; dTARGET;
97b03d64
FC
52 assert(SvTYPE(TARG) == SVt_PVCV);
53 XPUSHs(TARG);
54 RETURN;
ac217057
FC
55}
56
ecf9c8b7
FC
57PP(pp_introcv)
58{
20b7effb 59 dTARGET;
6d5c2147
FC
60 SvPADSTALE_off(TARG);
61 return NORMAL;
ecf9c8b7
FC
62}
63
13f89586
FC
64PP(pp_clonecv)
65{
20b7effb 66 dTARGET;
0f94cb1f
FC
67 CV * const protocv = PadnamePROTOCV(
68 PadlistNAMESARRAY(CvPADLIST(find_runcv(NULL)))[ARGTARG]
69 );
6d5c2147 70 assert(SvTYPE(TARG) == SVt_PVCV);
0f94cb1f
FC
71 assert(protocv);
72 if (CvISXSUB(protocv)) { /* constant */
6d5c2147 73 /* XXX Should we clone it here? */
6d5c2147
FC
74 /* If this changes to use SAVECLEARSV, we can move the SAVECLEARSV
75 to introcv and remove the SvPADSTALE_off. */
76 SAVEPADSVANDMORTALIZE(ARGTARG);
0f94cb1f 77 PAD_SVl(ARGTARG) = SvREFCNT_inc_simple_NN(protocv);
6d5c2147
FC
78 }
79 else {
0f94cb1f
FC
80 if (CvROOT(protocv)) {
81 assert(CvCLONE(protocv));
82 assert(!CvCLONED(protocv));
6d5c2147 83 }
0f94cb1f 84 cv_clone_into(protocv,(CV *)TARG);
6d5c2147
FC
85 SAVECLEARSV(PAD_SVl(ARGTARG));
86 }
87 return NORMAL;
13f89586
FC
88}
89
79072805
LW
90/* Translations. */
91
6f7909da
FC
92/* In some cases this function inspects PL_op. If this function is called
93 for new op types, more bool parameters may need to be added in place of
94 the checks.
95
96 When noinit is true, the absence of a gv will cause a retval of undef.
97 This is unrelated to the cv-to-gv assignment case.
6f7909da
FC
98*/
99
100static SV *
101S_rv2gv(pTHX_ SV *sv, const bool vivify_sv, const bool strict,
102 const bool noinit)
103{
f64c9ac5 104 if (!isGV(sv) || SvFAKE(sv)) SvGETMAGIC(sv);
ed6116ce 105 if (SvROK(sv)) {
93d7320b
DM
106 if (SvAMAGIC(sv)) {
107 sv = amagic_deref_call(sv, to_gv_amg);
93d7320b 108 }
e4a1664f 109 wasref:
ed6116ce 110 sv = SvRV(sv);
b1dadf13 111 if (SvTYPE(sv) == SVt_PVIO) {
159b6efe 112 GV * const gv = MUTABLE_GV(sv_newmortal());
885f468a 113 gv_init(gv, 0, "__ANONIO__", 10, 0);
a45c7426 114 GvIOp(gv) = MUTABLE_IO(sv);
b37c2d43 115 SvREFCNT_inc_void_NN(sv);
ad64d0ec 116 sv = MUTABLE_SV(gv);
ef54e1a4 117 }
81d52ecd
JH
118 else if (!isGV_with_GP(sv)) {
119 Perl_die(aTHX_ "Not a GLOB reference");
120 }
79072805
LW
121 }
122 else {
6e592b3a 123 if (!isGV_with_GP(sv)) {
f132ae69 124 if (!SvOK(sv)) {
b13b2135 125 /* If this is a 'my' scalar and flag is set then vivify
853846ea 126 * NI-S 1999/05/07
b13b2135 127 */
f132ae69 128 if (vivify_sv && sv != &PL_sv_undef) {
2c8ac474 129 GV *gv;
db9848c8 130 HV *stash;
ce74145d 131 if (SvREADONLY(sv))
cb077ed2 132 Perl_croak_no_modify();
db9848c8
Z
133 gv = MUTABLE_GV(newSV(0));
134 stash = CopSTASH(PL_curcop);
135 if (SvTYPE(stash) != SVt_PVHV) stash = NULL;
2c8ac474 136 if (cUNOP->op_targ) {
0bd48802 137 SV * const namesv = PAD_SV(cUNOP->op_targ);
94e7eb6f 138 gv_init_sv(gv, stash, namesv, 0);
2c8ac474
GS
139 }
140 else {
db9848c8 141 gv_init_pv(gv, stash, "__ANONIO__", 0);
1d8d4d2a 142 }
43230e26 143 prepare_SV_for_RV(sv);
ad64d0ec 144 SvRV_set(sv, MUTABLE_SV(gv));
853846ea 145 SvROK_on(sv);
1d8d4d2a 146 SvSETMAGIC(sv);
853846ea 147 goto wasref;
2c8ac474 148 }
81d52ecd
JH
149 if (PL_op->op_flags & OPf_REF || strict) {
150 Perl_die(aTHX_ PL_no_usym, "a symbol");
151 }
599cee73 152 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 153 report_uninit(sv);
6f7909da 154 return &PL_sv_undef;
a0d0e21e 155 }
6f7909da 156 if (noinit)
35cd451c 157 {
77cb3b01
FC
158 if (!(sv = MUTABLE_SV(gv_fetchsv_nomg(
159 sv, GV_ADDMG, SVt_PVGV
23496c6e 160 ))))
6f7909da 161 return &PL_sv_undef;
35cd451c
GS
162 }
163 else {
81d52ecd
JH
164 if (strict) {
165 Perl_die(aTHX_
fedf30e1 166 PL_no_symref_sv,
81d52ecd
JH
167 sv,
168 (SvPOKp(sv) && SvCUR(sv)>32 ? "..." : ""),
169 "a symbol"
170 );
171 }
e26df76a
NC
172 if ((PL_op->op_private & (OPpLVAL_INTRO|OPpDONT_INIT_GV))
173 == OPpDONT_INIT_GV) {
174 /* We are the target of a coderef assignment. Return
175 the scalar unchanged, and let pp_sasssign deal with
176 things. */
6f7909da 177 return sv;
e26df76a 178 }
77cb3b01 179 sv = MUTABLE_SV(gv_fetchsv_nomg(sv, GV_ADD, SVt_PVGV));
35cd451c 180 }
2acc3314 181 /* FAKE globs in the symbol table cause weird bugs (#77810) */
96293f45 182 SvFAKE_off(sv);
93a17b20 183 }
79072805 184 }
8dc99089 185 if (SvFAKE(sv) && !(PL_op->op_private & OPpALLOW_FAKE)) {
2acc3314 186 SV *newsv = sv_newmortal();
5cf4b255 187 sv_setsv_flags(newsv, sv, 0);
2acc3314 188 SvFAKE_off(newsv);
d8906c05 189 sv = newsv;
2acc3314 190 }
6f7909da
FC
191 return sv;
192}
193
194PP(pp_rv2gv)
195{
20b7effb 196 dSP; dTOPss;
6f7909da
FC
197
198 sv = S_rv2gv(aTHX_
199 sv, PL_op->op_private & OPpDEREF,
200 PL_op->op_private & HINT_STRICT_REFS,
201 ((PL_op->op_flags & OPf_SPECIAL) && !(PL_op->op_flags & OPf_MOD))
202 || PL_op->op_type == OP_READLINE
203 );
d8906c05
FC
204 if (PL_op->op_private & OPpLVAL_INTRO)
205 save_gp(MUTABLE_GV(sv), !(PL_op->op_flags & OPf_SPECIAL));
206 SETs(sv);
79072805
LW
207 RETURN;
208}
209
dc3c76f8
NC
210/* Helper function for pp_rv2sv and pp_rv2av */
211GV *
fe9845cc
RB
212Perl_softref2xv(pTHX_ SV *const sv, const char *const what,
213 const svtype type, SV ***spp)
dc3c76f8 214{
dc3c76f8
NC
215 GV *gv;
216
7918f24d
NC
217 PERL_ARGS_ASSERT_SOFTREF2XV;
218
dc3c76f8
NC
219 if (PL_op->op_private & HINT_STRICT_REFS) {
220 if (SvOK(sv))
fedf30e1 221 Perl_die(aTHX_ PL_no_symref_sv, sv,
bf3d870f 222 (SvPOKp(sv) && SvCUR(sv)>32 ? "..." : ""), what);
dc3c76f8
NC
223 else
224 Perl_die(aTHX_ PL_no_usym, what);
225 }
226 if (!SvOK(sv)) {
fd1d9b5c 227 if (
c8fe3bdf 228 PL_op->op_flags & OPf_REF
fd1d9b5c 229 )
dc3c76f8
NC
230 Perl_die(aTHX_ PL_no_usym, what);
231 if (ckWARN(WARN_UNINITIALIZED))
232 report_uninit(sv);
233 if (type != SVt_PV && GIMME_V == G_ARRAY) {
234 (*spp)--;
235 return NULL;
236 }
237 **spp = &PL_sv_undef;
238 return NULL;
239 }
240 if ((PL_op->op_flags & OPf_SPECIAL) &&
241 !(PL_op->op_flags & OPf_MOD))
242 {
77cb3b01 243 if (!(gv = gv_fetchsv_nomg(sv, GV_ADDMG, type)))
dc3c76f8
NC
244 {
245 **spp = &PL_sv_undef;
246 return NULL;
247 }
248 }
249 else {
77cb3b01 250 gv = gv_fetchsv_nomg(sv, GV_ADD, type);
dc3c76f8
NC
251 }
252 return gv;
253}
254
79072805
LW
255PP(pp_rv2sv)
256{
20b7effb 257 dSP; dTOPss;
c445ea15 258 GV *gv = NULL;
79072805 259
9026059d 260 SvGETMAGIC(sv);
ed6116ce 261 if (SvROK(sv)) {
93d7320b
DM
262 if (SvAMAGIC(sv)) {
263 sv = amagic_deref_call(sv, to_sv_amg);
93d7320b 264 }
f5284f61 265
ed6116ce 266 sv = SvRV(sv);
69f00f67 267 if (SvTYPE(sv) >= SVt_PVAV)
cea2e8a9 268 DIE(aTHX_ "Not a SCALAR reference");
79072805
LW
269 }
270 else {
159b6efe 271 gv = MUTABLE_GV(sv);
748a9306 272
6e592b3a 273 if (!isGV_with_GP(gv)) {
dc3c76f8
NC
274 gv = Perl_softref2xv(aTHX_ sv, "a SCALAR", SVt_PV, &sp);
275 if (!gv)
276 RETURN;
463ee0b2 277 }
29c711a3 278 sv = GvSVn(gv);
a0d0e21e 279 }
533c011a 280 if (PL_op->op_flags & OPf_MOD) {
82d03984
RGS
281 if (PL_op->op_private & OPpLVAL_INTRO) {
282 if (cUNOP->op_first->op_type == OP_NULL)
159b6efe 283 sv = save_scalar(MUTABLE_GV(TOPs));
82d03984
RGS
284 else if (gv)
285 sv = save_scalar(gv);
286 else
f1f66076 287 Perl_croak(aTHX_ "%s", PL_no_localize_ref);
82d03984 288 }
533c011a 289 else if (PL_op->op_private & OPpDEREF)
9026059d 290 sv = vivify_ref(sv, PL_op->op_private & OPpDEREF);
79072805 291 }
655f5b26 292 SPAGAIN; /* in case chasing soft refs reallocated the stack */
a0d0e21e 293 SETs(sv);
79072805
LW
294 RETURN;
295}
296
297PP(pp_av2arylen)
298{
20b7effb 299 dSP;
502c6561 300 AV * const av = MUTABLE_AV(TOPs);
02d85cc3
EB
301 const I32 lvalue = PL_op->op_flags & OPf_MOD || LVRET;
302 if (lvalue) {
8160c8f5
DM
303 SV ** const svp = Perl_av_arylen_p(aTHX_ MUTABLE_AV(av));
304 if (!*svp) {
305 *svp = newSV_type(SVt_PVMG);
306 sv_magic(*svp, MUTABLE_SV(av), PERL_MAGIC_arylen, NULL, 0);
02d85cc3 307 }
8160c8f5 308 SETs(*svp);
02d85cc3 309 } else {
e1dccc0d 310 SETs(sv_2mortal(newSViv(AvFILL(MUTABLE_AV(av)))));
79072805 311 }
79072805
LW
312 RETURN;
313}
314
a0d0e21e
LW
315PP(pp_pos)
316{
27a8dde8 317 dSP; dTOPss;
8ec5e241 318
78f9721b 319 if (PL_op->op_flags & OPf_MOD || LVRET) {
d14578b8 320 SV * const ret = sv_2mortal(newSV_type(SVt_PVLV));/* Not TARG RT#67838 */
16eb5365
FC
321 sv_magic(ret, NULL, PERL_MAGIC_pos, NULL, 0);
322 LvTYPE(ret) = '.';
323 LvTARG(ret) = SvREFCNT_inc_simple(sv);
27a8dde8 324 SETs(ret); /* no SvSETMAGIC */
a0d0e21e
LW
325 }
326 else {
96c2a8ff 327 const MAGIC * const mg = mg_find_mglob(sv);
6174b39a 328 if (mg && mg->mg_len != -1) {
6174b39a 329 STRLEN i = mg->mg_len;
7b394f12
DM
330 if (PL_op->op_private & OPpTRUEBOOL)
331 SETs(i ? &PL_sv_yes : &PL_sv_zero);
332 else {
333 dTARGET;
334 if (mg->mg_flags & MGf_BYTES && DO_UTF8(sv))
335 i = sv_pos_b2u_flags(sv, i, SV_GMAGIC|SV_CONST_RETURN);
336 SETu(i);
337 }
27a8dde8 338 return NORMAL;
a0d0e21e 339 }
27a8dde8 340 SETs(&PL_sv_undef);
a0d0e21e 341 }
27a8dde8 342 return NORMAL;
a0d0e21e
LW
343}
344
79072805
LW
345PP(pp_rv2cv)
346{
20b7effb 347 dSP;
79072805 348 GV *gv;
1eced8f8 349 HV *stash_unused;
c445ea15 350 const I32 flags = (PL_op->op_flags & OPf_SPECIAL)
9da346da 351 ? GV_ADDMG
d14578b8
KW
352 : ((PL_op->op_private & (OPpLVAL_INTRO|OPpMAY_RETURN_CONSTANT))
353 == OPpMAY_RETURN_CONSTANT)
c445ea15
AL
354 ? GV_ADD|GV_NOEXPAND
355 : GV_ADD;
4633a7c4
LW
356 /* We usually try to add a non-existent subroutine in case of AUTOLOAD. */
357 /* (But not in defined().) */
e26df76a 358
1eced8f8 359 CV *cv = sv_2cv(TOPs, &stash_unused, &gv, flags);
5a20ba3d 360 if (cv) NOOP;
e26df76a 361 else if ((flags == (GV_ADD|GV_NOEXPAND)) && gv && SvROK(gv)) {
2eaf799e
FC
362 cv = SvTYPE(SvRV(gv)) == SVt_PVCV
363 ? MUTABLE_CV(SvRV(gv))
364 : MUTABLE_CV(gv);
a8e41ef4 365 }
07055b4c 366 else
ea726b52 367 cv = MUTABLE_CV(&PL_sv_undef);
ad64d0ec 368 SETs(MUTABLE_SV(cv));
3d79e3ee 369 return NORMAL;
79072805
LW
370}
371
c07a80fd 372PP(pp_prototype)
373{
20b7effb 374 dSP;
c07a80fd 375 CV *cv;
376 HV *stash;
377 GV *gv;
fabdb6c0 378 SV *ret = &PL_sv_undef;
c07a80fd 379
6954f42f 380 if (SvGMAGICAL(TOPs)) SETs(sv_mortalcopy(TOPs));
b6c543e3 381 if (SvPOK(TOPs) && SvCUR(TOPs) >= 7) {
e3f73d4e 382 const char * s = SvPVX_const(TOPs);
0f12654f 383 if (memBEGINs(s, SvCUR(TOPs), "CORE::")) {
be1b855b 384 const int code = keyword(s + 6, SvCUR(TOPs) - 6, 1);
a96df643 385 if (!code)
147e3846 386 DIE(aTHX_ "Can't find an opnumber for \"%" UTF8f "\"",
b17a0679 387 UTF8fARG(SvFLAGS(TOPs) & SVf_UTF8, SvCUR(TOPs)-6, s+6));
4e338c21 388 {
b66130dd
FC
389 SV * const sv = core_prototype(NULL, s + 6, code, NULL);
390 if (sv) ret = sv;
391 }
b8c38f0a 392 goto set;
b6c543e3
IZ
393 }
394 }
f2c0649b 395 cv = sv_2cv(TOPs, &stash, &gv, 0);
5f05dabc 396 if (cv && SvPOK(cv))
8fa6a409
FC
397 ret = newSVpvn_flags(
398 CvPROTO(cv), CvPROTOLEN(cv), SVs_TEMP | SvUTF8(cv)
399 );
b6c543e3 400 set:
c07a80fd 401 SETs(ret);
402 RETURN;
403}
404
a0d0e21e
LW
405PP(pp_anoncode)
406{
20b7effb 407 dSP;
ea726b52 408 CV *cv = MUTABLE_CV(PAD_SV(PL_op->op_targ));
a5f75d66 409 if (CvCLONE(cv))
ad64d0ec 410 cv = MUTABLE_CV(sv_2mortal(MUTABLE_SV(cv_clone(cv))));
5f05dabc 411 EXTEND(SP,1);
ad64d0ec 412 PUSHs(MUTABLE_SV(cv));
a0d0e21e
LW
413 RETURN;
414}
415
416PP(pp_srefgen)
79072805 417{
20b7effb 418 dSP;
71be2cbc 419 *SP = refto(*SP);
3ed34c76 420 return NORMAL;
8ec5e241 421}
a0d0e21e
LW
422
423PP(pp_refgen)
424{
20b7effb 425 dSP; dMARK;
82334630 426 if (GIMME_V != G_ARRAY) {
5f0b1d4e
GS
427 if (++MARK <= SP)
428 *MARK = *SP;
429 else
1d51ab6c
FC
430 {
431 MEXTEND(SP, 1);
3280af22 432 *MARK = &PL_sv_undef;
1d51ab6c 433 }
5f0b1d4e
GS
434 *MARK = refto(*MARK);
435 SP = MARK;
436 RETURN;
a0d0e21e 437 }
bbce6d69 438 EXTEND_MORTAL(SP - MARK);
71be2cbc 439 while (++MARK <= SP)
440 *MARK = refto(*MARK);
a0d0e21e 441 RETURN;
79072805
LW
442}
443
76e3520e 444STATIC SV*
cea2e8a9 445S_refto(pTHX_ SV *sv)
71be2cbc 446{
447 SV* rv;
448
7918f24d
NC
449 PERL_ARGS_ASSERT_REFTO;
450
71be2cbc 451 if (SvTYPE(sv) == SVt_PVLV && LvTYPE(sv) == 'y') {
452 if (LvTARGLEN(sv))
68dc0745 453 vivify_defelem(sv);
454 if (!(sv = LvTARG(sv)))
3280af22 455 sv = &PL_sv_undef;
0dd88869 456 else
b37c2d43 457 SvREFCNT_inc_void_NN(sv);
71be2cbc 458 }
d8b46c1b 459 else if (SvTYPE(sv) == SVt_PVAV) {
502c6561
NC
460 if (!AvREAL((const AV *)sv) && AvREIFY((const AV *)sv))
461 av_reify(MUTABLE_AV(sv));
d8b46c1b 462 SvTEMP_off(sv);
b37c2d43 463 SvREFCNT_inc_void_NN(sv);
d8b46c1b 464 }
60779a30 465 else if (SvPADTMP(sv)) {
f2933f5f 466 sv = newSVsv(sv);
60779a30 467 }
1f1dcfb5
FC
468 else if (UNLIKELY(SvSMAGICAL(sv) && mg_find(sv, PERL_MAGIC_nonelem)))
469 sv_unmagic(SvREFCNT_inc_simple_NN(sv), PERL_MAGIC_nonelem);
71be2cbc 470 else {
471 SvTEMP_off(sv);
b37c2d43 472 SvREFCNT_inc_void_NN(sv);
71be2cbc 473 }
474 rv = sv_newmortal();
4df7f6af 475 sv_upgrade(rv, SVt_IV);
b162af07 476 SvRV_set(rv, sv);
71be2cbc 477 SvROK_on(rv);
478 return rv;
479}
480
79072805
LW
481PP(pp_ref)
482{
3c1e67ac
DD
483 dSP;
484 SV * const sv = TOPs;
f12c7020 485
511ddbdf 486 SvGETMAGIC(sv);
ba75e9a4 487 if (!SvROK(sv)) {
3c1e67ac 488 SETs(&PL_sv_no);
ba75e9a4
DM
489 return NORMAL;
490 }
491
492 /* op is in boolean context? */
493 if ( (PL_op->op_private & OPpTRUEBOOL)
494 || ( (PL_op->op_private & OPpMAYBE_TRUEBOOL)
495 && block_gimme() == G_VOID))
496 {
497 /* refs are always true - unless it's to an object blessed into a
498 * class with a false name, i.e. "0". So we have to check for
499 * that remote possibility. The following is is basically an
500 * unrolled SvTRUE(sv_reftype(rv)) */
501 SV * const rv = SvRV(sv);
502 if (SvOBJECT(rv)) {
503 HV *stash = SvSTASH(rv);
504 HEK *hek = HvNAME_HEK(stash);
505 if (hek) {
506 I32 len = HEK_LEN(hek);
507 /* bail out and do it the hard way? */
508 if (UNLIKELY(
509 len == HEf_SVKEY
510 || (len == 1 && HEK_KEY(hek)[0] == '0')
511 ))
512 goto do_sv_ref;
513 }
514 }
515 SETs(&PL_sv_yes);
516 return NORMAL;
517 }
518
519 do_sv_ref:
520 {
3c1e67ac
DD
521 dTARGET;
522 SETs(TARG);
ba75e9a4 523 sv_ref(TARG, SvRV(sv), TRUE);
a10e04b5 524 SvSETMAGIC(TARG);
ba75e9a4 525 return NORMAL;
3c1e67ac 526 }
79072805 527
79072805
LW
528}
529
ba75e9a4 530
79072805
LW
531PP(pp_bless)
532{
20b7effb 533 dSP;
463ee0b2 534 HV *stash;
79072805 535
463ee0b2 536 if (MAXARG == 1)
dcdfe746 537 {
c2f922f1 538 curstash:
11faa288 539 stash = CopSTASH(PL_curcop);
dcdfe746
FC
540 if (SvTYPE(stash) != SVt_PVHV)
541 Perl_croak(aTHX_ "Attempt to bless into a freed package");
542 }
7b8d334a 543 else {
1b6737cc 544 SV * const ssv = POPs;
7b8d334a 545 STRLEN len;
e1ec3a88 546 const char *ptr;
81689caa 547
c2f922f1 548 if (!ssv) goto curstash;
8d9dd4b9 549 SvGETMAGIC(ssv);
c7ea825d
FC
550 if (SvROK(ssv)) {
551 if (!SvAMAGIC(ssv)) {
552 frog:
81689caa 553 Perl_croak(aTHX_ "Attempt to bless into a reference");
c7ea825d
FC
554 }
555 /* SvAMAGIC is on here, but it only means potentially overloaded,
556 so after stringification: */
557 ptr = SvPV_nomg_const(ssv,len);
558 /* We need to check the flag again: */
559 if (!SvAMAGIC(ssv)) goto frog;
560 }
561 else ptr = SvPV_nomg_const(ssv,len);
a2a5de95
NC
562 if (len == 0)
563 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
564 "Explicit blessing to '' (assuming package main)");
e69c50fe 565 stash = gv_stashpvn(ptr, len, GV_ADD|SvUTF8(ssv));
7b8d334a 566 }
a0d0e21e 567
5d3fdfeb 568 (void)sv_bless(TOPs, stash);
79072805
LW
569 RETURN;
570}
571
fb73857a 572PP(pp_gelem)
573{
20b7effb 574 dSP;
b13b2135 575
1b6737cc 576 SV *sv = POPs;
a180b31a
BF
577 STRLEN len;
578 const char * const elem = SvPV_const(sv, len);
5695161e 579 GV * const gv = MUTABLE_GV(TOPs);
c445ea15 580 SV * tmpRef = NULL;
1b6737cc 581
c445ea15 582 sv = NULL;
c4ba80c3
NC
583 if (elem) {
584 /* elem will always be NUL terminated. */
c4ba80c3
NC
585 switch (*elem) {
586 case 'A':
500f3e18 587 if (memEQs(elem, len, "ARRAY"))
e14698d8 588 {
ad64d0ec 589 tmpRef = MUTABLE_SV(GvAV(gv));
e14698d8
FC
590 if (tmpRef && !AvREAL((const AV *)tmpRef)
591 && AvREIFY((const AV *)tmpRef))
592 av_reify(MUTABLE_AV(tmpRef));
593 }
c4ba80c3
NC
594 break;
595 case 'C':
500f3e18 596 if (memEQs(elem, len, "CODE"))
ad64d0ec 597 tmpRef = MUTABLE_SV(GvCVu(gv));
c4ba80c3
NC
598 break;
599 case 'F':
500f3e18 600 if (memEQs(elem, len, "FILEHANDLE")) {
ad64d0ec 601 tmpRef = MUTABLE_SV(GvIOp(gv));
c4ba80c3
NC
602 }
603 else
500f3e18 604 if (memEQs(elem, len, "FORMAT"))
ad64d0ec 605 tmpRef = MUTABLE_SV(GvFORM(gv));
c4ba80c3
NC
606 break;
607 case 'G':
500f3e18 608 if (memEQs(elem, len, "GLOB"))
ad64d0ec 609 tmpRef = MUTABLE_SV(gv);
c4ba80c3
NC
610 break;
611 case 'H':
500f3e18 612 if (memEQs(elem, len, "HASH"))
ad64d0ec 613 tmpRef = MUTABLE_SV(GvHV(gv));
c4ba80c3
NC
614 break;
615 case 'I':
500f3e18 616 if (memEQs(elem, len, "IO"))
ad64d0ec 617 tmpRef = MUTABLE_SV(GvIOp(gv));
c4ba80c3
NC
618 break;
619 case 'N':
500f3e18 620 if (memEQs(elem, len, "NAME"))
a663657d 621 sv = newSVhek(GvNAME_HEK(gv));
c4ba80c3
NC
622 break;
623 case 'P':
500f3e18 624 if (memEQs(elem, len, "PACKAGE")) {
7fa3a4ab
NC
625 const HV * const stash = GvSTASH(gv);
626 const HEK * const hek = stash ? HvNAME_HEK(stash) : NULL;
396482e1 627 sv = hek ? newSVhek(hek) : newSVpvs("__ANON__");
c4ba80c3
NC
628 }
629 break;
630 case 'S':
500f3e18 631 if (memEQs(elem, len, "SCALAR"))
f9d52e31 632 tmpRef = GvSVn(gv);
c4ba80c3 633 break;
39b99f21 634 }
fb73857a 635 }
76e3520e
GS
636 if (tmpRef)
637 sv = newRV(tmpRef);
fb73857a 638 if (sv)
639 sv_2mortal(sv);
640 else
3280af22 641 sv = &PL_sv_undef;
5695161e 642 SETs(sv);
fb73857a 643 RETURN;
644}
645
a0d0e21e 646/* Pattern matching */
79072805 647
a0d0e21e 648PP(pp_study)
79072805 649{
add3e777 650 dSP; dTOPss;
a0d0e21e
LW
651 STRLEN len;
652
1fa930f2 653 (void)SvPV(sv, len);
bc9a5256 654 if (len == 0 || len > I32_MAX || !SvPOK(sv) || SvUTF8(sv) || SvVALID(sv)) {
32f0ea87 655 /* Historically, study was skipped in these cases. */
add3e777
FC
656 SETs(&PL_sv_no);
657 return NORMAL;
a4f4e906
NC
658 }
659
a58a85fa 660 /* Make study a no-op. It's no longer useful and its existence
32f0ea87 661 complicates matters elsewhere. */
add3e777
FC
662 SETs(&PL_sv_yes);
663 return NORMAL;
79072805
LW
664}
665
b1c05ba5
DM
666
667/* also used for: pp_transr() */
668
a0d0e21e 669PP(pp_trans)
79072805 670{
a8e41ef4 671 dSP;
a0d0e21e
LW
672 SV *sv;
673
533c011a 674 if (PL_op->op_flags & OPf_STACKED)
a0d0e21e 675 sv = POPs;
79072805 676 else {
a0d0e21e 677 EXTEND(SP,1);
f605e527 678 if (ARGTARG)
6442877a 679 sv = PAD_SV(ARGTARG);
f605e527
FC
680 else {
681 sv = DEFSV;
682 }
79072805 683 }
bb16bae8 684 if(PL_op->op_type == OP_TRANSR) {
290797f7
FC
685 STRLEN len;
686 const char * const pv = SvPV(sv,len);
687 SV * const newsv = newSVpvn_flags(pv, len, SVs_TEMP|SvUTF8(sv));
bb16bae8 688 do_trans(newsv);
290797f7 689 PUSHs(newsv);
bb16bae8 690 }
5bbe7184 691 else {
f0fd0980
DM
692 Size_t i = do_trans(sv);
693 mPUSHi((UV)i);
5bbe7184 694 }
a0d0e21e 695 RETURN;
79072805
LW
696}
697
a0d0e21e 698/* Lvalue operators. */
79072805 699
f595e19f 700static size_t
81745e4e
NC
701S_do_chomp(pTHX_ SV *retval, SV *sv, bool chomping)
702{
81745e4e
NC
703 STRLEN len;
704 char *s;
f595e19f 705 size_t count = 0;
81745e4e
NC
706
707 PERL_ARGS_ASSERT_DO_CHOMP;
708
709 if (chomping && (RsSNARF(PL_rs) || RsRECORD(PL_rs)))
f595e19f 710 return 0;
81745e4e
NC
711 if (SvTYPE(sv) == SVt_PVAV) {
712 I32 i;
713 AV *const av = MUTABLE_AV(sv);
714 const I32 max = AvFILL(av);
715
716 for (i = 0; i <= max; i++) {
717 sv = MUTABLE_SV(av_fetch(av, i, FALSE));
718 if (sv && ((sv = *(SV**)sv), sv != &PL_sv_undef))
f595e19f 719 count += do_chomp(retval, sv, chomping);
81745e4e 720 }
f595e19f 721 return count;
81745e4e
NC
722 }
723 else if (SvTYPE(sv) == SVt_PVHV) {
724 HV* const hv = MUTABLE_HV(sv);
725 HE* entry;
726 (void)hv_iterinit(hv);
727 while ((entry = hv_iternext(hv)))
f595e19f
FC
728 count += do_chomp(retval, hv_iterval(hv,entry), chomping);
729 return count;
81745e4e
NC
730 }
731 else if (SvREADONLY(sv)) {
cb077ed2 732 Perl_croak_no_modify();
81745e4e
NC
733 }
734
81745e4e
NC
735 s = SvPV(sv, len);
736 if (chomping) {
81745e4e 737 if (s && len) {
997c424a
DD
738 char *temp_buffer = NULL;
739 SV *svrecode = NULL;
81745e4e
NC
740 s += --len;
741 if (RsPARA(PL_rs)) {
742 if (*s != '\n')
997c424a 743 goto nope_free_nothing;
f595e19f 744 ++count;
81745e4e
NC
745 while (len && s[-1] == '\n') {
746 --len;
747 --s;
f595e19f 748 ++count;
81745e4e
NC
749 }
750 }
751 else {
752 STRLEN rslen, rs_charlen;
753 const char *rsptr = SvPV_const(PL_rs, rslen);
754
755 rs_charlen = SvUTF8(PL_rs)
756 ? sv_len_utf8(PL_rs)
757 : rslen;
758
759 if (SvUTF8(PL_rs) != SvUTF8(sv)) {
760 /* Assumption is that rs is shorter than the scalar. */
761 if (SvUTF8(PL_rs)) {
762 /* RS is utf8, scalar is 8 bit. */
763 bool is_utf8 = TRUE;
764 temp_buffer = (char*)bytes_from_utf8((U8*)rsptr,
765 &rslen, &is_utf8);
766 if (is_utf8) {
997c424a
DD
767 /* Cannot downgrade, therefore cannot possibly match.
768 At this point, temp_buffer is not alloced, and
769 is the buffer inside PL_rs, so dont free it.
81745e4e
NC
770 */
771 assert (temp_buffer == rsptr);
997c424a 772 goto nope_free_sv;
81745e4e
NC
773 }
774 rsptr = temp_buffer;
775 }
81745e4e
NC
776 else {
777 /* RS is 8 bit, scalar is utf8. */
778 temp_buffer = (char*)bytes_to_utf8((U8*)rsptr, &rslen);
779 rsptr = temp_buffer;
780 }
781 }
782 if (rslen == 1) {
783 if (*s != *rsptr)
997c424a 784 goto nope_free_all;
f595e19f 785 ++count;
81745e4e
NC
786 }
787 else {
788 if (len < rslen - 1)
997c424a 789 goto nope_free_all;
81745e4e
NC
790 len -= rslen - 1;
791 s -= rslen - 1;
792 if (memNE(s, rsptr, rslen))
997c424a 793 goto nope_free_all;
f595e19f 794 count += rs_charlen;
81745e4e
NC
795 }
796 }
3b7ded39 797 SvPV_force_nomg_nolen(sv);
81745e4e
NC
798 SvCUR_set(sv, len);
799 *SvEND(sv) = '\0';
800 SvNIOK_off(sv);
801 SvSETMAGIC(sv);
81745e4e 802
997c424a
DD
803 nope_free_all:
804 Safefree(temp_buffer);
805 nope_free_sv:
806 SvREFCNT_dec(svrecode);
807 nope_free_nothing: ;
808 }
81745e4e 809 } else {
f8c80a8e 810 if (len && (!SvPOK(sv) || SvIsCOW(sv)))
81745e4e
NC
811 s = SvPV_force_nomg(sv, len);
812 if (DO_UTF8(sv)) {
813 if (s && len) {
814 char * const send = s + len;
815 char * const start = s;
816 s = send - 1;
817 while (s > start && UTF8_IS_CONTINUATION(*s))
818 s--;
819 if (is_utf8_string((U8*)s, send - s)) {
820 sv_setpvn(retval, s, send - s);
821 *s = '\0';
822 SvCUR_set(sv, s - start);
823 SvNIOK_off(sv);
824 SvUTF8_on(retval);
825 }
826 }
827 else
500f3e18 828 SvPVCLEAR(retval);
81745e4e
NC
829 }
830 else if (s && len) {
831 s += --len;
832 sv_setpvn(retval, s, 1);
833 *s = '\0';
834 SvCUR_set(sv, len);
835 SvUTF8_off(sv);
836 SvNIOK_off(sv);
837 }
838 else
500f3e18 839 SvPVCLEAR(retval);
81745e4e
NC
840 SvSETMAGIC(sv);
841 }
f595e19f 842 return count;
81745e4e
NC
843}
844
b1c05ba5
DM
845
846/* also used for: pp_schomp() */
847
a0d0e21e
LW
848PP(pp_schop)
849{
20b7effb 850 dSP; dTARGET;
fa54efae
NC
851 const bool chomping = PL_op->op_type == OP_SCHOMP;
852
f595e19f 853 const size_t count = do_chomp(TARG, TOPs, chomping);
fa54efae 854 if (chomping)
f595e19f 855 sv_setiv(TARG, count);
a0d0e21e 856 SETTARG;
ee41d8c7 857 return NORMAL;
79072805
LW
858}
859
b1c05ba5
DM
860
861/* also used for: pp_chomp() */
862
a0d0e21e 863PP(pp_chop)
79072805 864{
20b7effb 865 dSP; dMARK; dTARGET; dORIGMARK;
fa54efae 866 const bool chomping = PL_op->op_type == OP_CHOMP;
f595e19f 867 size_t count = 0;
8ec5e241 868
20cf1f79 869 while (MARK < SP)
f595e19f
FC
870 count += do_chomp(TARG, *++MARK, chomping);
871 if (chomping)
872 sv_setiv(TARG, count);
20cf1f79
NC
873 SP = ORIGMARK;
874 XPUSHTARG;
a0d0e21e 875 RETURN;
79072805
LW
876}
877
a0d0e21e
LW
878PP(pp_undef)
879{
20b7effb 880 dSP;
a0d0e21e
LW
881 SV *sv;
882
533c011a 883 if (!PL_op->op_private) {
774d564b 884 EXTEND(SP, 1);
a0d0e21e 885 RETPUSHUNDEF;
774d564b 886 }
79072805 887
821f14b0 888 sv = TOPs;
a0d0e21e 889 if (!sv)
821f14b0
FC
890 {
891 SETs(&PL_sv_undef);
892 return NORMAL;
893 }
85e6fe83 894
4dda930b
FC
895 if (SvTHINKFIRST(sv))
896 sv_force_normal_flags(sv, SV_COW_DROP_PV|SV_IMMEDIATE_UNREF);
85e6fe83 897
a0d0e21e
LW
898 switch (SvTYPE(sv)) {
899 case SVt_NULL:
900 break;
901 case SVt_PVAV:
60edcf09 902 av_undef(MUTABLE_AV(sv));
a0d0e21e
LW
903 break;
904 case SVt_PVHV:
60edcf09 905 hv_undef(MUTABLE_HV(sv));
a0d0e21e
LW
906 break;
907 case SVt_PVCV:
a2a5de95 908 if (cv_const_sv((const CV *)sv))
714cd18f 909 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
147e3846 910 "Constant subroutine %" SVf " undefined",
714cd18f
BF
911 SVfARG(CvANON((const CV *)sv)
912 ? newSVpvs_flags("(anonymous)", SVs_TEMP)
bdbfc51a
FC
913 : sv_2mortal(newSVhek(
914 CvNAMED(sv)
915 ? CvNAME_HEK((CV *)sv)
916 : GvENAME_HEK(CvGV((const CV *)sv))
917 ))
918 ));
5f66b61c 919 /* FALLTHROUGH */
9607fc9c 920 case SVt_PVFM:
6fc92669 921 /* let user-undef'd sub keep its identity */
b7acb0a3 922 cv_undef_flags(MUTABLE_CV(sv), CV_UNDEF_KEEP_NAME);
a0d0e21e 923 break;
8e07c86e 924 case SVt_PVGV:
bc1df6c2
FC
925 assert(isGV_with_GP(sv));
926 assert(!SvFAKE(sv));
927 {
20408e3c 928 GP *gp;
dd69841b
BB
929 HV *stash;
930
dd69841b 931 /* undef *Pkg::meth_name ... */
e530fb81
FC
932 bool method_changed
933 = GvCVu((const GV *)sv) && (stash = GvSTASH((const GV *)sv))
934 && HvENAME_get(stash);
935 /* undef *Foo:: */
936 if((stash = GvHV((const GV *)sv))) {
937 if(HvENAME_get(stash))
938 SvREFCNT_inc_simple_void_NN(sv_2mortal((SV *)stash));
939 else stash = NULL;
940 }
dd69841b 941
795eb8c8 942 SvREFCNT_inc_simple_void_NN(sv_2mortal(sv));
159b6efe 943 gp_free(MUTABLE_GV(sv));
a02a5408 944 Newxz(gp, 1, GP);
c43ae56f 945 GvGP_set(sv, gp_ref(gp));
2e3295e3 946#ifndef PERL_DONT_CREATE_GVSV
561b68a9 947 GvSV(sv) = newSV(0);
2e3295e3 948#endif
57843af0 949 GvLINE(sv) = CopLINE(PL_curcop);
159b6efe 950 GvEGV(sv) = MUTABLE_GV(sv);
20408e3c 951 GvMULTI_on(sv);
e530fb81
FC
952
953 if(stash)
afdbe55d 954 mro_package_moved(NULL, stash, (const GV *)sv, 0);
e530fb81
FC
955 stash = NULL;
956 /* undef *Foo::ISA */
957 if( strEQ(GvNAME((const GV *)sv), "ISA")
958 && (stash = GvSTASH((const GV *)sv))
959 && (method_changed || HvENAME(stash)) )
960 mro_isa_changed_in(stash);
961 else if(method_changed)
962 mro_method_changed_in(
da9043f5 963 GvSTASH((const GV *)sv)
e530fb81
FC
964 );
965
6e592b3a 966 break;
20408e3c 967 }
a0d0e21e 968 default:
b15aece3 969 if (SvTYPE(sv) >= SVt_PV && SvPVX_const(sv) && SvLEN(sv)) {
8bd4d4c5 970 SvPV_free(sv);
c445ea15 971 SvPV_set(sv, NULL);
4633a7c4 972 SvLEN_set(sv, 0);
a0d0e21e 973 }
0c34ef67 974 SvOK_off(sv);
4633a7c4 975 SvSETMAGIC(sv);
79072805 976 }
a0d0e21e 977
821f14b0
FC
978 SETs(&PL_sv_undef);
979 return NORMAL;
79072805
LW
980}
981
b1c05ba5 982
20e96431 983/* common "slow" code for pp_postinc and pp_postdec */
b1c05ba5 984
20e96431
DM
985static OP *
986S_postincdec_common(pTHX_ SV *sv, SV *targ)
a0d0e21e 987{
20e96431 988 dSP;
c22c99bc
FC
989 const bool inc =
990 PL_op->op_type == OP_POSTINC || PL_op->op_type == OP_I_POSTINC;
20e96431
DM
991
992 if (SvROK(sv))
7dcb9b98 993 TARG = sv_newmortal();
20e96431
DM
994 sv_setsv(TARG, sv);
995 if (inc)
996 sv_inc_nomg(sv);
997 else
998 sv_dec_nomg(sv);
999 SvSETMAGIC(sv);
1e54a23f 1000 /* special case for undef: see thread at 2003-03/msg00536.html in archive */
c22c99bc 1001 if (inc && !SvOK(TARG))
a0d0e21e 1002 sv_setiv(TARG, 0);
e87de4ab 1003 SETTARG;
a0d0e21e
LW
1004 return NORMAL;
1005}
79072805 1006
20e96431
DM
1007
1008/* also used for: pp_i_postinc() */
1009
1010PP(pp_postinc)
1011{
1012 dSP; dTARGET;
1013 SV *sv = TOPs;
1014
1015 /* special-case sv being a simple integer */
1016 if (LIKELY(((sv->sv_flags &
1017 (SVf_THINKFIRST|SVs_GMG|SVf_IVisUV|
1018 SVf_IOK|SVf_NOK|SVf_POK|SVp_NOK|SVp_POK|SVf_ROK))
1019 == SVf_IOK))
1020 && SvIVX(sv) != IV_MAX)
1021 {
1022 IV iv = SvIVX(sv);
1023 SvIV_set(sv, iv + 1);
1024 TARGi(iv, 0); /* arg not GMG, so can't be tainted */
1025 SETs(TARG);
1026 return NORMAL;
1027 }
1028
1029 return S_postincdec_common(aTHX_ sv, TARG);
1030}
1031
1032
1033/* also used for: pp_i_postdec() */
1034
1035PP(pp_postdec)
1036{
1037 dSP; dTARGET;
1038 SV *sv = TOPs;
1039
1040 /* special-case sv being a simple integer */
1041 if (LIKELY(((sv->sv_flags &
1042 (SVf_THINKFIRST|SVs_GMG|SVf_IVisUV|
1043 SVf_IOK|SVf_NOK|SVf_POK|SVp_NOK|SVp_POK|SVf_ROK))
1044 == SVf_IOK))
1045 && SvIVX(sv) != IV_MIN)
1046 {
1047 IV iv = SvIVX(sv);
1048 SvIV_set(sv, iv - 1);
1049 TARGi(iv, 0); /* arg not GMG, so can't be tainted */
1050 SETs(TARG);
1051 return NORMAL;
1052 }
1053
1054 return S_postincdec_common(aTHX_ sv, TARG);
1055}
1056
1057
a0d0e21e
LW
1058/* Ordinary operators. */
1059
1060PP(pp_pow)
1061{
20b7effb 1062 dSP; dATARGET; SV *svl, *svr;
58d76dfd 1063#ifdef PERL_PRESERVE_IVUV
52a96ae6
HS
1064 bool is_int = 0;
1065#endif
6f1401dc
DM
1066 tryAMAGICbin_MG(pow_amg, AMGf_assign|AMGf_numeric);
1067 svr = TOPs;
1068 svl = TOPm1s;
52a96ae6
HS
1069#ifdef PERL_PRESERVE_IVUV
1070 /* For integer to integer power, we do the calculation by hand wherever
1071 we're sure it is safe; otherwise we call pow() and try to convert to
1072 integer afterwards. */
01f91bf2 1073 if (SvIV_please_nomg(svr) && SvIV_please_nomg(svl)) {
900658e3
PF
1074 UV power;
1075 bool baseuok;
1076 UV baseuv;
1077
800401ee
JH
1078 if (SvUOK(svr)) {
1079 power = SvUVX(svr);
900658e3 1080 } else {
800401ee 1081 const IV iv = SvIVX(svr);
900658e3
PF
1082 if (iv >= 0) {
1083 power = iv;
1084 } else {
1085 goto float_it; /* Can't do negative powers this way. */
1086 }
1087 }
1088
800401ee 1089 baseuok = SvUOK(svl);
900658e3 1090 if (baseuok) {
800401ee 1091 baseuv = SvUVX(svl);
900658e3 1092 } else {
800401ee 1093 const IV iv = SvIVX(svl);
900658e3
PF
1094 if (iv >= 0) {
1095 baseuv = iv;
1096 baseuok = TRUE; /* effectively it's a UV now */
1097 } else {
1098 baseuv = -iv; /* abs, baseuok == false records sign */
1099 }
1100 }
52a96ae6
HS
1101 /* now we have integer ** positive integer. */
1102 is_int = 1;
1103
1104 /* foo & (foo - 1) is zero only for a power of 2. */
58d76dfd 1105 if (!(baseuv & (baseuv - 1))) {
52a96ae6 1106 /* We are raising power-of-2 to a positive integer.
58d76dfd
JH
1107 The logic here will work for any base (even non-integer
1108 bases) but it can be less accurate than
1109 pow (base,power) or exp (power * log (base)) when the
1110 intermediate values start to spill out of the mantissa.
1111 With powers of 2 we know this can't happen.
1112 And powers of 2 are the favourite thing for perl
1113 programmers to notice ** not doing what they mean. */
1114 NV result = 1.0;
1115 NV base = baseuok ? baseuv : -(NV)baseuv;
900658e3
PF
1116
1117 if (power & 1) {
1118 result *= base;
1119 }
1120 while (power >>= 1) {
1121 base *= base;
1122 if (power & 1) {
1123 result *= base;
1124 }
1125 }
58d76dfd
JH
1126 SP--;
1127 SETn( result );
6f1401dc 1128 SvIV_please_nomg(svr);
58d76dfd 1129 RETURN;
52a96ae6 1130 } else {
eb578fdb
KW
1131 unsigned int highbit = 8 * sizeof(UV);
1132 unsigned int diff = 8 * sizeof(UV);
900658e3
PF
1133 while (diff >>= 1) {
1134 highbit -= diff;
1135 if (baseuv >> highbit) {
1136 highbit += diff;
1137 }
52a96ae6
HS
1138 }
1139 /* we now have baseuv < 2 ** highbit */
1140 if (power * highbit <= 8 * sizeof(UV)) {
1141 /* result will definitely fit in UV, so use UV math
1142 on same algorithm as above */
eb578fdb
KW
1143 UV result = 1;
1144 UV base = baseuv;
f2338a2e 1145 const bool odd_power = cBOOL(power & 1);
900658e3
PF
1146 if (odd_power) {
1147 result *= base;
1148 }
1149 while (power >>= 1) {
1150 base *= base;
1151 if (power & 1) {
52a96ae6 1152 result *= base;
52a96ae6
HS
1153 }
1154 }
1155 SP--;
0615a994 1156 if (baseuok || !odd_power)
52a96ae6
HS
1157 /* answer is positive */
1158 SETu( result );
1159 else if (result <= (UV)IV_MAX)
1160 /* answer negative, fits in IV */
1161 SETi( -(IV)result );
a8e41ef4 1162 else if (result == (UV)IV_MIN)
52a96ae6
HS
1163 /* 2's complement assumption: special case IV_MIN */
1164 SETi( IV_MIN );
1165 else
1166 /* answer negative, doesn't fit */
1167 SETn( -(NV)result );
1168 RETURN;
a8e41ef4 1169 }
52a96ae6 1170 }
58d76dfd 1171 }
52a96ae6 1172 float_it:
a8e41ef4 1173#endif
a0d0e21e 1174 {
6f1401dc
DM
1175 NV right = SvNV_nomg(svr);
1176 NV left = SvNV_nomg(svl);
4efa5a16 1177 (void)POPs;
3aaeb624
JA
1178
1179#if defined(USE_LONG_DOUBLE) && defined(HAS_AIX_POWL_NEG_BASE_BUG)
1180 /*
1181 We are building perl with long double support and are on an AIX OS
1182 afflicted with a powl() function that wrongly returns NaNQ for any
1183 negative base. This was reported to IBM as PMR #23047-379 on
1184 03/06/2006. The problem exists in at least the following versions
1185 of AIX and the libm fileset, and no doubt others as well:
1186
1187 AIX 4.3.3-ML10 bos.adt.libm 4.3.3.50
1188 AIX 5.1.0-ML04 bos.adt.libm 5.1.0.29
1189 AIX 5.2.0 bos.adt.libm 5.2.0.85
1190
1191 So, until IBM fixes powl(), we provide the following workaround to
1192 handle the problem ourselves. Our logic is as follows: for
1193 negative bases (left), we use fmod(right, 2) to check if the
1194 exponent is an odd or even integer:
1195
1196 - if odd, powl(left, right) == -powl(-left, right)
1197 - if even, powl(left, right) == powl(-left, right)
1198
1199 If the exponent is not an integer, the result is rightly NaNQ, so
1200 we just return that (as NV_NAN).
1201 */
1202
1203 if (left < 0.0) {
1204 NV mod2 = Perl_fmod( right, 2.0 );
1205 if (mod2 == 1.0 || mod2 == -1.0) { /* odd integer */
1206 SETn( -Perl_pow( -left, right) );
1207 } else if (mod2 == 0.0) { /* even integer */
1208 SETn( Perl_pow( -left, right) );
1209 } else { /* fractional power */
1210 SETn( NV_NAN );
1211 }
1212 } else {
1213 SETn( Perl_pow( left, right) );
1214 }
1215#else
52a96ae6 1216 SETn( Perl_pow( left, right) );
3aaeb624
JA
1217#endif /* HAS_AIX_POWL_NEG_BASE_BUG */
1218
52a96ae6
HS
1219#ifdef PERL_PRESERVE_IVUV
1220 if (is_int)
6f1401dc 1221 SvIV_please_nomg(svr);
52a96ae6
HS
1222#endif
1223 RETURN;
93a17b20 1224 }
a0d0e21e
LW
1225}
1226
1227PP(pp_multiply)
1228{
20b7effb 1229 dSP; dATARGET; SV *svl, *svr;
6f1401dc
DM
1230 tryAMAGICbin_MG(mult_amg, AMGf_assign|AMGf_numeric);
1231 svr = TOPs;
1232 svl = TOPm1s;
230ee21f 1233
28e5dec8 1234#ifdef PERL_PRESERVE_IVUV
230ee21f
DM
1235
1236 /* special-case some simple common cases */
1237 if (!((svl->sv_flags|svr->sv_flags) & (SVf_IVisUV|SVs_GMG))) {
1238 IV il, ir;
1239 U32 flags = (svl->sv_flags & svr->sv_flags);
1240 if (flags & SVf_IOK) {
1241 /* both args are simple IVs */
1242 UV topl, topr;
1243 il = SvIVX(svl);
1244 ir = SvIVX(svr);
1245 do_iv:
1246 topl = ((UV)il) >> (UVSIZE * 4 - 1);
1247 topr = ((UV)ir) >> (UVSIZE * 4 - 1);
1248
1249 /* if both are in a range that can't under/overflow, do a
1250 * simple integer multiply: if the top halves(*) of both numbers
1251 * are 00...00 or 11...11, then it's safe.
1252 * (*) for 32-bits, the "top half" is the top 17 bits,
1253 * for 64-bits, its 33 bits */
1254 if (!(
1255 ((topl+1) | (topr+1))
1256 & ( (((UV)1) << (UVSIZE * 4 + 1)) - 2) /* 11..110 */
1257 )) {
1258 SP--;
1259 TARGi(il * ir, 0); /* args not GMG, so can't be tainted */
1260 SETs(TARG);
1261 RETURN;
1262 }
1263 goto generic;
1264 }
1265 else if (flags & SVf_NOK) {
1266 /* both args are NVs */
1267 NV nl = SvNVX(svl);
1268 NV nr = SvNVX(svr);
1269 NV result;
1270
3a019afd 1271 if (lossless_NV_to_IV(nl, &il) && lossless_NV_to_IV(nr, &ir)) {
230ee21f
DM
1272 /* nothing was lost by converting to IVs */
1273 goto do_iv;
3a019afd 1274 }
230ee21f
DM
1275 SP--;
1276 result = nl * nr;
1f02ab1d 1277# if defined(__sgi) && defined(USE_LONG_DOUBLE) && LONG_DOUBLEKIND == LONG_DOUBLE_IS_DOUBLEDOUBLE_128_BIT_BE_BE && NVSIZE == 16
230ee21f
DM
1278 if (Perl_isinf(result)) {
1279 Zero((U8*)&result + 8, 8, U8);
1280 }
1281# endif
1282 TARGn(result, 0); /* args not GMG, so can't be tainted */
1283 SETs(TARG);
1284 RETURN;
1285 }
1286 }
1287
1288 generic:
1289
01f91bf2 1290 if (SvIV_please_nomg(svr)) {
28e5dec8
JH
1291 /* Unless the left argument is integer in range we are going to have to
1292 use NV maths. Hence only attempt to coerce the right argument if
1293 we know the left is integer. */
1294 /* Left operand is defined, so is it IV? */
01f91bf2 1295 if (SvIV_please_nomg(svl)) {
800401ee
JH
1296 bool auvok = SvUOK(svl);
1297 bool buvok = SvUOK(svr);
28e5dec8
JH
1298 const UV topmask = (~ (UV)0) << (4 * sizeof (UV));
1299 const UV botmask = ~((~ (UV)0) << (4 * sizeof (UV)));
1300 UV alow;
1301 UV ahigh;
1302 UV blow;
1303 UV bhigh;
1304
1305 if (auvok) {
800401ee 1306 alow = SvUVX(svl);
28e5dec8 1307 } else {
800401ee 1308 const IV aiv = SvIVX(svl);
28e5dec8
JH
1309 if (aiv >= 0) {
1310 alow = aiv;
1311 auvok = TRUE; /* effectively it's a UV now */
1312 } else {
10be8dab
KW
1313 /* abs, auvok == false records sign; Using 0- here and
1314 * later to silence bogus warning from MS VC */
1315 alow = (UV) (0 - (UV) aiv);
28e5dec8
JH
1316 }
1317 }
1318 if (buvok) {
800401ee 1319 blow = SvUVX(svr);
28e5dec8 1320 } else {
800401ee 1321 const IV biv = SvIVX(svr);
28e5dec8
JH
1322 if (biv >= 0) {
1323 blow = biv;
1324 buvok = TRUE; /* effectively it's a UV now */
1325 } else {
53e2bfb7 1326 /* abs, buvok == false records sign */
10be8dab 1327 blow = (UV) (0 - (UV) biv);
28e5dec8
JH
1328 }
1329 }
1330
1331 /* If this does sign extension on unsigned it's time for plan B */
1332 ahigh = alow >> (4 * sizeof (UV));
1333 alow &= botmask;
1334 bhigh = blow >> (4 * sizeof (UV));
1335 blow &= botmask;
1336 if (ahigh && bhigh) {
6f207bd3 1337 NOOP;
28e5dec8
JH
1338 /* eg 32 bit is at least 0x10000 * 0x10000 == 0x100000000
1339 which is overflow. Drop to NVs below. */
1340 } else if (!ahigh && !bhigh) {
1341 /* eg 32 bit is at most 0xFFFF * 0xFFFF == 0xFFFE0001
1342 so the unsigned multiply cannot overflow. */
c445ea15 1343 const UV product = alow * blow;
28e5dec8
JH
1344 if (auvok == buvok) {
1345 /* -ve * -ve or +ve * +ve gives a +ve result. */
1346 SP--;
1347 SETu( product );
1348 RETURN;
1349 } else if (product <= (UV)IV_MIN) {
1350 /* 2s complement assumption that (UV)-IV_MIN is correct. */
1351 /* -ve result, which could overflow an IV */
1352 SP--;
02b08bbc
DM
1353 /* can't negate IV_MIN, but there are aren't two
1354 * integers such that !ahigh && !bhigh, where the
1355 * product equals 0x800....000 */
1356 assert(product != (UV)IV_MIN);
25716404 1357 SETi( -(IV)product );
28e5dec8
JH
1358 RETURN;
1359 } /* else drop to NVs below. */
1360 } else {
1361 /* One operand is large, 1 small */
1362 UV product_middle;
1363 if (bhigh) {
1364 /* swap the operands */
1365 ahigh = bhigh;
1366 bhigh = blow; /* bhigh now the temp var for the swap */
1367 blow = alow;
1368 alow = bhigh;
1369 }
1370 /* now, ((ahigh * blow) << half_UV_len) + (alow * blow)
1371 multiplies can't overflow. shift can, add can, -ve can. */
1372 product_middle = ahigh * blow;
1373 if (!(product_middle & topmask)) {
1374 /* OK, (ahigh * blow) won't lose bits when we shift it. */
1375 UV product_low;
1376 product_middle <<= (4 * sizeof (UV));
1377 product_low = alow * blow;
1378
1379 /* as for pp_add, UV + something mustn't get smaller.
1380 IIRC ANSI mandates this wrapping *behaviour* for
1381 unsigned whatever the actual representation*/
1382 product_low += product_middle;
1383 if (product_low >= product_middle) {
1384 /* didn't overflow */
1385 if (auvok == buvok) {
1386 /* -ve * -ve or +ve * +ve gives a +ve result. */
1387 SP--;
1388 SETu( product_low );
1389 RETURN;
1390 } else if (product_low <= (UV)IV_MIN) {
1391 /* 2s complement assumption again */
1392 /* -ve result, which could overflow an IV */
1393 SP--;
53e2bfb7
DM
1394 SETi(product_low == (UV)IV_MIN
1395 ? IV_MIN : -(IV)product_low);
28e5dec8
JH
1396 RETURN;
1397 } /* else drop to NVs below. */
1398 }
1399 } /* product_middle too large */
1400 } /* ahigh && bhigh */
800401ee
JH
1401 } /* SvIOK(svl) */
1402 } /* SvIOK(svr) */
28e5dec8 1403#endif
a0d0e21e 1404 {
6f1401dc
DM
1405 NV right = SvNV_nomg(svr);
1406 NV left = SvNV_nomg(svl);
230ee21f
DM
1407 NV result = left * right;
1408
4efa5a16 1409 (void)POPs;
1f02ab1d 1410#if defined(__sgi) && defined(USE_LONG_DOUBLE) && LONG_DOUBLEKIND == LONG_DOUBLE_IS_DOUBLEDOUBLE_128_BIT_BE_BE && NVSIZE == 16
230ee21f
DM
1411 if (Perl_isinf(result)) {
1412 Zero((U8*)&result + 8, 8, U8);
3ec400f5 1413 }
3ec400f5 1414#endif
230ee21f 1415 SETn(result);
a0d0e21e 1416 RETURN;
79072805 1417 }
a0d0e21e
LW
1418}
1419
1420PP(pp_divide)
1421{
20b7effb 1422 dSP; dATARGET; SV *svl, *svr;
6f1401dc
DM
1423 tryAMAGICbin_MG(div_amg, AMGf_assign|AMGf_numeric);
1424 svr = TOPs;
1425 svl = TOPm1s;
5479d192 1426 /* Only try to do UV divide first
68795e93 1427 if ((SLOPPYDIVIDE is true) or
5479d192
NC
1428 (PERL_PRESERVE_IVUV is true and one or both SV is a UV too large
1429 to preserve))
1430 The assumption is that it is better to use floating point divide
1431 whenever possible, only doing integer divide first if we can't be sure.
1432 If NV_PRESERVES_UV is true then we know at compile time that no UV
1433 can be too large to preserve, so don't need to compile the code to
1434 test the size of UVs. */
1435
00b6a411 1436#if defined(SLOPPYDIVIDE) || (defined(PERL_PRESERVE_IVUV) && !defined(NV_PRESERVES_UV))
5479d192
NC
1437# define PERL_TRY_UV_DIVIDE
1438 /* ensure that 20./5. == 4. */
a0d0e21e 1439#endif
5479d192
NC
1440
1441#ifdef PERL_TRY_UV_DIVIDE
01f91bf2 1442 if (SvIV_please_nomg(svr) && SvIV_please_nomg(svl)) {
800401ee
JH
1443 bool left_non_neg = SvUOK(svl);
1444 bool right_non_neg = SvUOK(svr);
5479d192
NC
1445 UV left;
1446 UV right;
1447
1448 if (right_non_neg) {
800401ee 1449 right = SvUVX(svr);
5479d192
NC
1450 }
1451 else {
800401ee 1452 const IV biv = SvIVX(svr);
5479d192
NC
1453 if (biv >= 0) {
1454 right = biv;
1455 right_non_neg = TRUE; /* effectively it's a UV now */
1456 }
1457 else {
ad9b9a49 1458 right = -(UV)biv;
5479d192
NC
1459 }
1460 }
1461 /* historically undef()/0 gives a "Use of uninitialized value"
1462 warning before dieing, hence this test goes here.
1463 If it were immediately before the second SvIV_please, then
1464 DIE() would be invoked before left was even inspected, so
486ec47a 1465 no inspection would give no warning. */
5479d192
NC
1466 if (right == 0)
1467 DIE(aTHX_ "Illegal division by zero");
1468
1469 if (left_non_neg) {
800401ee 1470 left = SvUVX(svl);
5479d192
NC
1471 }
1472 else {
800401ee 1473 const IV aiv = SvIVX(svl);
5479d192
NC
1474 if (aiv >= 0) {
1475 left = aiv;
1476 left_non_neg = TRUE; /* effectively it's a UV now */
1477 }
1478 else {
ad9b9a49 1479 left = -(UV)aiv;
5479d192
NC
1480 }
1481 }
1482
1483 if (left >= right
1484#ifdef SLOPPYDIVIDE
1485 /* For sloppy divide we always attempt integer division. */
1486#else
1487 /* Otherwise we only attempt it if either or both operands
1488 would not be preserved by an NV. If both fit in NVs
0c2ee62a
NC
1489 we fall through to the NV divide code below. However,
1490 as left >= right to ensure integer result here, we know that
1491 we can skip the test on the right operand - right big
1492 enough not to be preserved can't get here unless left is
1493 also too big. */
1494
1495 && (left > ((UV)1 << NV_PRESERVES_UV_BITS))
5479d192
NC
1496#endif
1497 ) {
1498 /* Integer division can't overflow, but it can be imprecise. */
f1966580
TK
1499
1500 /* Modern compilers optimize division followed by
1501 * modulo into a single div instruction */
1b6737cc 1502 const UV result = left / right;
f1966580 1503 if (left % right == 0) {
5479d192
NC
1504 SP--; /* result is valid */
1505 if (left_non_neg == right_non_neg) {
1506 /* signs identical, result is positive. */
1507 SETu( result );
1508 RETURN;
1509 }
1510 /* 2s complement assumption */
1511 if (result <= (UV)IV_MIN)
02b08bbc 1512 SETi(result == (UV)IV_MIN ? IV_MIN : -(IV)result);
5479d192
NC
1513 else {
1514 /* It's exact but too negative for IV. */
1515 SETn( -(NV)result );
1516 }
1517 RETURN;
1518 } /* tried integer divide but it was not an integer result */
32fdb065 1519 } /* else (PERL_ABS(result) < 1.0) or (both UVs in range for NV) */
01f91bf2 1520 } /* one operand wasn't SvIOK */
5479d192
NC
1521#endif /* PERL_TRY_UV_DIVIDE */
1522 {
6f1401dc
DM
1523 NV right = SvNV_nomg(svr);
1524 NV left = SvNV_nomg(svl);
4efa5a16 1525 (void)POPs;(void)POPs;
ebc6a117
PD
1526#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1527 if (! Perl_isnan(right) && right == 0.0)
1528#else
659c4b96 1529 if (right == 0.0)
ebc6a117 1530#endif
5479d192
NC
1531 DIE(aTHX_ "Illegal division by zero");
1532 PUSHn( left / right );
1533 RETURN;
79072805 1534 }
a0d0e21e
LW
1535}
1536
1537PP(pp_modulo)
1538{
20b7effb 1539 dSP; dATARGET;
6f1401dc 1540 tryAMAGICbin_MG(modulo_amg, AMGf_assign|AMGf_numeric);
a0d0e21e 1541 {
9c5ffd7c
JH
1542 UV left = 0;
1543 UV right = 0;
dc656993
JH
1544 bool left_neg = FALSE;
1545 bool right_neg = FALSE;
e2c88acc
NC
1546 bool use_double = FALSE;
1547 bool dright_valid = FALSE;
9c5ffd7c
JH
1548 NV dright = 0.0;
1549 NV dleft = 0.0;
6f1401dc
DM
1550 SV * const svr = TOPs;
1551 SV * const svl = TOPm1s;
01f91bf2 1552 if (SvIV_please_nomg(svr)) {
800401ee 1553 right_neg = !SvUOK(svr);
e2c88acc 1554 if (!right_neg) {
800401ee 1555 right = SvUVX(svr);
e2c88acc 1556 } else {
800401ee 1557 const IV biv = SvIVX(svr);
e2c88acc
NC
1558 if (biv >= 0) {
1559 right = biv;
1560 right_neg = FALSE; /* effectively it's a UV now */
1561 } else {
10be8dab 1562 right = (UV) (0 - (UV) biv);
e2c88acc
NC
1563 }
1564 }
1565 }
1566 else {
6f1401dc 1567 dright = SvNV_nomg(svr);
787eafbd
IZ
1568 right_neg = dright < 0;
1569 if (right_neg)
1570 dright = -dright;
e2c88acc
NC
1571 if (dright < UV_MAX_P1) {
1572 right = U_V(dright);
1573 dright_valid = TRUE; /* In case we need to use double below. */
1574 } else {
1575 use_double = TRUE;
1576 }
787eafbd 1577 }
a0d0e21e 1578
e2c88acc
NC
1579 /* At this point use_double is only true if right is out of range for
1580 a UV. In range NV has been rounded down to nearest UV and
1581 use_double false. */
01f91bf2 1582 if (!use_double && SvIV_please_nomg(svl)) {
800401ee 1583 left_neg = !SvUOK(svl);
e2c88acc 1584 if (!left_neg) {
800401ee 1585 left = SvUVX(svl);
e2c88acc 1586 } else {
800401ee 1587 const IV aiv = SvIVX(svl);
e2c88acc
NC
1588 if (aiv >= 0) {
1589 left = aiv;
1590 left_neg = FALSE; /* effectively it's a UV now */
1591 } else {
10be8dab 1592 left = (UV) (0 - (UV) aiv);
e2c88acc
NC
1593 }
1594 }
e2c88acc 1595 }
787eafbd 1596 else {
6f1401dc 1597 dleft = SvNV_nomg(svl);
787eafbd
IZ
1598 left_neg = dleft < 0;
1599 if (left_neg)
1600 dleft = -dleft;
68dc0745 1601
e2c88acc
NC
1602 /* This should be exactly the 5.6 behaviour - if left and right are
1603 both in range for UV then use U_V() rather than floor. */
1604 if (!use_double) {
1605 if (dleft < UV_MAX_P1) {
1606 /* right was in range, so is dleft, so use UVs not double.
1607 */
1608 left = U_V(dleft);
1609 }
1610 /* left is out of range for UV, right was in range, so promote
1611 right (back) to double. */
1612 else {
1613 /* The +0.5 is used in 5.6 even though it is not strictly
1614 consistent with the implicit +0 floor in the U_V()
1615 inside the #if 1. */
1616 dleft = Perl_floor(dleft + 0.5);
1617 use_double = TRUE;
1618 if (dright_valid)
1619 dright = Perl_floor(dright + 0.5);
1620 else
1621 dright = right;
1622 }
1623 }
1624 }
6f1401dc 1625 sp -= 2;
787eafbd 1626 if (use_double) {
65202027 1627 NV dans;
787eafbd 1628
659c4b96 1629 if (!dright)
cea2e8a9 1630 DIE(aTHX_ "Illegal modulus zero");
787eafbd 1631
65202027 1632 dans = Perl_fmod(dleft, dright);
659c4b96 1633 if ((left_neg != right_neg) && dans)
787eafbd
IZ
1634 dans = dright - dans;
1635 if (right_neg)
1636 dans = -dans;
1637 sv_setnv(TARG, dans);
1638 }
1639 else {
1640 UV ans;
1641
787eafbd 1642 if (!right)
cea2e8a9 1643 DIE(aTHX_ "Illegal modulus zero");
787eafbd
IZ
1644
1645 ans = left % right;
1646 if ((left_neg != right_neg) && ans)
1647 ans = right - ans;
1648 if (right_neg) {
1649 /* XXX may warn: unary minus operator applied to unsigned type */
1650 /* could change -foo to be (~foo)+1 instead */
1651 if (ans <= ~((UV)IV_MAX)+1)
1652 sv_setiv(TARG, ~ans+1);
1653 else
65202027 1654 sv_setnv(TARG, -(NV)ans);
787eafbd
IZ
1655 }
1656 else
1657 sv_setuv(TARG, ans);
1658 }
1659 PUSHTARG;
1660 RETURN;
79072805 1661 }
a0d0e21e 1662}
79072805 1663
a0d0e21e
LW
1664PP(pp_repeat)
1665{
20b7effb 1666 dSP; dATARGET;
eb578fdb 1667 IV count;
6f1401dc 1668 SV *sv;
02a7a248 1669 bool infnan = FALSE;
490b24f6 1670 const U8 gimme = GIMME_V;
6f1401dc 1671
490b24f6 1672 if (gimme == G_ARRAY && PL_op->op_private & OPpREPEAT_DOLIST) {
6f1401dc
DM
1673 /* TODO: think of some way of doing list-repeat overloading ??? */
1674 sv = POPs;
1675 SvGETMAGIC(sv);
1676 }
1677 else {
3a100dab
FC
1678 if (UNLIKELY(PL_op->op_private & OPpREPEAT_DOLIST)) {
1679 /* The parser saw this as a list repeat, and there
1680 are probably several items on the stack. But we're
1681 in scalar/void context, and there's no pp_list to save us
1682 now. So drop the rest of the items -- robin@kitsite.com
1683 */
1684 dMARK;
1685 if (MARK + 1 < SP) {
1686 MARK[1] = TOPm1s;
1687 MARK[2] = TOPs;
1688 }
1689 else {
1690 dTOPss;
1691 ASSUME(MARK + 1 == SP);
d81b7735
TC
1692 MEXTEND(SP, 1);
1693 PUSHs(sv);
3a100dab
FC
1694 MARK[1] = &PL_sv_undef;
1695 }
1696 SP = MARK + 2;
1697 }
6f1401dc
DM
1698 tryAMAGICbin_MG(repeat_amg, AMGf_assign);
1699 sv = POPs;
1700 }
1701
2b573ace
JH
1702 if (SvIOKp(sv)) {
1703 if (SvUOK(sv)) {
6f1401dc 1704 const UV uv = SvUV_nomg(sv);
2b573ace
JH
1705 if (uv > IV_MAX)
1706 count = IV_MAX; /* The best we can do? */
1707 else
1708 count = uv;
1709 } else {
b3211734 1710 count = SvIV_nomg(sv);
2b573ace
JH
1711 }
1712 }
1713 else if (SvNOKp(sv)) {
02a7a248
JH
1714 const NV nv = SvNV_nomg(sv);
1715 infnan = Perl_isinfnan(nv);
1716 if (UNLIKELY(infnan)) {
1717 count = 0;
1718 } else {
1719 if (nv < 0.0)
1720 count = -1; /* An arbitrary negative integer */
1721 else
1722 count = (IV)nv;
1723 }
2b573ace
JH
1724 }
1725 else
02a7a248 1726 count = SvIV_nomg(sv);
6f1401dc 1727
02a7a248
JH
1728 if (infnan) {
1729 Perl_ck_warner(aTHX_ packWARN(WARN_NUMERIC),
1730 "Non-finite repeat count does nothing");
1731 } else if (count < 0) {
b3211734
KW
1732 count = 0;
1733 Perl_ck_warner(aTHX_ packWARN(WARN_NUMERIC),
02a7a248 1734 "Negative repeat count does nothing");
b3211734
KW
1735 }
1736
490b24f6 1737 if (gimme == G_ARRAY && PL_op->op_private & OPpREPEAT_DOLIST) {
a0d0e21e 1738 dMARK;
052a7c76 1739 const SSize_t items = SP - MARK;
da9e430b 1740 const U8 mod = PL_op->op_flags & OPf_MOD;
79072805 1741
a0d0e21e 1742 if (count > 1) {
052a7c76 1743 SSize_t max;
b3b27d01 1744
052a7c76
DM
1745 if ( items > SSize_t_MAX / count /* max would overflow */
1746 /* repeatcpy would overflow */
1747 || items > I32_MAX / (I32)sizeof(SV *)
b3b27d01
DM
1748 )
1749 Perl_croak(aTHX_ "%s","Out of memory during list extend");
1750 max = items * count;
1751 MEXTEND(MARK, max);
1752
a0d0e21e 1753 while (SP > MARK) {
60779a30
DM
1754 if (*SP) {
1755 if (mod && SvPADTMP(*SP)) {
da9e430b 1756 *SP = sv_mortalcopy(*SP);
60779a30 1757 }
976c8a39 1758 SvTEMP_off((*SP));
da9e430b 1759 }
a0d0e21e 1760 SP--;
79072805 1761 }
a0d0e21e
LW
1762 MARK++;
1763 repeatcpy((char*)(MARK + items), (char*)MARK,
ad64d0ec 1764 items * sizeof(const SV *), count - 1);
a0d0e21e 1765 SP += max;
79072805 1766 }
a0d0e21e 1767 else if (count <= 0)
052a7c76 1768 SP = MARK;
79072805 1769 }
a0d0e21e 1770 else { /* Note: mark already snarfed by pp_list */
0bd48802 1771 SV * const tmpstr = POPs;
a0d0e21e 1772 STRLEN len;
9b877dbb 1773 bool isutf;
a0d0e21e 1774
6f1401dc
DM
1775 if (TARG != tmpstr)
1776 sv_setsv_nomg(TARG, tmpstr);
1777 SvPV_force_nomg(TARG, len);
9b877dbb 1778 isutf = DO_UTF8(TARG);
8ebc5c01 1779 if (count != 1) {
1780 if (count < 1)
1781 SvCUR_set(TARG, 0);
1782 else {
b3b27d01
DM
1783 STRLEN max;
1784
1785 if ( len > (MEM_SIZE_MAX-1) / (UV)count /* max would overflow */
1786 || len > (U32)I32_MAX /* repeatcpy would overflow */
1787 )
1788 Perl_croak(aTHX_ "%s",
1789 "Out of memory during string extend");
1790 max = (UV)count * len + 1;
1791 SvGROW(TARG, max);
1792
a0d0e21e 1793 repeatcpy(SvPVX(TARG) + len, SvPVX(TARG), len, count - 1);
b162af07 1794 SvCUR_set(TARG, SvCUR(TARG) * count);
7a4c00b4 1795 }
a0d0e21e 1796 *SvEND(TARG) = '\0';
a0d0e21e 1797 }
dfcb284a
GS
1798 if (isutf)
1799 (void)SvPOK_only_UTF8(TARG);
1800 else
1801 (void)SvPOK_only(TARG);
b80b6069 1802
a0d0e21e 1803 PUSHTARG;
79072805 1804 }
a0d0e21e
LW
1805 RETURN;
1806}
79072805 1807
a0d0e21e
LW
1808PP(pp_subtract)
1809{
20b7effb 1810 dSP; dATARGET; bool useleft; SV *svl, *svr;
6f1401dc
DM
1811 tryAMAGICbin_MG(subtr_amg, AMGf_assign|AMGf_numeric);
1812 svr = TOPs;
1813 svl = TOPm1s;
230ee21f 1814
28e5dec8 1815#ifdef PERL_PRESERVE_IVUV
230ee21f
DM
1816
1817 /* special-case some simple common cases */
1818 if (!((svl->sv_flags|svr->sv_flags) & (SVf_IVisUV|SVs_GMG))) {
1819 IV il, ir;
1820 U32 flags = (svl->sv_flags & svr->sv_flags);
1821 if (flags & SVf_IOK) {
1822 /* both args are simple IVs */
1823 UV topl, topr;
1824 il = SvIVX(svl);
1825 ir = SvIVX(svr);
1826 do_iv:
1827 topl = ((UV)il) >> (UVSIZE * 8 - 2);
1828 topr = ((UV)ir) >> (UVSIZE * 8 - 2);
1829
1830 /* if both are in a range that can't under/overflow, do a
1831 * simple integer subtract: if the top of both numbers
1832 * are 00 or 11, then it's safe */
1833 if (!( ((topl+1) | (topr+1)) & 2)) {
1834 SP--;
1835 TARGi(il - ir, 0); /* args not GMG, so can't be tainted */
1836 SETs(TARG);
1837 RETURN;
1838 }
1839 goto generic;
1840 }
1841 else if (flags & SVf_NOK) {
1842 /* both args are NVs */
1843 NV nl = SvNVX(svl);
1844 NV nr = SvNVX(svr);
1845
3a019afd 1846 if (lossless_NV_to_IV(nl, &il) && lossless_NV_to_IV(nr, &ir)) {
230ee21f
DM
1847 /* nothing was lost by converting to IVs */
1848 goto do_iv;
3a019afd 1849 }
230ee21f
DM
1850 SP--;
1851 TARGn(nl - nr, 0); /* args not GMG, so can't be tainted */
1852 SETs(TARG);
1853 RETURN;
1854 }
1855 }
1856
1857 generic:
1858
1859 useleft = USE_LEFT(svl);
7dca457a
NC
1860 /* See comments in pp_add (in pp_hot.c) about Overflow, and how
1861 "bad things" happen if you rely on signed integers wrapping. */
01f91bf2 1862 if (SvIV_please_nomg(svr)) {
28e5dec8
JH
1863 /* Unless the left argument is integer in range we are going to have to
1864 use NV maths. Hence only attempt to coerce the right argument if
1865 we know the left is integer. */
eb578fdb 1866 UV auv = 0;
9c5ffd7c 1867 bool auvok = FALSE;
7dca457a
NC
1868 bool a_valid = 0;
1869
28e5dec8 1870 if (!useleft) {
7dca457a
NC
1871 auv = 0;
1872 a_valid = auvok = 1;
1873 /* left operand is undef, treat as zero. */
28e5dec8
JH
1874 } else {
1875 /* Left operand is defined, so is it IV? */
01f91bf2 1876 if (SvIV_please_nomg(svl)) {
800401ee
JH
1877 if ((auvok = SvUOK(svl)))
1878 auv = SvUVX(svl);
7dca457a 1879 else {
eb578fdb 1880 const IV aiv = SvIVX(svl);
7dca457a
NC
1881 if (aiv >= 0) {
1882 auv = aiv;
1883 auvok = 1; /* Now acting as a sign flag. */
ad9b9a49 1884 } else {
10be8dab 1885 auv = (UV) (0 - (UV) aiv);
28e5dec8 1886 }
7dca457a
NC
1887 }
1888 a_valid = 1;
1889 }
1890 }
1891 if (a_valid) {
1892 bool result_good = 0;
1893 UV result;
eb578fdb 1894 UV buv;
800401ee 1895 bool buvok = SvUOK(svr);
a8e41ef4 1896
7dca457a 1897 if (buvok)
800401ee 1898 buv = SvUVX(svr);
7dca457a 1899 else {
eb578fdb 1900 const IV biv = SvIVX(svr);
7dca457a
NC
1901 if (biv >= 0) {
1902 buv = biv;
1903 buvok = 1;
1904 } else
10be8dab 1905 buv = (UV) (0 - (UV) biv);
7dca457a
NC
1906 }
1907 /* ?uvok if value is >= 0. basically, flagged as UV if it's +ve,
602f51c4 1908 else "IV" now, independent of how it came in.
7dca457a
NC
1909 if a, b represents positive, A, B negative, a maps to -A etc
1910 a - b => (a - b)
1911 A - b => -(a + b)
1912 a - B => (a + b)
1913 A - B => -(a - b)
1914 all UV maths. negate result if A negative.
1915 subtract if signs same, add if signs differ. */
1916
1917 if (auvok ^ buvok) {
1918 /* Signs differ. */
1919 result = auv + buv;
1920 if (result >= auv)
1921 result_good = 1;
1922 } else {
1923 /* Signs same */
1924 if (auv >= buv) {
1925 result = auv - buv;
1926 /* Must get smaller */
1927 if (result <= auv)
1928 result_good = 1;
1929 } else {
1930 result = buv - auv;
1931 if (result <= buv) {
1932 /* result really should be -(auv-buv). as its negation
1933 of true value, need to swap our result flag */
1934 auvok = !auvok;
1935 result_good = 1;
28e5dec8 1936 }
28e5dec8
JH
1937 }
1938 }
7dca457a
NC
1939 if (result_good) {
1940 SP--;
1941 if (auvok)
1942 SETu( result );
1943 else {
1944 /* Negate result */
1945 if (result <= (UV)IV_MIN)
53e2bfb7
DM
1946 SETi(result == (UV)IV_MIN
1947 ? IV_MIN : -(IV)result);
7dca457a
NC
1948 else {
1949 /* result valid, but out of range for IV. */
1950 SETn( -(NV)result );
1951 }
1952 }
1953 RETURN;
1954 } /* Overflow, drop through to NVs. */
28e5dec8
JH
1955 }
1956 }
230ee21f
DM
1957#else
1958 useleft = USE_LEFT(svl);
28e5dec8 1959#endif
a0d0e21e 1960 {
6f1401dc 1961 NV value = SvNV_nomg(svr);
4efa5a16
RD
1962 (void)POPs;
1963
28e5dec8
JH
1964 if (!useleft) {
1965 /* left operand is undef, treat as zero - value */
1966 SETn(-value);
1967 RETURN;
1968 }
6f1401dc 1969 SETn( SvNV_nomg(svl) - value );
28e5dec8 1970 RETURN;
79072805 1971 }
a0d0e21e 1972}
79072805 1973
b3498293
JH
1974#define IV_BITS (IVSIZE * 8)
1975
1976static UV S_uv_shift(UV uv, int shift, bool left)
1977{
1978 if (shift < 0) {
1979 shift = -shift;
1980 left = !left;
1981 }
bae047b6 1982 if (UNLIKELY(shift >= IV_BITS)) {
b3498293
JH
1983 return 0;
1984 }
1985 return left ? uv << shift : uv >> shift;
1986}
1987
1988static IV S_iv_shift(IV iv, int shift, bool left)
1989{
190e86d7
KW
1990 if (shift < 0) {
1991 shift = -shift;
1992 left = !left;
1993 }
814735a3 1994
bae047b6 1995 if (UNLIKELY(shift >= IV_BITS)) {
190e86d7
KW
1996 return iv < 0 && !left ? -1 : 0;
1997 }
1998
814735a3
KW
1999 /* For left shifts, perl 5 has chosen to treat the value as unsigned for
2000 * the * purposes of shifting, then cast back to signed. This is very
17b35041 2001 * different from Raku:
814735a3 2002 *
17b35041 2003 * $ raku -e 'say -2 +< 5'
814735a3
KW
2004 * -64
2005 *
2006 * $ ./perl -le 'print -2 << 5'
2007 * 18446744073709551552
2008 * */
2009 if (left) {
2010 if (iv == IV_MIN) { /* Casting this to a UV is undefined behavior */
2011 return 0;
2012 }
2013 return (IV) (((UV) iv) << shift);
2014 }
2015
2016 /* Here is right shift */
2017 return iv >> shift;
b3498293
JH
2018}
2019
2020#define UV_LEFT_SHIFT(uv, shift) S_uv_shift(uv, shift, TRUE)
2021#define UV_RIGHT_SHIFT(uv, shift) S_uv_shift(uv, shift, FALSE)
2022#define IV_LEFT_SHIFT(iv, shift) S_iv_shift(iv, shift, TRUE)
2023#define IV_RIGHT_SHIFT(iv, shift) S_iv_shift(iv, shift, FALSE)
2024
a0d0e21e
LW
2025PP(pp_left_shift)
2026{
20b7effb 2027 dSP; dATARGET; SV *svl, *svr;
a42d0242 2028 tryAMAGICbin_MG(lshift_amg, AMGf_assign|AMGf_numeric);
6f1401dc
DM
2029 svr = POPs;
2030 svl = TOPs;
a0d0e21e 2031 {
6f1401dc 2032 const IV shift = SvIV_nomg(svr);
d0ba1bd2 2033 if (PL_op->op_private & HINT_INTEGER) {
b3498293 2034 SETi(IV_LEFT_SHIFT(SvIV_nomg(svl), shift));
d0ba1bd2
JH
2035 }
2036 else {
b3498293 2037 SETu(UV_LEFT_SHIFT(SvUV_nomg(svl), shift));
d0ba1bd2 2038 }
55497cff 2039 RETURN;
79072805 2040 }
a0d0e21e 2041}
79072805 2042
a0d0e21e
LW
2043PP(pp_right_shift)
2044{
20b7effb 2045 dSP; dATARGET; SV *svl, *svr;
a42d0242 2046 tryAMAGICbin_MG(rshift_amg, AMGf_assign|AMGf_numeric);
6f1401dc
DM
2047 svr = POPs;
2048 svl = TOPs;
a0d0e21e 2049 {
6f1401dc 2050 const IV shift = SvIV_nomg(svr);
d0ba1bd2 2051 if (PL_op->op_private & HINT_INTEGER) {
b3498293 2052 SETi(IV_RIGHT_SHIFT(SvIV_nomg(svl), shift));
d0ba1bd2
JH
2053 }
2054 else {
b3498293 2055 SETu(UV_RIGHT_SHIFT(SvUV_nomg(svl), shift));
d0ba1bd2 2056 }
a0d0e21e 2057 RETURN;
93a17b20 2058 }
79072805
LW
2059}
2060
a0d0e21e 2061PP(pp_lt)
79072805 2062{
20b7effb 2063 dSP;
33efebe6
DM
2064 SV *left, *right;
2065
0872de45 2066 tryAMAGICbin_MG(lt_amg, AMGf_numeric);
33efebe6
DM
2067 right = POPs;
2068 left = TOPs;
2069 SETs(boolSV(
2070 (SvIOK_notUV(left) && SvIOK_notUV(right))
2071 ? (SvIVX(left) < SvIVX(right))
2072 : (do_ncmp(left, right) == -1)
2073 ));
2074 RETURN;
a0d0e21e 2075}
79072805 2076
a0d0e21e
LW
2077PP(pp_gt)
2078{
20b7effb 2079 dSP;
33efebe6 2080 SV *left, *right;
1b6737cc 2081
0872de45 2082 tryAMAGICbin_MG(gt_amg, AMGf_numeric);
33efebe6
DM
2083 right = POPs;
2084 left = TOPs;
2085 SETs(boolSV(
2086 (SvIOK_notUV(left) && SvIOK_notUV(right))
2087 ? (SvIVX(left) > SvIVX(right))
2088 : (do_ncmp(left, right) == 1)
2089 ));
2090 RETURN;
a0d0e21e
LW
2091}
2092
2093PP(pp_le)
2094{
20b7effb 2095 dSP;
33efebe6 2096 SV *left, *right;
1b6737cc 2097
0872de45 2098 tryAMAGICbin_MG(le_amg, AMGf_numeric);
33efebe6
DM
2099 right = POPs;
2100 left = TOPs;
2101 SETs(boolSV(
2102 (SvIOK_notUV(left) && SvIOK_notUV(right))
2103 ? (SvIVX(left) <= SvIVX(right))
2104 : (do_ncmp(left, right) <= 0)
2105 ));
2106 RETURN;
a0d0e21e
LW
2107}
2108
2109PP(pp_ge)
2110{
20b7effb 2111 dSP;
33efebe6
DM
2112 SV *left, *right;
2113
0872de45 2114 tryAMAGICbin_MG(ge_amg, AMGf_numeric);
33efebe6
DM
2115 right = POPs;
2116 left = TOPs;
2117 SETs(boolSV(
2118 (SvIOK_notUV(left) && SvIOK_notUV(right))
2119 ? (SvIVX(left) >= SvIVX(right))
2120 : ( (do_ncmp(left, right) & 2) == 0)
2121 ));
2122 RETURN;
2123}
1b6737cc 2124
33efebe6
DM
2125PP(pp_ne)
2126{
20b7effb 2127 dSP;
33efebe6
DM
2128 SV *left, *right;
2129
0872de45 2130 tryAMAGICbin_MG(ne_amg, AMGf_numeric);
33efebe6
DM
2131 right = POPs;
2132 left = TOPs;
2133 SETs(boolSV(
2134 (SvIOK_notUV(left) && SvIOK_notUV(right))
2135 ? (SvIVX(left) != SvIVX(right))
2136 : (do_ncmp(left, right) != 0)
2137 ));
2138 RETURN;
2139}
1b6737cc 2140
33efebe6
DM
2141/* compare left and right SVs. Returns:
2142 * -1: <
2143 * 0: ==
2144 * 1: >
2145 * 2: left or right was a NaN
2146 */
2147I32
2148Perl_do_ncmp(pTHX_ SV* const left, SV * const right)
2149{
33efebe6
DM
2150 PERL_ARGS_ASSERT_DO_NCMP;
2151#ifdef PERL_PRESERVE_IVUV
33efebe6 2152 /* Fortunately it seems NaN isn't IOK */
01f91bf2 2153 if (SvIV_please_nomg(right) && SvIV_please_nomg(left)) {
33efebe6
DM
2154 if (!SvUOK(left)) {
2155 const IV leftiv = SvIVX(left);
2156 if (!SvUOK(right)) {
2157 /* ## IV <=> IV ## */
2158 const IV rightiv = SvIVX(right);
2159 return (leftiv > rightiv) - (leftiv < rightiv);
28e5dec8 2160 }
33efebe6
DM
2161 /* ## IV <=> UV ## */
2162 if (leftiv < 0)
2163 /* As (b) is a UV, it's >=0, so it must be < */
2164 return -1;
2165 {
2166 const UV rightuv = SvUVX(right);
2167 return ((UV)leftiv > rightuv) - ((UV)leftiv < rightuv);
28e5dec8 2168 }
28e5dec8 2169 }
79072805 2170
33efebe6
DM
2171 if (SvUOK(right)) {
2172 /* ## UV <=> UV ## */
2173 const UV leftuv = SvUVX(left);
2174 const UV rightuv = SvUVX(right);
2175 return (leftuv > rightuv) - (leftuv < rightuv);
28e5dec8 2176 }
33efebe6
DM
2177 /* ## UV <=> IV ## */
2178 {
2179 const IV rightiv = SvIVX(right);
2180 if (rightiv < 0)
2181 /* As (a) is a UV, it's >=0, so it cannot be < */
2182 return 1;
2183 {
2184 const UV leftuv = SvUVX(left);
2185 return (leftuv > (UV)rightiv) - (leftuv < (UV)rightiv);
28e5dec8 2186 }
28e5dec8 2187 }
e5964223 2188 NOT_REACHED; /* NOTREACHED */
28e5dec8
JH
2189 }
2190#endif
a0d0e21e 2191 {
33efebe6
DM
2192 NV const rnv = SvNV_nomg(right);
2193 NV const lnv = SvNV_nomg(left);
2194
cab190d4 2195#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
33efebe6
DM
2196 if (Perl_isnan(lnv) || Perl_isnan(rnv)) {
2197 return 2;
2198 }
2199 return (lnv > rnv) - (lnv < rnv);
cab190d4 2200#else
33efebe6
DM
2201 if (lnv < rnv)
2202 return -1;
2203 if (lnv > rnv)
2204 return 1;
659c4b96 2205 if (lnv == rnv)
33efebe6
DM
2206 return 0;
2207 return 2;
cab190d4 2208#endif
a0d0e21e 2209 }
79072805
LW
2210}
2211
33efebe6 2212
a0d0e21e 2213PP(pp_ncmp)
79072805 2214{
20b7effb 2215 dSP;
33efebe6
DM
2216 SV *left, *right;
2217 I32 value;
a42d0242 2218 tryAMAGICbin_MG(ncmp_amg, AMGf_numeric);
33efebe6
DM
2219 right = POPs;
2220 left = TOPs;
2221 value = do_ncmp(left, right);
2222 if (value == 2) {
3280af22 2223 SETs(&PL_sv_undef);
79072805 2224 }
33efebe6
DM
2225 else {
2226 dTARGET;
2227 SETi(value);
2228 }
2229 RETURN;
a0d0e21e 2230}
79072805 2231
b1c05ba5
DM
2232
2233/* also used for: pp_sge() pp_sgt() pp_slt() */
2234
afd9910b 2235PP(pp_sle)
a0d0e21e 2236{
20b7effb 2237 dSP;
79072805 2238
afd9910b
NC
2239 int amg_type = sle_amg;
2240 int multiplier = 1;
2241 int rhs = 1;
79072805 2242
afd9910b
NC
2243 switch (PL_op->op_type) {
2244 case OP_SLT:
2245 amg_type = slt_amg;
2246 /* cmp < 0 */
2247 rhs = 0;
2248 break;
2249 case OP_SGT:
2250 amg_type = sgt_amg;
2251 /* cmp > 0 */
2252 multiplier = -1;
2253 rhs = 0;
2254 break;
2255 case OP_SGE:
2256 amg_type = sge_amg;
2257 /* cmp >= 0 */
2258 multiplier = -1;
2259 break;
79072805 2260 }
79072805 2261
0872de45 2262 tryAMAGICbin_MG(amg_type, 0);
a0d0e21e
LW
2263 {
2264 dPOPTOPssrl;
130c5df3 2265 const int cmp =
5778acb6 2266#ifdef USE_LOCALE_COLLATE
130c5df3
KW
2267 (IN_LC_RUNTIME(LC_COLLATE))
2268 ? sv_cmp_locale_flags(left, right, 0)
2269 :
2270#endif
2271 sv_cmp_flags(left, right, 0);
afd9910b 2272 SETs(boolSV(cmp * multiplier < rhs));
a0d0e21e
LW
2273 RETURN;
2274 }
2275}
79072805 2276
36477c24 2277PP(pp_seq)
2278{
20b7effb 2279 dSP;
0872de45 2280 tryAMAGICbin_MG(seq_amg, 0);
36477c24 2281 {
2282 dPOPTOPssrl;
078504b2 2283 SETs(boolSV(sv_eq_flags(left, right, 0)));
a0d0e21e
LW
2284 RETURN;
2285 }
2286}
79072805 2287
a0d0e21e 2288PP(pp_sne)
79072805 2289{
20b7effb 2290 dSP;
0872de45 2291 tryAMAGICbin_MG(sne_amg, 0);
a0d0e21e
LW
2292 {
2293 dPOPTOPssrl;
078504b2 2294 SETs(boolSV(!sv_eq_flags(left, right, 0)));
a0d0e21e 2295 RETURN;
463ee0b2 2296 }
79072805
LW
2297}
2298
a0d0e21e 2299PP(pp_scmp)
79072805 2300{
20b7effb 2301 dSP; dTARGET;
6f1401dc 2302 tryAMAGICbin_MG(scmp_amg, 0);
a0d0e21e
LW
2303 {
2304 dPOPTOPssrl;
130c5df3 2305 const int cmp =
5778acb6 2306#ifdef USE_LOCALE_COLLATE
130c5df3
KW
2307 (IN_LC_RUNTIME(LC_COLLATE))
2308 ? sv_cmp_locale_flags(left, right, 0)
2309 :
2310#endif
2311 sv_cmp_flags(left, right, 0);
bbce6d69 2312 SETi( cmp );
a0d0e21e
LW
2313 RETURN;
2314 }
2315}
79072805 2316
55497cff 2317PP(pp_bit_and)
2318{
20b7effb 2319 dSP; dATARGET;
6f1401dc 2320 tryAMAGICbin_MG(band_amg, AMGf_assign);
a0d0e21e
LW
2321 {
2322 dPOPTOPssrl;
4633a7c4 2323 if (SvNIOKp(left) || SvNIOKp(right)) {
b20c4ee1
FC
2324 const bool left_ro_nonnum = !SvNIOKp(left) && SvREADONLY(left);
2325 const bool right_ro_nonnum = !SvNIOKp(right) && SvREADONLY(right);
d0ba1bd2 2326 if (PL_op->op_private & HINT_INTEGER) {
1b6737cc 2327 const IV i = SvIV_nomg(left) & SvIV_nomg(right);
972b05a9 2328 SETi(i);
d0ba1bd2
JH
2329 }
2330 else {
1b6737cc 2331 const UV u = SvUV_nomg(left) & SvUV_nomg(right);
972b05a9 2332 SETu(u);
d0ba1bd2 2333 }
5ee80e13 2334 if (left_ro_nonnum && left != TARG) SvNIOK_off(left);
b20c4ee1 2335 if (right_ro_nonnum) SvNIOK_off(right);
a0d0e21e
LW
2336 }
2337 else {
533c011a 2338 do_vop(PL_op->op_type, TARG, left, right);
a0d0e21e
LW
2339 SETTARG;
2340 }
2341 RETURN;
2342 }
2343}
79072805 2344
5d01050a
FC
2345PP(pp_nbit_and)
2346{
2347 dSP;
636ac8fc 2348 tryAMAGICbin_MG(band_amg, AMGf_assign|AMGf_numarg);
5d01050a
FC
2349 {
2350 dATARGET; dPOPTOPssrl;
2351 if (PL_op->op_private & HINT_INTEGER) {
2352 const IV i = SvIV_nomg(left) & SvIV_nomg(right);
2353 SETi(i);
2354 }
2355 else {
2356 const UV u = SvUV_nomg(left) & SvUV_nomg(right);
2357 SETu(u);
2358 }
2359 }
2360 RETURN;
2361}
2362
2363PP(pp_sbit_and)
2364{
2365 dSP;
2366 tryAMAGICbin_MG(sband_amg, AMGf_assign);
2367 {
2368 dATARGET; dPOPTOPssrl;
2369 do_vop(OP_BIT_AND, TARG, left, right);
2370 RETSETTARG;
2371 }
2372}
b1c05ba5
DM
2373
2374/* also used for: pp_bit_xor() */
2375
a0d0e21e
LW
2376PP(pp_bit_or)
2377{
20b7effb 2378 dSP; dATARGET;
3658c1f1
NC
2379 const int op_type = PL_op->op_type;
2380
6f1401dc 2381 tryAMAGICbin_MG((op_type == OP_BIT_OR ? bor_amg : bxor_amg), AMGf_assign);
a0d0e21e
LW
2382 {
2383 dPOPTOPssrl;
4633a7c4 2384 if (SvNIOKp(left) || SvNIOKp(right)) {
b20c4ee1
FC
2385 const bool left_ro_nonnum = !SvNIOKp(left) && SvREADONLY(left);
2386 const bool right_ro_nonnum = !SvNIOKp(right) && SvREADONLY(right);
d0ba1bd2 2387 if (PL_op->op_private & HINT_INTEGER) {
3658c1f1
NC
2388 const IV l = (USE_LEFT(left) ? SvIV_nomg(left) : 0);
2389 const IV r = SvIV_nomg(right);
2390 const IV result = op_type == OP_BIT_OR ? (l | r) : (l ^ r);
2391 SETi(result);
d0ba1bd2
JH
2392 }
2393 else {
3658c1f1
NC
2394 const UV l = (USE_LEFT(left) ? SvUV_nomg(left) : 0);
2395 const UV r = SvUV_nomg(right);
2396 const UV result = op_type == OP_BIT_OR ? (l | r) : (l ^ r);
2397 SETu(result);
d0ba1bd2 2398 }
5ee80e13 2399 if (left_ro_nonnum && left != TARG) SvNIOK_off(left);
b20c4ee1 2400 if (right_ro_nonnum) SvNIOK_off(right);
a0d0e21e
LW
2401 }
2402 else {
3658c1f1 2403 do_vop(op_type, TARG, left, right);
a0d0e21e
LW
2404 SETTARG;
2405 }
2406 RETURN;
79072805 2407 }
a0d0e21e 2408}
79072805 2409
5d01050a
FC
2410/* also used for: pp_nbit_xor() */
2411
2412PP(pp_nbit_or)
2413{
2414 dSP;
2415 const int op_type = PL_op->op_type;
2416
2417 tryAMAGICbin_MG((op_type == OP_NBIT_OR ? bor_amg : bxor_amg),
636ac8fc 2418 AMGf_assign|AMGf_numarg);
5d01050a
FC
2419 {
2420 dATARGET; dPOPTOPssrl;
2421 if (PL_op->op_private & HINT_INTEGER) {
2422 const IV l = (USE_LEFT(left) ? SvIV_nomg(left) : 0);
2423 const IV r = SvIV_nomg(right);
2424 const IV result = op_type == OP_NBIT_OR ? (l | r) : (l ^ r);
2425 SETi(result);
2426 }
2427 else {
2428 const UV l = (USE_LEFT(left) ? SvUV_nomg(left) : 0);
2429 const UV r = SvUV_nomg(right);
2430 const UV result = op_type == OP_NBIT_OR ? (l | r) : (l ^ r);
2431 SETu(result);
2432 }
2433 }
2434 RETURN;
2435}
2436
2437/* also used for: pp_sbit_xor() */
2438
2439PP(pp_sbit_or)
2440{
2441 dSP;
2442 const int op_type = PL_op->op_type;
2443
2444 tryAMAGICbin_MG((op_type == OP_SBIT_OR ? sbor_amg : sbxor_amg),
2445 AMGf_assign);
2446 {
2447 dATARGET; dPOPTOPssrl;
2448 do_vop(op_type == OP_SBIT_OR ? OP_BIT_OR : OP_BIT_XOR, TARG, left,
2449 right);
2450 RETSETTARG;
2451 }
2452}
2453
1c2b3fd6
FC
2454PERL_STATIC_INLINE bool
2455S_negate_string(pTHX)
2456{
2457 dTARGET; dSP;
2458 STRLEN len;
2459 const char *s;
2460 SV * const sv = TOPs;
2461 if (!SvPOKp(sv) || SvNIOK(sv) || (!SvPOK(sv) && SvNIOKp(sv)))
2462 return FALSE;
2463 s = SvPV_nomg_const(sv, len);
2464 if (isIDFIRST(*s)) {
2465 sv_setpvs(TARG, "-");
2466 sv_catsv(TARG, sv);
2467 }
2468 else if (*s == '+' || (*s == '-' && !looks_like_number(sv))) {
2469 sv_setsv_nomg(TARG, sv);
2470 *SvPV_force_nomg(TARG, len) = *s == '-' ? '+' : '-';
2471 }
2472 else return FALSE;
245d035e 2473 SETTARG;
1c2b3fd6
FC
2474 return TRUE;
2475}
2476
a0d0e21e
LW
2477PP(pp_negate)
2478{
20b7effb 2479 dSP; dTARGET;
6f1401dc 2480 tryAMAGICun_MG(neg_amg, AMGf_numeric);
1c2b3fd6 2481 if (S_negate_string(aTHX)) return NORMAL;
a0d0e21e 2482 {
6f1401dc 2483 SV * const sv = TOPs;
a5b92898 2484
d96ab1b5 2485 if (SvIOK(sv)) {
7dbe3150 2486 /* It's publicly an integer */
28e5dec8 2487 oops_its_an_int:
9b0e499b
GS
2488 if (SvIsUV(sv)) {
2489 if (SvIVX(sv) == IV_MIN) {
28e5dec8 2490 /* 2s complement assumption. */
d14578b8
KW
2491 SETi(SvIVX(sv)); /* special case: -((UV)IV_MAX+1) ==
2492 IV_MIN */
245d035e 2493 return NORMAL;
9b0e499b
GS
2494 }
2495 else if (SvUVX(sv) <= IV_MAX) {
beccb14c 2496 SETi(-SvIVX(sv));
245d035e 2497 return NORMAL;
9b0e499b
GS
2498 }
2499 }
2500 else if (SvIVX(sv) != IV_MIN) {
2501 SETi(-SvIVX(sv));
245d035e 2502 return NORMAL;
9b0e499b 2503 }
28e5dec8
JH
2504#ifdef PERL_PRESERVE_IVUV
2505 else {
2506 SETu((UV)IV_MIN);
245d035e 2507 return NORMAL;
28e5dec8
JH
2508 }
2509#endif
9b0e499b 2510 }
8a5decd8 2511 if (SvNIOKp(sv) && (SvNIOK(sv) || !SvPOK(sv)))
6f1401dc 2512 SETn(-SvNV_nomg(sv));
1c2b3fd6 2513 else if (SvPOKp(sv) && SvIV_please_nomg(sv))
8eb28a70 2514 goto oops_its_an_int;
4633a7c4 2515 else
6f1401dc 2516 SETn(-SvNV_nomg(sv));
79072805 2517 }
245d035e 2518 return NORMAL;
79072805
LW
2519}
2520
a0d0e21e 2521PP(pp_not)
79072805 2522{
20b7effb 2523 dSP;
f4c975aa
DM
2524 SV *sv;
2525
0872de45 2526 tryAMAGICun_MG(not_amg, 0);
f4c975aa
DM
2527 sv = *PL_stack_sp;
2528 *PL_stack_sp = boolSV(!SvTRUE_nomg_NN(sv));
a0d0e21e 2529 return NORMAL;
79072805
LW
2530}
2531
5d01050a
FC
2532static void
2533S_scomplement(pTHX_ SV *targ, SV *sv)
79072805 2534{
eb578fdb
KW
2535 U8 *tmps;
2536 I32 anum;
a0d0e21e
LW
2537 STRLEN len;
2538
85b0ee6e
FC
2539 sv_copypv_nomg(TARG, sv);
2540 tmps = (U8*)SvPV_nomg(TARG, len);
08b6664b 2541
1d68d6cd 2542 if (SvUTF8(TARG)) {
08b6664b 2543 if (len && ! utf8_to_bytes(tmps, &len)) {
814eedc8 2544 Perl_croak(aTHX_ FATAL_ABOVE_FF_MSG, PL_op_desc[PL_op->op_type]);
08b6664b 2545 }
2324bdb9 2546 SvCUR_set(TARG, len);
08b6664b
KW
2547 SvUTF8_off(TARG);
2548 }
2549
2550 anum = len;
1d68d6cd 2551
51723571 2552 {
eb578fdb 2553 long *tmpl;
d398c6bf 2554 for ( ; anum && PTR2nat(tmps) % sizeof(long); anum--, tmps++)
51723571
JH
2555 *tmps = ~*tmps;
2556 tmpl = (long*)tmps;
bb7a0f54 2557 for ( ; anum >= (I32)sizeof(long); anum -= (I32)sizeof(long), tmpl++)
51723571
JH
2558 *tmpl = ~*tmpl;
2559 tmps = (U8*)tmpl;
2560 }
17d44595 2561
a0d0e21e
LW
2562 for ( ; anum > 0; anum--, tmps++)
2563 *tmps = ~*tmps;
5d01050a
FC
2564}
2565
2566PP(pp_complement)
2567{
2568 dSP; dTARGET;
2569 tryAMAGICun_MG(compl_amg, AMGf_numeric);
2570 {
2571 dTOPss;
2572 if (SvNIOKp(sv)) {
2573 if (PL_op->op_private & HINT_INTEGER) {
2574 const IV i = ~SvIV_nomg(sv);
2575 SETi(i);
2576 }
2577 else {
2578 const UV u = ~SvUV_nomg(sv);
2579 SETu(u);
2580 }
2581 }
2582 else {
2583 S_scomplement(aTHX_ TARG, sv);
ec93b65f 2584 SETTARG;
a0d0e21e 2585 }
24840750 2586 return NORMAL;
5d01050a
FC
2587 }
2588}
2589
2590PP(pp_ncomplement)
2591{
2592 dSP;
636ac8fc 2593 tryAMAGICun_MG(compl_amg, AMGf_numeric|AMGf_numarg);
5d01050a
FC
2594 {
2595 dTARGET; dTOPss;
2596 if (PL_op->op_private & HINT_INTEGER) {
2597 const IV i = ~SvIV_nomg(sv);
2598 SETi(i);
2599 }
2600 else {
2601 const UV u = ~SvUV_nomg(sv);
2602 SETu(u);
2603 }
2604 }
2605 return NORMAL;
2606}
2607
2608PP(pp_scomplement)
2609{
2610 dSP;
2611 tryAMAGICun_MG(scompl_amg, AMGf_numeric);
2612 {
2613 dTARGET; dTOPss;
2614 S_scomplement(aTHX_ TARG, sv);
2615 SETTARG;
2616 return NORMAL;
a0d0e21e 2617 }
79072805
LW
2618}
2619
a0d0e21e
LW
2620/* integer versions of some of the above */
2621
a0d0e21e 2622PP(pp_i_multiply)
79072805 2623{
20b7effb 2624 dSP; dATARGET;
6f1401dc 2625 tryAMAGICbin_MG(mult_amg, AMGf_assign);
a0d0e21e 2626 {
6f1401dc 2627 dPOPTOPiirl_nomg;
a0d0e21e
LW
2628 SETi( left * right );
2629 RETURN;
2630 }
79072805
LW
2631}
2632
a0d0e21e 2633PP(pp_i_divide)
79072805 2634{
85935d8e 2635 IV num;
20b7effb 2636 dSP; dATARGET;
6f1401dc 2637 tryAMAGICbin_MG(div_amg, AMGf_assign);
a0d0e21e 2638 {
6f1401dc 2639 dPOPTOPssrl;
85935d8e 2640 IV value = SvIV_nomg(right);
a0d0e21e 2641 if (value == 0)
ece1bcef 2642 DIE(aTHX_ "Illegal division by zero");
85935d8e 2643 num = SvIV_nomg(left);
a0cec769
YST
2644
2645 /* avoid FPE_INTOVF on some platforms when num is IV_MIN */
2646 if (value == -1)
2647 value = - num;
2648 else
2649 value = num / value;
6f1401dc 2650 SETi(value);
a0d0e21e
LW
2651 RETURN;
2652 }
79072805
LW
2653}
2654
befad5d1 2655PP(pp_i_modulo)
224ec323 2656{
20b7effb 2657 dSP; dATARGET;
6f1401dc 2658 tryAMAGICbin_MG(modulo_amg, AMGf_assign);
224ec323 2659 {
6f1401dc 2660 dPOPTOPiirl_nomg;
224ec323
JH
2661 if (!right)
2662 DIE(aTHX_ "Illegal modulus zero");
a0cec769
YST
2663 /* avoid FPE_INTOVF on some platforms when left is IV_MIN */
2664 if (right == -1)
2665 SETi( 0 );
2666 else
2667 SETi( left % right );
224ec323
JH
2668 RETURN;
2669 }
2670}
2671
a0d0e21e 2672PP(pp_i_add)
79072805 2673{
20b7effb 2674 dSP; dATARGET;
6f1401dc 2675 tryAMAGICbin_MG(add_amg, AMGf_assign);
a0d0e21e 2676 {
6f1401dc 2677 dPOPTOPiirl_ul_nomg;
a0d0e21e
LW
2678 SETi( left + right );
2679 RETURN;
79072805 2680 }
79072805
LW
2681}
2682
a0d0e21e 2683PP(pp_i_subtract)
79072805 2684{
20b7effb 2685 dSP; dATARGET;
6f1401dc 2686 tryAMAGICbin_MG(subtr_amg, AMGf_assign);
a0d0e21e 2687 {
6f1401dc 2688 dPOPTOPiirl_ul_nomg;
a0d0e21e
LW
2689 SETi( left - right );
2690 RETURN;
79072805 2691 }
79072805
LW
2692}
2693
a0d0e21e 2694PP(pp_i_lt)
79072805 2695{
20b7effb 2696 dSP;
0872de45 2697 tryAMAGICbin_MG(lt_amg, 0);
a0d0e21e 2698 {
96b6b87f 2699 dPOPTOPiirl_nomg;
54310121 2700 SETs(boolSV(left < right));
a0d0e21e
LW
2701 RETURN;
2702 }
79072805
LW
2703}
2704
a0d0e21e 2705PP(pp_i_gt)
79072805 2706{
20b7effb 2707 dSP;
0872de45 2708 tryAMAGICbin_MG(gt_amg, 0);
a0d0e21e 2709 {
96b6b87f 2710 dPOPTOPiirl_nomg;
54310121 2711 SETs(boolSV(left > right));
a0d0e21e
LW
2712 RETURN;
2713 }
79072805
LW
2714}
2715
a0d0e21e 2716PP(pp_i_le)
79072805 2717{
20b7effb 2718 dSP;
0872de45 2719 tryAMAGICbin_MG(le_amg, 0);
a0d0e21e 2720 {
96b6b87f 2721 dPOPTOPiirl_nomg;
54310121 2722 SETs(boolSV(left <= right));
a0d0e21e 2723 RETURN;
85e6fe83 2724 }
79072805
LW
2725}
2726
a0d0e21e 2727PP(pp_i_ge)
79072805 2728{
20b7effb 2729 dSP;
0872de45 2730 tryAMAGICbin_MG(ge_amg, 0);
a0d0e21e 2731 {
96b6b87f 2732 dPOPTOPiirl_nomg;
54310121 2733 SETs(boolSV(left >= right));
a0d0e21e
LW
2734 RETURN;
2735 }
79072805
LW
2736}
2737
a0d0e21e 2738PP(pp_i_eq)
79072805 2739{
20b7effb 2740 dSP;
0872de45 2741 tryAMAGICbin_MG(eq_amg, 0);
a0d0e21e 2742 {
96b6b87f 2743 dPOPTOPiirl_nomg;
54310121 2744 SETs(boolSV(left == right));
a0d0e21e
LW
2745 RETURN;
2746 }
79072805
LW
2747}
2748
a0d0e21e 2749PP(pp_i_ne)
79072805 2750{
20b7effb 2751 dSP;
0872de45 2752 tryAMAGICbin_MG(ne_amg, 0);
a0d0e21e 2753 {
96b6b87f 2754 dPOPTOPiirl_nomg;
54310121 2755 SETs(boolSV(left != right));
a0d0e21e
LW
2756 RETURN;
2757 }
79072805
LW
2758}
2759
a0d0e21e 2760PP(pp_i_ncmp)
79072805 2761{
20b7effb 2762 dSP; dTARGET;
6f1401dc 2763 tryAMAGICbin_MG(ncmp_amg, 0);
a0d0e21e 2764 {
96b6b87f 2765 dPOPTOPiirl_nomg;
a0d0e21e 2766 I32 value;
79072805 2767
a0d0e21e 2768 if (left > right)
79072805 2769 value = 1;
a0d0e21e 2770 else if (left < right)
79072805 2771 value = -1;
a0d0e21e 2772 else
79072805 2773 value = 0;
a0d0e21e
LW
2774 SETi(value);
2775 RETURN;
79072805 2776 }
85e6fe83
LW
2777}
2778
2779PP(pp_i_negate)
2780{
20b7effb 2781 dSP; dTARGET;
6f1401dc 2782 tryAMAGICun_MG(neg_amg, 0);
1c2b3fd6 2783 if (S_negate_string(aTHX)) return NORMAL;
6f1401dc
DM
2784 {
2785 SV * const sv = TOPs;
2786 IV const i = SvIV_nomg(sv);
2787 SETi(-i);
ae642386 2788 return NORMAL;
6f1401dc 2789 }
85e6fe83
LW
2790}
2791
79072805
LW
2792/* High falutin' math. */
2793
2794PP(pp_atan2)
2795{
20b7effb 2796 dSP; dTARGET;
6f1401dc 2797 tryAMAGICbin_MG(atan2_amg, 0);
a0d0e21e 2798 {
096c060c 2799 dPOPTOPnnrl_nomg;
a1021d57 2800 SETn(Perl_atan2(left, right));
a0d0e21e
LW
2801 RETURN;
2802 }
79072805
LW
2803}
2804
b1c05ba5
DM
2805
2806/* also used for: pp_cos() pp_exp() pp_log() pp_sqrt() */
2807
79072805
LW
2808PP(pp_sin)
2809{
20b7effb 2810 dSP; dTARGET;
af71714e 2811 int amg_type = fallback_amg;
71302fe3 2812 const char *neg_report = NULL;
71302fe3
NC
2813 const int op_type = PL_op->op_type;
2814
2815 switch (op_type) {
af71714e
JH
2816 case OP_SIN: amg_type = sin_amg; break;
2817 case OP_COS: amg_type = cos_amg; break;
2818 case OP_EXP: amg_type = exp_amg; break;
2819 case OP_LOG: amg_type = log_amg; neg_report = "log"; break;
2820 case OP_SQRT: amg_type = sqrt_amg; neg_report = "sqrt"; break;
a0d0e21e 2821 }
79072805 2822
af71714e 2823 assert(amg_type != fallback_amg);
6f1401dc
DM
2824
2825 tryAMAGICun_MG(amg_type, 0);
a0d0e21e 2826 {
8c78ed36 2827 SV * const arg = TOPs;
6f1401dc 2828 const NV value = SvNV_nomg(arg);
a5dc2484 2829#ifdef NV_NAN
f256868e 2830 NV result = NV_NAN;
a5dc2484
JH
2831#else
2832 NV result = 0.0;
2833#endif
af71714e 2834 if (neg_report) { /* log or sqrt */
a3463d96
DD
2835 if (
2836#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2837 ! Perl_isnan(value) &&
2838#endif
2839 (op_type == OP_LOG ? (value <= 0.0) : (value < 0.0))) {
71302fe3 2840 SET_NUMERIC_STANDARD();
dcbac5bb 2841 /* diag_listed_as: Can't take log of %g */
147e3846 2842 DIE(aTHX_ "Can't take %s of %" NVgf, neg_report, value);
71302fe3
NC
2843 }
2844 }
af71714e 2845 switch (op_type) {
f256868e 2846 default:
af71714e
JH
2847 case OP_SIN: result = Perl_sin(value); break;
2848 case OP_COS: result = Perl_cos(value); break;
2849 case OP_EXP: result = Perl_exp(value); break;
2850 case OP_LOG: result = Perl_log(value); break;
2851 case OP_SQRT: result = Perl_sqrt(value); break;
2852 }
8c78ed36
FC
2853 SETn(result);
2854 return NORMAL;
a0d0e21e 2855 }
79072805
LW
2856}
2857
56cb0a1c
AD
2858/* Support Configure command-line overrides for rand() functions.
2859 After 5.005, perhaps we should replace this by Configure support
2860 for drand48(), random(), or rand(). For 5.005, though, maintain
2861 compatibility by calling rand() but allow the user to override it.
2862 See INSTALL for details. --Andy Dougherty 15 July 1998
2863*/
85ab1d1d
JH
2864/* Now it's after 5.005, and Configure supports drand48() and random(),
2865 in addition to rand(). So the overrides should not be needed any more.
2866 --Jarkko Hietaniemi 27 September 1998
2867 */
2868
79072805
LW
2869PP(pp_rand)
2870{
80252599 2871 if (!PL_srand_called) {
85ab1d1d 2872 (void)seedDrand01((Rand_seed_t)seed());
80252599 2873 PL_srand_called = TRUE;
93dc8474 2874 }
fdf4dddd
DD
2875 {
2876 dSP;
2877 NV value;
a8e41ef4 2878
fdf4dddd 2879 if (MAXARG < 1)
7e9044f9
FC
2880 {
2881 EXTEND(SP, 1);
fdf4dddd 2882 value = 1.0;
7e9044f9 2883 }
fdf4dddd
DD
2884 else {
2885 SV * const sv = POPs;
2886 if(!sv)
2887 value = 1.0;
2888 else
2889 value = SvNV(sv);
2890 }
2891 /* 1 of 2 things can be carried through SvNV, SP or TARG, SP was carried */
a3463d96
DD
2892#if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2893 if (! Perl_isnan(value) && value == 0.0)
2894#else
659c4b96 2895 if (value == 0.0)
a3463d96 2896#endif
fdf4dddd
DD
2897 value = 1.0;
2898 {
2899 dTARGET;
2900 PUSHs(TARG);
2901 PUTBACK;
2902 value *= Drand01();
2903 sv_setnv_mg(TARG, value);
2904 }
2905 }
2906 return NORMAL;
79072805
LW
2907}
2908
2909PP(pp_srand)
2910{
20b7effb 2911 dSP; dTARGET;
f914a682
JL
2912 UV anum;
2913
0a5f3363 2914 if (MAXARG >= 1 && (TOPs || POPs)) {
f914a682
JL
2915 SV *top;
2916 char *pv;
2917 STRLEN len;
2918 int flags;
2919
2920 top = POPs;
2921 pv = SvPV(top, len);
2922 flags = grok_number(pv, len, &anum);
2923
2924 if (!(flags & IS_NUMBER_IN_UV)) {
2925 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
2926 "Integer overflow in srand");
2927 anum = UV_MAX;
2928 }
2929 }
2930 else {
2931 anum = seed();
2932 }
2933
85ab1d1d 2934 (void)seedDrand01((Rand_seed_t)anum);
80252599 2935 PL_srand_called = TRUE;
da1010ec
NC
2936 if (anum)
2937 XPUSHu(anum);
2938 else {
2939 /* Historically srand always returned true. We can avoid breaking
2940 that like this: */
2941 sv_setpvs(TARG, "0 but true");
2942 XPUSHTARG;
2943 }
83832992 2944 RETURN;
79072805
LW
2945}
2946
79072805
LW
2947PP(pp_int)
2948{
20b7effb 2949 dSP; dTARGET;
6f1401dc 2950 tryAMAGICun_MG(int_amg, AMGf_numeric);
774d564b 2951 {
6f1401dc
DM
2952 SV * const sv = TOPs;
2953 const IV iv = SvIV_nomg(sv);
28e5dec8
JH
2954 /* XXX it's arguable that compiler casting to IV might be subtly
2955 different from modf (for numbers inside (IV_MIN,UV_MAX)) in which
2956 else preferring IV has introduced a subtle behaviour change bug. OTOH
2957 relying on floating point to be accurate is a bug. */
2958
c781a409 2959 if (!SvOK(sv)) {
922c4365 2960 SETu(0);
c781a409
RD
2961 }
2962 else if (SvIOK(sv)) {
2963 if (SvIsUV(sv))
6f1401dc 2964 SETu(SvUV_nomg(sv));
c781a409 2965 else
28e5dec8 2966 SETi(iv);
c781a409 2967 }
c781a409 2968 else {
6f1401dc 2969 const NV value = SvNV_nomg(sv);
b9d05018
FC
2970 if (UNLIKELY(Perl_isinfnan(value)))
2971 SETn(value);
5bf8b78e 2972 else if (value >= 0.0) {
28e5dec8
JH
2973 if (value < (NV)UV_MAX + 0.5) {
2974 SETu(U_V(value));
2975 } else {
059a1014 2976 SETn(Perl_floor(value));
28e5dec8 2977 }
1048ea30 2978 }
28e5dec8
JH
2979 else {
2980 if (value > (NV)IV_MIN - 0.5) {
2981 SETi(I_V(value));
2982 } else {
1bbae031 2983 SETn(Perl_ceil(value));
28e5dec8
JH
2984 }
2985 }
774d564b 2986 }
79072805 2987 }
699e9491 2988 return NORMAL;
79072805
LW
2989}
2990
463ee0b2
LW
2991PP(pp_abs)
2992{
20b7effb 2993 dSP; dTARGET;
6f1401dc 2994 tryAMAGICun_MG(abs_amg, AMGf_numeric);
a0d0e21e 2995 {
6f1401dc 2996 SV * const sv = TOPs;
28e5dec8 2997 /* This will cache the NV value if string isn't actually integer */
6f1401dc 2998 const IV iv = SvIV_nomg(sv);
a227d84d 2999
800401ee 3000 if (!SvOK(sv)) {
922c4365 3001 SETu(0);
800401ee
JH
3002 }
3003 else if (SvIOK(sv)) {
28e5dec8 3004 /* IVX is precise */
800401ee 3005 if (SvIsUV(sv)) {
6f1401dc 3006 SETu(SvUV_nomg(sv)); /* force it to be numeric only */
28e5dec8
JH
3007 } else {
3008 if (iv >= 0) {
3009 SETi(iv);
3010 } else {
3011 if (iv != IV_MIN) {
3012 SETi(-iv);
3013 } else {
3014 /* 2s complement assumption. Also, not really needed as
3015 IV_MIN and -IV_MIN should both be %100...00 and NV-able */
b396d0d8 3016 SETu((UV)IV_MIN);
28e5dec8 3017 }
a227d84d 3018 }
28e5dec8
JH
3019 }
3020 } else{
6f1401dc 3021 const NV value = SvNV_nomg(sv);
774d564b 3022 if (value < 0.0)
1b6737cc 3023 SETn(-value);
a4474c9e
DD
3024 else
3025 SETn(value);
774d564b 3026 }
a0d0e21e 3027 }
067b7929 3028 return NORMAL;
463ee0b2
LW
3029}
3030
b1c05ba5
DM
3031
3032/* also used for: pp_hex() */
3033
79072805
LW
3034PP(pp_oct)
3035{
20b7effb 3036 dSP; dTARGET;
5c144d81 3037 const char *tmps;
53305cf1 3038 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES;
6f894ead 3039 STRLEN len;
53305cf1
NC
3040 NV result_nv;
3041 UV result_uv;
4e51bcca 3042 SV* const sv = TOPs;
79072805 3043
349d4f2f 3044 tmps = (SvPV_const(sv, len));
2bc69dc4
NIS
3045 if (DO_UTF8(sv)) {
3046 /* If Unicode, try to downgrade
3047 * If not possible, croak. */
1b6737cc 3048 SV* const tsv = sv_2mortal(newSVsv(sv));
a8e41ef4 3049
2bc69dc4
NIS
3050 SvUTF8_on(tsv);
3051 sv_utf8_downgrade(tsv, FALSE);
349d4f2f 3052 tmps = SvPV_const(tsv, len);
2bc69dc4 3053 }
daa2adfd
NC
3054 if (PL_op->op_type == OP_HEX)
3055 goto hex;
3056
6f894ead 3057 while (*tmps && len && isSPACE(*tmps))
53305cf1 3058 tmps++, len--;
9e24b6e2 3059 if (*tmps == '0')
53305cf1 3060 tmps++, len--;
305b8651 3061 if (isALPHA_FOLD_EQ(*tmps, 'x')) {
c969ff22
KW
3062 tmps++, len--;
3063 flags |= PERL_SCAN_DISALLOW_PREFIX;
daa2adfd 3064 hex:
53305cf1 3065 result_uv = grok_hex (tmps, &len, &flags, &result_nv);
daa2adfd 3066 }
c969ff22
KW
3067 else if (isALPHA_FOLD_EQ(*tmps, 'b')) {
3068 tmps++, len--;
3069 flags |= PERL_SCAN_DISALLOW_PREFIX;
53305cf1 3070 result_uv = grok_bin (tmps, &len, &flags, &result_nv);
c969ff22 3071 }
464e2e8a 3072 else
53305cf1
NC
3073 result_uv = grok_oct (tmps, &len, &flags, &result_nv);
3074
3075 if (flags & PERL_SCAN_GREATER_THAN_UV_MAX) {
4e51bcca 3076 SETn(result_nv);
53305cf1
NC
3077 }
3078 else {
4e51bcca 3079 SETu(result_uv);
53305cf1 3080 }
4e51bcca 3081 return NORMAL;
79072805
LW
3082}
3083
3084/* String stuff. */
3085
5febd2ff 3086
79072805
LW
3087PP(pp_length)
3088{
20b7effb 3089 dSP; dTARGET;
0bd48802 3090 SV * const sv = TOPs;
a0ed51b3 3091
7776003e 3092 U32 in_bytes = IN_BYTES;
5febd2ff
DM
3093 /* Simplest case shortcut:
3094 * set svflags to just the SVf_POK|SVs_GMG|SVf_UTF8 from the SV,
3095 * with the SVf_UTF8 flag inverted if under 'use bytes' (HINT_BYTES
3096 * set)
3097 */
7776003e 3098 U32 svflags = (SvFLAGS(sv) ^ (in_bytes << 26)) & (SVf_POK|SVs_GMG|SVf_UTF8);
5febd2ff
DM
3099
3100 STATIC_ASSERT_STMT(SVf_UTF8 == (HINT_BYTES << 26));
7776003e
DD
3101 SETs(TARG);
3102
5febd2ff 3103 if (LIKELY(svflags == SVf_POK))
7776003e 3104 goto simple_pv;
5febd2ff
DM
3105
3106 if (svflags & SVs_GMG)
7776003e 3107 mg_get(sv);
5febd2ff 3108
0f43fd57 3109 if (SvOK(sv)) {
5b750817 3110 STRLEN len;
f446eca7
DM
3111 if (!IN_BYTES) { /* reread to avoid using an C auto/register */
3112 if ((SvFLAGS(sv) & (SVf_POK|SVf_UTF8)) == SVf_POK)
3113 goto simple_pv;
7b394f12
DM
3114 if ( SvPOK(sv) && (PL_op->op_private & OPpTRUEBOOL)) {
3115 /* no need to convert from bytes to chars */
3116 len = SvCUR(sv);
3117 goto return_bool;
3118 }
5b750817 3119 len = sv_len_utf8_nomg(sv);
f446eca7 3120 }
5febd2ff 3121 else {
7776003e 3122 /* unrolled SvPV_nomg_const(sv,len) */
5febd2ff
DM
3123 if (SvPOK_nog(sv)) {
3124 simple_pv:
7776003e 3125 len = SvCUR(sv);
7b394f12
DM
3126 if (PL_op->op_private & OPpTRUEBOOL) {
3127 return_bool:
3128 SETs(len ? &PL_sv_yes : &PL_sv_zero);
3129 return NORMAL;
3130 }
5febd2ff
DM
3131 }
3132 else {
7776003e
DD
3133 (void)sv_2pv_flags(sv, &len, 0|SV_CONST_RETURN);
3134 }
0f43fd57 3135 }
5b750817 3136 TARGi((IV)(len), 1);
5febd2ff
DM
3137 }
3138 else {
9407f9c1 3139 if (!SvPADTMP(TARG)) {
5febd2ff 3140 /* OPpTARGET_MY: targ is var in '$lex = length()' */
e03e82a0 3141 sv_set_undef(TARG);
5b750817 3142 SvSETMAGIC(TARG);
5febd2ff
DM
3143 }
3144 else
3145 /* TARG is on stack at this point and is overwriten by SETs.
3146 * This branch is the odd one out, so put TARG by default on
3147 * stack earlier to let local SP go out of liveness sooner */
7776003e 3148 SETs(&PL_sv_undef);
92331800 3149 }
7776003e 3150 return NORMAL; /* no putback, SP didn't move in this opcode */
79072805
LW
3151}
3152
5febd2ff 3153
83f78d1a
FC
3154/* Returns false if substring is completely outside original string.
3155 No length is indicated by len_iv = 0 and len_is_uv = 0. len_is_uv must
3156 always be true for an explicit 0.
3157*/
3158bool
ddeaf645
DD
3159Perl_translate_substr_offsets( STRLEN curlen, IV pos1_iv,
3160 bool pos1_is_uv, IV len_iv,
3161 bool len_is_uv, STRLEN *posp,
3162 STRLEN *lenp)
83f78d1a
FC
3163{
3164 IV pos2_iv;
3165 int pos2_is_uv;
3166
3167 PERL_ARGS_ASSERT_TRANSLATE_SUBSTR_OFFSETS;
3168
3169 if (!pos1_is_uv && pos1_iv < 0 && curlen) {
3170 pos1_is_uv = curlen-1 > ~(UV)pos1_iv;
3171 pos1_iv += curlen;
3172 }
3173 if ((pos1_is_uv || pos1_iv > 0) && (UV)pos1_iv > curlen)
3174 return FALSE;
3175
3176 if (len_iv || len_is_uv) {
3177 if (!len_is_uv && len_iv < 0) {
3178 pos2_iv = curlen + len_iv;
3179 if (curlen)
3180 pos2_is_uv = curlen-1 > ~(UV)len_iv;
3181 else
3182 pos2_is_uv = 0;
3183 } else { /* len_iv >= 0 */
3184 if (!pos1_is_uv && pos1_iv < 0) {
3185 pos2_iv = pos1_iv + len_iv;
3186 pos2_is_uv = (UV)len_iv > (UV)IV_MAX;
3187 } else {
3188 if ((UV)len_iv > curlen-(UV)pos1_iv)
3189 pos2_iv = curlen;
3190 else
3191 pos2_iv = pos1_iv+len_iv;
3192 pos2_is_uv = 1;
3193 }
3194 }
3195 }
3196 else {
3197 pos2_iv = curlen;
3198 pos2_is_uv = 1;
3199 }
3200
3201 if (!pos2_is_uv && pos2_iv < 0) {
3202 if (!pos1_is_uv && pos1_iv < 0)
3203 return FALSE;
3204 pos2_iv = 0;
3205 }
3206 else if (!pos1_is_uv && pos1_iv < 0)
3207 pos1_iv = 0;
3208
3209 if ((UV)pos2_iv < (UV)pos1_iv)
3210 pos2_iv = pos1_iv;
3211 if ((UV)pos2_iv > curlen)
3212 pos2_iv = curlen;
3213
3214 /* pos1_iv and pos2_iv both in 0..curlen, so the cast is safe */
3215 *posp = (STRLEN)( (UV)pos1_iv );
3216 *lenp = (STRLEN)( (UV)pos2_iv - (UV)pos1_iv );
3217
3218 return TRUE;
3219}
3220
79072805
LW
3221PP(pp_substr)
3222{
20b7effb 3223 dSP; dTARGET;
79072805 3224 SV *sv;
463ee0b2 3225 STRLEN curlen;
9402d6ed 3226 STRLEN utf8_curlen;
777f7c56
EB
3227 SV * pos_sv;
3228 IV pos1_iv;
3229 int pos1_is_uv;
777f7c56
EB
3230 SV * len_sv;
3231 IV len_iv = 0;
83f78d1a 3232 int len_is_uv = 0;
24fcb59f 3233 I32 lvalue = PL_op->op_flags & OPf_MOD || LVRET;
bbddc9e0 3234 const bool rvalue = (GIMME_V != G_VOID);
e1ec3a88 3235 const char *tmps;
9402d6ed 3236 SV *repl_sv = NULL;
cbbf8932 3237 const char *repl = NULL;
7b8d334a 3238 STRLEN repl_len;
7bc95ae1 3239 int num_args = PL_op->op_private & 7;
13e30c65 3240 bool repl_need_utf8_upgrade = FALSE;
79072805 3241
78f9721b
SM
3242 if (num_args > 2) {
3243 if (num_args > 3) {
24fcb59f 3244 if(!(repl_sv = POPs)) num_args--;
7bc95ae1
FC
3245 }
3246 if ((len_sv = POPs)) {
3247 len_iv = SvIV(len_sv);
83f78d1a 3248 len_is_uv = len_iv ? SvIOK_UV(len_sv) : 1;
7b8d334a 3249 }
7bc95ae1 3250 else num_args--;
5d82c453 3251 }
777f7c56
EB
3252 pos_sv = POPs;
3253 pos1_iv = SvIV(pos_sv);
3254 pos1_is_uv = SvIOK_UV(pos_sv);
79072805 3255 sv = POPs;
24fcb59f
FC
3256 if (PL_op->op_private & OPpSUBSTR_REPL_FIRST) {
3257 assert(!repl_sv);
3258 repl_sv = POPs;
3259 }
6582db62 3260 if (lvalue && !repl_sv) {
83f78d1a
FC
3261 SV * ret;
3262 ret = sv_2mortal(newSV_type(SVt_PVLV)); /* Not TARG RT#67838 */
3263 sv_magic(ret, NULL, PERL_MAGIC_substr, NULL, 0);
3264 LvTYPE(ret) = 'x';
3265 LvTARG(ret) = SvREFCNT_inc_simple(sv);
3266 LvTARGOFF(ret) =
3267 pos1_is_uv || pos1_iv >= 0
3268 ? (STRLEN)(UV)pos1_iv
b063b0a8 3269 : (LvFLAGS(ret) |= LVf_NEG_OFF, (STRLEN)(UV)-pos1_iv);
83f78d1a
FC
3270 LvTARGLEN(ret) =
3271 len_is_uv || len_iv > 0
3272 ? (STRLEN)(UV)len_iv
b063b0a8 3273 : (LvFLAGS(ret) |= LVf_NEG_LEN, (STRLEN)(UV)-len_iv);
83f78d1a 3274
83f78d1a
FC
3275 PUSHs(ret); /* avoid SvSETMAGIC here */
3276 RETURN;
a74fb2cd 3277 }
6582db62
FC
3278 if (repl_sv) {
3279 repl = SvPV_const(repl_sv, repl_len);
3280 SvGETMAGIC(sv);
3281 if (SvROK(sv))
3282 Perl_ck_warner(aTHX_ packWARN(WARN_SUBSTR),
3283 "Attempt to use reference as lvalue in substr"
3284 );
3285 tmps = SvPV_force_nomg(sv, curlen);
3286 if (DO_UTF8(repl_sv) && repl_len) {
3287 if (!DO_UTF8(sv)) {
41b1e858
AC
3288 /* Upgrade the dest, and recalculate tmps in case the buffer
3289 * got reallocated; curlen may also have been changed */
01680ee9 3290 sv_utf8_upgrade_nomg(sv);
41b1e858 3291 tmps = SvPV_nomg(sv, curlen);
6582db62
FC
3292 }
3293 }
3294 else if (DO_UTF8(sv))
3295 repl_need_utf8_upgrade = TRUE;
3296 }
3297 else tmps = SvPV_const(sv, curlen);
7e2040f0 3298 if (DO_UTF8(sv)) {
0d788f38 3299 utf8_curlen = sv_or_pv_len_utf8(sv, tmps, curlen);
9402d6ed
JH
3300 if (utf8_curlen == curlen)
3301 utf8_curlen = 0;
a0ed51b3 3302 else
9402d6ed 3303 curlen = utf8_curlen;
a0ed51b3 3304 }
d1c2b58a 3305 else
9402d6ed 3306 utf8_curlen = 0;
a0ed51b3 3307
83f78d1a
FC
3308 {
3309 STRLEN pos, len, byte_len, byte_pos;
777f7c56 3310
83f78d1a
FC
3311 if (!translate_substr_offsets(
3312 curlen, pos1_iv, pos1_is_uv, len_iv, len_is_uv, &pos, &len
3313 )) goto bound_fail;
777f7c56 3314
83f78d1a
FC
3315 byte_len = len;
3316 byte_pos = utf8_curlen
0d788f38 3317 ? sv_or_pv_pos_u2b(sv, tmps, pos, &byte_len) : pos;
d931b1be 3318
2154eca7 3319 tmps += byte_pos;
bbddc9e0
CS
3320
3321 if (rvalue) {
3322 SvTAINTED_off(TARG); /* decontaminate */
3323 SvUTF8_off(TARG); /* decontaminate */
3324 sv_setpvn(TARG, tmps, byte_len);
12aa1545 3325#ifdef USE_LOCALE_COLLATE
bbddc9e0 3326 sv_unmagic(TARG, PERL_MAGIC_collxfrm);
12aa1545 3327#endif
bbddc9e0
CS
3328 if (utf8_curlen)
3329 SvUTF8_on(TARG);
3330 }
2154eca7 3331
f7928d6c 3332 if (repl) {
13e30c65
JH
3333 SV* repl_sv_copy = NULL;
3334
3335 if (repl_need_utf8_upgrade) {
3336 repl_sv_copy = newSVsv(repl_sv);
3337 sv_utf8_upgrade(repl_sv_copy);
349d4f2f 3338 repl = SvPV_const(repl_sv_copy, repl_len);
13e30c65 3339 }
502d9230 3340 if (!SvOK(sv))
500f3e18 3341 SvPVCLEAR(sv);
777f7c56 3342 sv_insert_flags(sv, byte_pos, byte_len, repl, repl_len, 0);
ef8d46e8 3343 SvREFCNT_dec(repl_sv_copy);
f7928d6c 3344 }
79072805 3345 }
6a9665b0
FC
3346 if (PL_op->op_private & OPpSUBSTR_REPL_FIRST)
3347 SP++;
3348 else if (rvalue) {
bbddc9e0
CS
3349 SvSETMAGIC(TARG);
3350 PUSHs(TARG);
3351 }
79072805 3352 RETURN;
777f7c56 3353
7b52d656 3354 bound_fail:
83f78d1a 3355 if (repl)
777f7c56
EB
3356 Perl_croak(aTHX_ "substr outside of string");
3357 Perl_ck_warner(aTHX_ packWARN(WARN_SUBSTR), "substr outside of string");
3358 RETPUSHUNDEF;
79072805
LW
3359}
3360
3361PP(pp_vec)
3362{
20b7effb 3363 dSP;
eb578fdb 3364 const IV size = POPi;
d69c4304 3365 SV* offsetsv = POPs;
eb578fdb 3366 SV * const src = POPs;
1b6737cc 3367 const I32 lvalue = PL_op->op_flags & OPf_MOD || LVRET;
2154eca7 3368 SV * ret;
1b92e694
DM
3369 UV retuv;
3370 STRLEN offset = 0;
3371 char errflags = 0;
d69c4304
DM
3372
3373 /* extract a STRLEN-ranged integer value from offsetsv into offset,
1b92e694 3374 * or flag that its out of range */
d69c4304
DM
3375 {
3376 IV iv = SvIV(offsetsv);
3377
3378 /* avoid a large UV being wrapped to a negative value */
1b92e694 3379 if (SvIOK_UV(offsetsv) && SvUVX(offsetsv) > (UV)IV_MAX)
b063b0a8 3380 errflags = LVf_OUT_OF_RANGE;
1b92e694 3381 else if (iv < 0)
b063b0a8 3382 errflags = (LVf_NEG_OFF|LVf_OUT_OF_RANGE);
d69c4304 3383#if PTRSIZE < IVSIZE
1b92e694 3384 else if (iv > Size_t_MAX)
b063b0a8 3385 errflags = LVf_OUT_OF_RANGE;
d69c4304 3386#endif
1b92e694
DM
3387 else
3388 offset = (STRLEN)iv;
d69c4304
DM
3389 }
3390
1b92e694 3391 retuv = errflags ? 0 : do_vecget(src, offset, size);
a0d0e21e 3392
81e118e0 3393 if (lvalue) { /* it's an lvalue! */
2154eca7
EB
3394 ret = sv_2mortal(newSV_type(SVt_PVLV)); /* Not TARG RT#67838 */
3395 sv_magic(ret, NULL, PERL_MAGIC_vec, NULL, 0);
3396 LvTYPE(ret) = 'v';
3397 LvTARG(ret) = SvREFCNT_inc_simple(src);
3398 LvTARGOFF(ret) = offset;
3399 LvTARGLEN(ret) = size;
1b92e694 3400 LvFLAGS(ret) = errflags;
2154eca7
EB
3401 }
3402 else {
3403 dTARGET;
3404 SvTAINTED_off(TARG); /* decontaminate */
3405 ret = TARG;
79072805
LW
3406 }
3407
d69c4304 3408 sv_setuv(ret, retuv);
f9e95907
FC
3409 if (!lvalue)
3410 SvSETMAGIC(ret);
2154eca7 3411 PUSHs(ret);
79072805
LW
3412 RETURN;
3413}
3414
b1c05ba5
DM
3415
3416/* also used for: pp_rindex() */
3417
79072805
LW
3418PP(pp_index)
3419{
20b7effb 3420 dSP; dTARGET;
79072805
LW
3421 SV *big;
3422 SV *little;
c445ea15 3423 SV *temp = NULL;
ad66a58c 3424 STRLEN biglen;
2723d216 3425 STRLEN llen = 0;
b464e2b7
TC
3426 SSize_t offset = 0;
3427 SSize_t retval;
73ee8be2
NC
3428 const char *big_p;
3429 const char *little_p;
2f040f7f
NC
3430 bool big_utf8;
3431 bool little_utf8;
2723d216 3432 const bool is_index = PL_op->op_type == OP_INDEX;
d3e26383 3433 const bool threeargs = MAXARG >= 3 && (TOPs || ((void)POPs,0));
79072805 3434
e1dccc0d
Z
3435 if (threeargs)
3436 offset = POPi;
79072805
LW
3437 little = POPs;
3438 big = POPs;
73ee8be2
NC
3439 big_p = SvPV_const(big, biglen);
3440 little_p = SvPV_const(little, llen);
3441
e609e586
NC
3442 big_utf8 = DO_UTF8(big);
3443 little_utf8 = DO_UTF8(little);
3444 if (big_utf8 ^ little_utf8) {
3445 /* One needs to be upgraded. */
8df0e7a2 3446 if (little_utf8) {
2f040f7f
NC
3447 /* Well, maybe instead we might be able to downgrade the small
3448 string? */
1eced8f8 3449 char * const pv = (char*)bytes_from_utf8((U8 *)little_p, &llen,
2f040f7f
NC
3450 &little_utf8);
3451 if (little_utf8) {
3452 /* If the large string is ISO-8859-1, and it's not possible to
3453 convert the small string to ISO-8859-1, then there is no
3454 way that it could be found anywhere by index. */
3455 retval = -1;
7e8d786b 3456 goto push_result;
2f040f7f 3457 }
e609e586 3458
2f040f7f
NC
3459 /* At this point, pv is a malloc()ed string. So donate it to temp
3460 to ensure it will get free()d */
3461 little = temp = newSV(0);
73ee8be2
NC
3462 sv_usepvn(temp, pv, llen);
3463 little_p = SvPVX(little);
e609e586 3464 } else {
20e67ba1 3465 temp = newSVpvn(little_p, llen);
2f040f7f 3466
8df0e7a2 3467 sv_utf8_upgrade(temp);
20e67ba1
FC
3468 little = temp;
3469 little_p = SvPV_const(little, llen);
e609e586
NC
3470 }
3471 }
73ee8be2
NC
3472 if (SvGAMAGIC(big)) {
3473 /* Life just becomes a lot easier if I use a temporary here.
3474 Otherwise I need to avoid calls to sv_pos_u2b(), which (dangerously)
3475 will trigger magic and overloading again, as will fbm_instr()
3476 */
59cd0e26
NC
3477 big = newSVpvn_flags(big_p, biglen,
3478 SVs_TEMP | (big_utf8 ? SVf_UTF8 : 0));
73ee8be2
NC
3479 big_p = SvPVX(big);
3480 }
e4e44778 3481 if (SvGAMAGIC(little) || (is_index && !SvOK(little))) {
73ee8be2
NC
3482 /* index && SvOK() is a hack. fbm_instr() calls SvPV_const, which will
3483 warn on undef, and we've already triggered a warning with the
3484 SvPV_const some lines above. We can't remove that, as we need to
3485 call some SvPV to trigger overloading early and find out if the
3486 string is UTF-8.
8bd97c0c 3487 This is all getting too messy. The API isn't quite clean enough,
73ee8be2
NC
3488 because data access has side effects.
3489 */
59cd0e26
NC
3490 little = newSVpvn_flags(little_p, llen,
3491 SVs_TEMP | (little_utf8 ? SVf_UTF8 : 0));
73ee8be2
NC
3492 little_p = SvPVX(little);
3493 }
e609e586 3494
d3e26383 3495 if (!threeargs)
2723d216 3496 offset = is_index ? 0 : biglen;
a0ed51b3 3497 else {
ad66a58c 3498 if (big_utf8 && offset > 0)
b464e2b7 3499 offset = sv_pos_u2b_flags(big, offset, 0, SV_CONST_RETURN);
73ee8be2
NC
3500 if (!is_index)
3501 offset += llen;
a0ed51b3 3502 }
79072805
LW
3503 if (offset < 0)
3504 offset = 0;
b464e2b7 3505 else if (offset > (SSize_t)biglen)
ad66a58c 3506 offset = biglen;
73ee8be2
NC
3507 if (!(little_p = is_index
3508 ? fbm_instr((unsigned char*)big_p + offset,
3509 (unsigned char*)big_p + biglen, little, 0)
3510 : rninstr(big_p, big_p + offset,
3511 little_p, little_p + llen)))
a0ed51b3 3512 retval = -1;
ad66a58c 3513 else {
73ee8be2 3514 retval = little_p - big_p;
15c41403 3515 if (retval > 1 && big_utf8)
b464e2b7 3516 retval = sv_pos_b2u_flags(big, retval, SV_CONST_RETURN);
ad66a58c 3517 }
ef8d46e8 3518 SvREFCNT_dec(temp);
7e8d786b
DM
3519
3520 push_result:
3521 /* OPpTRUEBOOL indicates an '== -1' has been optimised away */
3522 if (PL_op->op_private & OPpTRUEBOOL) {
e2d0e9a5
TC
3523 SV *result = ((retval != -1) ^ cBOOL(PL_op->op_private & OPpINDEX_BOOLNEG))
3524 ? &PL_sv_yes : &PL_sv_no;
3525 if (PL_op->op_private & OPpTARGET_MY) {
7e8d786b 3526 /* $lex = (index() == -1) */
e2d0e9a5
TC
3527 sv_setsv_mg(TARG, result);
3528 PUSHs(TARG);
3529 }
3530 else {
3531 PUSHs(result);
3532 }
7e8d786b 3533 }
a8e41ef4 3534 else
7e8d786b 3535 PUSHi(retval);
79072805
LW
3536 RETURN;
3537}
3538
3539PP(pp_sprintf)
3540{
20b7effb 3541 dSP; dMARK; dORIGMARK; dTARGET;
3e6bd4bf 3542 SvTAINTED_off(TARG);
79072805 3543 do_sprintf(TARG, SP-MARK, MARK+1);
bbce6d69 3544 TAINT_IF(SvTAINTED(TARG));
79072805
LW
3545 SP = ORIGMARK;
3546 PUSHTARG;
3547 RETURN;
3548}
3549
79072805
LW
3550PP(pp_ord)
3551{
20b7effb 3552 dSP; dTARGET;
1eced8f8 3553
6ba92227 3554 SV *argsv = TOPs;
ba210ebe 3555 STRLEN len;
349d4f2f 3556 const U8 *s = (U8*)SvPV_const(argsv, len);
121910a4 3557
6ba92227 3558 SETu(DO_UTF8(argsv)
aee9b917 3559 ? (len ? utf8n_to_uvchr(s, len, 0, UTF8_ALLOW_ANYUV) : 0)
f3943cf2 3560 : (UV)(*s));
68795e93 3561
6ba92227 3562 return NORMAL;
79072805
LW
3563}
3564
463ee0b2
LW
3565PP(pp_chr)
3566{
20b7effb 3567 dSP; dTARGET;
463ee0b2 3568 char *tmps;
8a064bd6 3569 UV value;
d3261b99 3570 SV *top = TOPs;
8a064bd6 3571
71739502 3572 SvGETMAGIC(top);
9911fc4e
FC
3573 if (UNLIKELY(SvAMAGIC(top)))
3574 top = sv_2num(top);
99f450cc 3575 if (UNLIKELY(isinfnansv(top)))
147e3846 3576 Perl_croak(aTHX_ "Cannot chr %" NVgf, SvNV(top));
1cd88304
JH
3577 else {
3578 if (!IN_BYTES /* under bytes, chr(-1) eq chr(0xff), etc. */
3579 && ((SvIOKp(top) && !SvIsUV(top) && SvIV_nomg(top) < 0)
3580 ||
3581 ((SvNOKp(top) || (SvOK(top) && !SvIsUV(top)))
2cc2a5a0
KW
3582 && SvNV_nomg(top) < 0.0)))
3583 {
b3fe8680
FC
3584 if (ckWARN(WARN_UTF8)) {
3585 if (SvGMAGICAL(top)) {
3586 SV *top2 = sv_newmortal();
3587 sv_setsv_nomg(top2, top);
3588 top = top2;
3589 }
1cd88304 3590 Perl_warner(aTHX_ packWARN(WARN_UTF8),
147e3846 3591 "Invalid negative number (%" SVf ") in chr", SVfARG(top));
1cd88304
JH
3592 }
3593 value = UNICODE_REPLACEMENT;
3594 } else {
3595 value = SvUV_nomg(top);
3596 }
8a064bd6 3597 }
463ee0b2 3598
862a34c6 3599 SvUPGRADE(TARG,SVt_PV);
a0ed51b3 3600
0064a8a9 3601 if (value > 255 && !IN_BYTES) {
5f560d8a 3602 SvGROW(TARG, (STRLEN)UVCHR_SKIP(value)+1);
62961d2e 3603 tmps = (char*)uvchr_to_utf8_flags((U8*)SvPVX(TARG), value, 0);
349d4f2f 3604 SvCUR_set(TARG, tmps - SvPVX_const(TARG));
a0ed51b3
LW
3605 *tmps = '\0';
3606 (void)SvPOK_only(TARG);
aa6ffa16 3607 SvUTF8_on(TARG);
d3261b99
FC
3608 SETTARG;
3609 return NORMAL;
a0ed51b3
LW
3610 }
3611
748a9306 3612 SvGROW(TARG,2);
463ee0b2
LW
3613 SvCUR_set(TARG, 1);
3614 tmps = SvPVX(TARG);
eb160463 3615 *tmps++ = (char)value;
748a9306 3616 *tmps = '\0';
a0d0e21e 3617 (void)SvPOK_only(TARG);
4c5ed6e2 3618
d3261b99
FC
3619 SETTARG;
3620 return NORMAL;
463ee0b2
LW
3621}
3622
79072805
LW
3623PP(pp_crypt)
3624{
79072805 3625#ifdef HAS_CRYPT
20b7effb 3626 dSP; dTARGET;
5f74f29c 3627 dPOPTOPssrl;
85c16d83 3628 STRLEN len;
10516c54 3629 const char *tmps = SvPV_const(left, len);
2bc69dc4 3630
85c16d83 3631 if (DO_UTF8(left)) {
2bc69dc4 3632 /* If Unicode, try to downgrade.
f2791508
JH
3633 * If not possible, croak.
3634 * Yes, we made this up. */
659fbb76 3635 SV* const tsv = newSVpvn_flags(tmps, len, SVf_UTF8|SVs_TEMP);
2bc69dc4 3636
2bc69dc4 3637 sv_utf8_downgrade(tsv, FALSE);
349d4f2f 3638 tmps = SvPV_const(tsv, len);
85c16d83 3639 }
cf7477a0
DM
3640# ifdef USE_ITHREADS
3641# ifdef HAS_CRYPT_R
05404ffe
JH
3642 if (!PL_reentrant_buffer->_crypt_struct_buffer) {
3643 /* This should be threadsafe because in ithreads there is only
3644 * one thread per interpreter. If this would not be true,
3645 * we would need a mutex to protect this malloc. */
3646 PL_reentrant_buffer->_crypt_struct_buffer =
3647 (struct crypt_data *)safemalloc(sizeof(struct crypt_data));
cf7477a0 3648# if defined(__GLIBC__) || defined(__EMX__)
05404ffe
JH
3649 if (PL_reentrant_buffer->_crypt_struct_buffer) {
3650 PL_reentrant_buffer->_crypt_struct_buffer->initialized = 0;
05404ffe 3651 }
cf7477a0 3652# endif
6ab58e4d 3653 }
cf7477a0
DM
3654# endif /* HAS_CRYPT_R */
3655# endif /* USE_ITHREADS */
3656
3657# ifdef FCRYPT
83003860 3658 sv_setpv(TARG, fcrypt(tmps, SvPV_nolen_const(right)));
cf7477a0 3659# else
83003860 3660 sv_setpv(TARG, PerlProc_crypt(tmps, SvPV_nolen_const(right)));
cf7477a0 3661# endif
fbc76eb3 3662 SvUTF8_off(TARG);
ec93b65f 3663 SETTARG;
4808266b 3664 RETURN;
79072805 3665#else
b13b2135 3666 DIE(aTHX_
79072805
LW
3667 "The crypt() function is unimplemented due to excessive paranoia.");
3668#endif
79072805
LW
3669}
3670
a8e41ef4 3671/* Generally UTF-8 and UTF-EBCDIC are indistinguishable at this level. So
00f254e2
KW
3672 * most comments below say UTF-8, when in fact they mean UTF-EBCDIC as well */
3673
b1c05ba5
DM
3674
3675/* also used for: pp_lcfirst() */
3676
79072805
LW
3677PP(pp_ucfirst)
3678{
00f254e2
KW
3679 /* Actually is both lcfirst() and ucfirst(). Only the first character
3680 * changes. This means that possibly we can change in-place, ie., just
3681 * take the source and change that one character and store it back, but not
3682 * if read-only etc, or if the length changes */
3683
39644a26 3684 dSP;
d54190f6 3685 SV *source = TOPs;
00f254e2 3686 STRLEN slen; /* slen is the byte length of the whole SV. */
d54190f6
NC
3687 STRLEN need;
3688 SV *dest;
00f254e2
KW
3689 bool inplace; /* ? Convert first char only, in-place */
3690 bool doing_utf8 = FALSE; /* ? using utf8 */
3691 bool convert_source_to_utf8 = FALSE; /* ? need to convert */
12e9c124 3692 const int op_type = PL_op->op_type;
d54190f6
NC
3693 const U8 *s;
3694 U8 *d;
3695 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
00f254e2
KW
3696 STRLEN ulen; /* ulen is the byte length of the original Unicode character
3697 * stored as UTF-8 at s. */
3698 STRLEN tculen; /* tculen is the byte length of the freshly titlecased (or
3699 * lowercased) character stored in tmpbuf. May be either
3700 * UTF-8 or not, but in either case is the number of bytes */
be42d347 3701 bool remove_dot_above = FALSE;
d54190f6 3702
841a5e18 3703 s = (const U8*)SvPV_const(source, slen);
a0ed51b3 3704
00f254e2
KW
3705 /* We may be able to get away with changing only the first character, in
3706 * place, but not if read-only, etc. Later we may discover more reasons to
3707 * not convert in-place. */
1921e031 3708 inplace = !SvREADONLY(source) && SvPADTMP(source);
00f254e2 3709
8b7358b9
KW
3710#ifdef USE_LOCALE_CTYPE
3711
3712 if (IN_LC_RUNTIME(LC_CTYPE)) {
3713 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
3714 }
3715
3716#endif
3717
00f254e2
KW
3718 /* First calculate what the changed first character should be. This affects
3719 * whether we can just swap it out, leaving the rest of the string unchanged,
3720 * or even if have to convert the dest to UTF-8 when the source isn't */
3721
3722 if (! slen) { /* If empty */
3723 need = 1; /* still need a trailing NUL */
b7576bcb 3724 ulen = 0;
62e6b705 3725 *tmpbuf = '\0';
00f254e2
KW
3726 }
3727 else if (DO_UTF8(source)) { /* Is the source utf8? */
d54190f6 3728 doing_utf8 = TRUE;
17e95c9d 3729 ulen = UTF8SKIP(s);
190e86d7 3730
094a2f8c 3731 if (op_type == OP_UCFIRST) {
130c5df3 3732#ifdef USE_LOCALE_CTYPE
a1a5ec35 3733 _toTITLE_utf8_flags(s, s +slen, tmpbuf, &tculen, IN_LC_RUNTIME(LC_CTYPE));
130c5df3 3734#else
a1a5ec35 3735 _toTITLE_utf8_flags(s, s +slen, tmpbuf, &tculen, 0);
130c5df3 3736#endif
094a2f8c
KW
3737 }
3738 else {
a8e41ef4 3739
130c5df3 3740#ifdef USE_LOCALE_CTYPE
a8e41ef4 3741
a1a5ec35 3742 _toLOWER_utf8_flags(s, s + slen, tmpbuf, &tculen, IN_LC_RUNTIME(LC_CTYPE));
be42d347
KW
3743
3744 /* In turkic locales, lower casing an 'I' normally yields U+0131,
3745 * LATIN SMALL LETTER DOTLESS I, but not if the grapheme also
3746 * contains a COMBINING DOT ABOVE. Instead it is treated like
3747 * LATIN CAPITAL LETTER I WITH DOT ABOVE lowercased to 'i'. The
3748 * call to lowercase above has handled this. But SpecialCasing.txt
3749 * says we are supposed to remove the COMBINING DOT ABOVE. We can
3750 * tell if we have this situation if I ==> i in a turkic locale. */
3751 if ( UNLIKELY(PL_in_utf8_turkic_locale)
3752 && IN_LC_RUNTIME(LC_CTYPE)
3753 && (UNLIKELY(*s == 'I' && tmpbuf[0] == 'i')))
3754 {
3755 /* Here, we know there was a COMBINING DOT ABOVE. We won't be
3756 * able to handle this in-place. */
3757 inplace = FALSE;
3758
3759 /* It seems likely that the DOT will immediately follow the
3760 * 'I'. If so, we can remove it simply by indicating to the
3761 * code below to start copying the source just beyond the DOT.
3762 * We know its length is 2 */
3763 if (LIKELY(memBEGINs(s + 1, s + slen, COMBINING_DOT_ABOVE_UTF8))) {
3764 ulen += 2;
3765 }
3766 else { /* But if it doesn't follow immediately, set a flag for
3767 the code below */
3768 remove_dot_above = TRUE;
3769 }
3770 }
130c5df3 3771#else
be42d347
KW
3772 PERL_UNUSED_VAR(remove_dot_above);
3773
a1a5ec35 3774 _toLOWER_utf8_flags(s, s + slen, tmpbuf, &tculen, 0);
130c5df3 3775#endif
a8e41ef4
KW
3776
3777 }
00f254e2 3778
17e95c9d
KW
3779 /* we can't do in-place if the length changes. */
3780 if (ulen != tculen) inplace = FALSE;
3781 need = slen + 1 - ulen + tculen;
d54190f6 3782 }
00f254e2
KW
3783 else { /* Non-zero length, non-UTF-8, Need to consider locale and if
3784 * latin1 is treated as caseless. Note that a locale takes
a8e41ef4 3785 * precedence */
167d19f2 3786 ulen = 1; /* Original character is 1 byte */
00f254e2
KW
3787 tculen = 1; /* Most characters will require one byte, but this will
3788 * need to be overridden for the tricky ones */
3789 need = slen + 1;
3790
d54190f6 3791
130c5df3 3792#ifdef USE_LOCALE_CTYPE
be42d347
KW
3793
3794 if (IN_LC_RUNTIME(LC_CTYPE)) {
3795 if ( UNLIKELY(PL_in_utf8_turkic_locale)
3796 && ( (op_type == OP_LCFIRST && UNLIKELY(*s == 'I'))
3797 || (op_type == OP_UCFIRST && UNLIKELY(*s == 'i'))))
780fcc9f 3798 {
be42d347
KW
3799 if (*s == 'I') { /* lcfirst('I') */
3800 tmpbuf[0] = UTF8_TWO_BYTE_HI(LATIN_SMALL_LETTER_DOTLESS_I);
3801 tmpbuf[1] = UTF8_TWO_BYTE_LO(LATIN_SMALL_LETTER_DOTLESS_I);
3802 }
3803 else { /* ucfirst('i') */
3804 tmpbuf[0] = UTF8_TWO_BYTE_HI(LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE);
3805 tmpbuf[1] = UTF8_TWO_BYTE_LO(LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE);
3806 }
3807 tculen = 2;
3808 inplace = FALSE;
3809 doing_utf8 = TRUE;
3810 convert_source_to_utf8 = TRUE;
3811 need += variant_under_utf8_count(s, s + slen);
780fcc9f 3812 }
be42d347
KW
3813 else if (op_type == OP_LCFIRST) {
3814
3815 /* For lc, there are no gotchas for UTF-8 locales (other than
3816 * the turkish ones already handled above) */
3817 *tmpbuf = toLOWER_LC(*s);
31f05a37 3818 }
be42d347 3819 else { /* ucfirst */
31f05a37 3820
be42d347
KW
3821 /* But for uc, some characters require special handling */
3822 if (IN_UTF8_CTYPE_LOCALE) {
3823 goto do_uni_rules;
3824 }
3825
3826 /* This would be a bug if any locales have upper and title case
3827 * different */
3828 *tmpbuf = (U8) toUPPER_LC(*s);
3829 }
3830 }
3831 else
130c5df3 3832#endif
be42d347
KW
3833 /* Here, not in locale. If not using Unicode rules, is a simple
3834 * lower/upper, depending */
3835 if (! IN_UNI_8_BIT) {
3836 *tmpbuf = (op_type == OP_LCFIRST)
3837 ? toLOWER(*s)
3838 : toUPPER(*s);
3839 }
3840 else if (op_type == OP_LCFIRST) {
3841 /* lower case the first letter: no trickiness for any character */
3842 *tmpbuf = toLOWER_LATIN1(*s);
3843 }
31f05a37
KW
3844 else {
3845 /* Here, is ucfirst non-UTF-8, not in locale (unless that locale is
be42d347
KW
3846 * non-turkic UTF-8, which we treat as not in locale), and cased
3847 * latin1 */
31f05a37 3848 UV title_ord;
91191cf7 3849#ifdef USE_LOCALE_CTYPE
31f05a37 3850 do_uni_rules:
91191cf7 3851#endif
31f05a37
KW
3852
3853 title_ord = _to_upper_title_latin1(*s, tmpbuf, &tculen, 's');
167d19f2
KW
3854 if (tculen > 1) {
3855 assert(tculen == 2);
3856
3857 /* If the result is an upper Latin1-range character, it can
3858 * still be represented in one byte, which is its ordinal */
3859 if (UTF8_IS_DOWNGRADEABLE_START(*tmpbuf)) {
3860 *tmpbuf = (U8) title_ord;
3861 tculen = 1;
00f254e2
KW
3862 }
3863 else {
167d19f2
KW
3864 /* Otherwise it became more than one ASCII character (in
3865 * the case of LATIN_SMALL_LETTER_SHARP_S) or changed to
3866 * beyond Latin1, so the number of bytes changed, so can't
3867 * replace just the first character in place. */
3868 inplace = FALSE;
3869
d14578b8 3870 /* If the result won't fit in a byte, the entire result
2f8f985a
KW
3871 * will have to be in UTF-8. Allocate enough space for the
3872 * expanded first byte, and if UTF-8, the rest of the input
3873 * string, some or all of which may also expand to two
3874 * bytes, plus the terminating NUL. */
167d19f2
KW
3875 if (title_ord > 255) {
3876 doing_utf8 = TRUE;
3877 convert_source_to_utf8 = TRUE;
2f8f985a
KW
3878 need = slen
3879 + variant_under_utf8_count(s, s + slen)
3880 + 1;
167d19f2
KW
3881
3882 /* The (converted) UTF-8 and UTF-EBCDIC lengths of all
be42d347 3883 * characters whose title case is above 255 is
167d19f2
KW
3884 * 2. */
3885 ulen = 2;
3886 }
3887 else { /* LATIN_SMALL_LETTER_SHARP_S expands by 1 byte */
3888 need = slen + 1 + 1;
3889 }
00f254e2 3890 }
167d19f2 3891 }
00f254e2
KW
3892 } /* End of use Unicode (Latin1) semantics */
3893 } /* End of changing the case of the first character */
3894
3895 /* Here, have the first character's changed case stored in tmpbuf. Ready to
3896 * generate the result */
3897 if (inplace) {
3898
3899 /* We can convert in place. This means we change just the first
3900 * character without disturbing the rest; no need to grow */
d54190f6
NC
3901 dest = source;
3902 s = d = (U8*)SvPV_force_nomg(source, slen);
3903 } else {
3904 dTARGET;
3905
3906 dest = TARG;
3907
00f254e2
KW
3908 /* Here, we can't convert in place; we earlier calculated how much
3909 * space we will need, so grow to accommodate that */
d54190f6 3910 SvUPGRADE(dest, SVt_PV);
3b416f41 3911 d = (U8*)SvGROW(dest, need);
d54190f6
NC
3912 (void)SvPOK_only(dest);
3913
3914 SETs(dest);
d54190f6 3915 }
44bc797b 3916
d54190f6 3917 if (doing_utf8) {
00f254e2
KW
3918 if (! inplace) {
3919 if (! convert_source_to_utf8) {
3920
3921 /* Here both source and dest are in UTF-8, but have to create
3922 * the entire output. We initialize the result to be the
3923 * title/lower cased first character, and then append the rest
3924 * of the string. */
3925 sv_setpvn(dest, (char*)tmpbuf, tculen);
3926 if (slen > ulen) {
be42d347
KW
3927
3928 /* But this boolean being set means we are in a turkic
3929 * locale, and there is a DOT character that needs to be
3930 * removed, and it isn't immediately after the current
3931 * character. Keep concatenating characters to the output
3932 * one at a time, until we find the DOT, which we simply
3933 * skip */
3934 if (UNLIKELY(remove_dot_above)) {
3935 do {
3936 Size_t this_len = UTF8SKIP(s + ulen);
3937
3938 sv_catpvn(dest, (char*)(s + ulen), this_len);
3939
3940 ulen += this_len;
3941 if (memBEGINs(s + ulen, s + slen, COMBINING_DOT_ABOVE_UTF8)) {
3942 ulen += 2;
3943 break;
3944 }
3945 } while (s + ulen < s + slen);
3946 }
3947
3948 /* The rest of the string can be concatenated unchanged,
3949 * all at once */
00f254e2
KW
3950 sv_catpvn(dest, (char*)(s + ulen), slen - ulen);
3951 }
3952 }
3953 else {
3954 const U8 *const send = s + slen;
3955
3956 /* Here the dest needs to be in UTF-8, but the source isn't,
3957 * except we earlier UTF-8'd the first character of the source
3958 * into tmpbuf. First put that into dest, and then append the
3959 * rest of the source, converting it to UTF-8 as we go. */
3960
be42d347 3961 /* Assert tculen is 2 here because the only characters that
00f254e2 3962 * get to this part of the code have 2-byte UTF-8 equivalents */
f4cd1cd9 3963 assert(tculen == 2);
00f254e2
KW
3964 *d++ = *tmpbuf;
3965 *d++ = *(tmpbuf + 1);
3966 s++; /* We have just processed the 1st char */
3967
df7d4938
KW
3968 while (s < send) {
3969 append_utf8_from_native_byte(*s, &d);
3970 s++;
3971 }
3972
00f254e2
KW
3973 *d = '\0';
3974 SvCUR_set(dest, d - (U8*)SvPVX_const(dest));
3975 }
d54190f6 3976 SvUTF8_on(dest);
a0ed51b3 3977 }
00f254e2 3978 else { /* in-place UTF-8. Just overwrite the first character */
d54190f6
NC
3979 Copy(tmpbuf, d, tculen, U8);
3980 SvCUR_set(dest, need - 1);
a0ed51b3 3981 }
094a2f8c 3982
a0ed51b3 3983 }
a8e41ef4 3984 else { /* Neither source nor dest are, nor need to be UTF-8 */
00f254e2 3985 if (slen) {
00f254e2
KW
3986 if (inplace) { /* in-place, only need to change the 1st char */
3987 *d = *tmpbuf;
3988 }
3989 else { /* Not in-place */
3990
3991 /* Copy the case-changed character(s) from tmpbuf */
3992 Copy(tmpbuf, d, tculen, U8);
3993 d += tculen - 1; /* Code below expects d to point to final
3994 * character stored */
3995 }
3996 }
3997 else { /* empty source */
3998 /* See bug #39028: Don't taint if empty */
d54190f6
NC
3999 *d = *s;
4000 }
4001
00f254e2
KW
4002 /* In a "use bytes" we don't treat the source as UTF-8, but, still want
4003 * the destination to retain that flag */
9edbb8b2 4004 if (DO_UTF8(source))
d54190f6
NC
4005 SvUTF8_on(dest);
4006
00f254e2 4007 if (!inplace) { /* Finish the rest of the string, unchanged */
d54190f6
NC
4008 /* This will copy the trailing NUL */
4009 Copy(s + 1, d + 1, slen, U8);
4010 SvCUR_set(dest, need - 1);
bbce6d69 4011 }
bbce6d69 4012 }
130c5df3 4013#ifdef USE_LOCALE_CTYPE
d6ded950 4014 if (IN_LC_RUNTIME(LC_CTYPE)) {
445bf929
KW
4015 TAINT;
4016 SvTAINTED_on(dest);
4017 }
130c5df3 4018#endif
539689e7
FC
4019 if (dest != source && SvTAINTED(source))
4020 SvTAINT(dest);
d54190f6 4021 SvSETMAGIC(dest);
3cb4e04f 4022 return NORMAL;
79072805
LW
4023}
4024
4025PP(pp_uc)
4026{
1565c085 4027 dVAR;
39644a26 4028 dSP;
67306194 4029 SV *source = TOPs;
463ee0b2 4030 STRLEN len;
67306194
NC
4031 STRLEN min;
4032 SV *dest;
4033 const U8 *s;
4034 U8 *d;
79072805 4035
67306194
NC
4036 SvGETMAGIC(source);
4037
1921e031 4038 if ( SvPADTMP(source)
5cd5e2d6
FC
4039 && !SvREADONLY(source) && SvPOK(source)
4040 && !DO_UTF8(source)
130c5df3
KW
4041 && (
4042#ifdef USE_LOCALE_CTYPE
4043 (IN_LC_RUNTIME(LC_CTYPE))
31f05a37 4044 ? ! IN_UTF8_CTYPE_LOCALE
130c5df3
KW
4045 :
4046#endif
4047 ! IN_UNI_8_BIT))
31f05a37
KW
4048 {
4049
4050 /* We can convert in place. The reason we can't if in UNI_8_BIT is to
4051 * make the loop tight, so we overwrite the source with the dest before
4052 * looking at it, and we need to look at the original source
4053 * afterwards. There would also need to be code added to handle
4054 * switching to not in-place in midstream if we run into characters
4055 * that change the length. Since being in locale overrides UNI_8_BIT,
4056 * that latter becomes irrelevant in the above test; instead for
4057 * locale, the size can't normally change, except if the locale is a
4058 * UTF-8 one */
67306194
NC
4059 dest = source;
4060 s = d = (U8*)SvPV_force_nomg(source, len);
4061 min = len + 1;
4062 } else {
a0ed51b3 4063 dTARGET;
a0ed51b3 4064
67306194 4065 dest = TARG;
128c9517 4066
841a5e18 4067 s = (const U8*)SvPV_nomg_const(source, len);
67306194
NC
4068 min = len + 1;
4069
4070 SvUPGRADE(dest, SVt_PV);
3b416f41 4071 d = (U8*)SvGROW(dest, min);
67306194
NC
4072 (void)SvPOK_only(dest);
4073
4074 SETs(dest);
a0ed51b3 4075 }
31351b04 4076
8b7358b9
KW
4077#ifdef USE_LOCALE_CTYPE
4078
4079 if (IN_LC_RUNTIME(LC_CTYPE)) {
4080 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
4081 }
4082
4083#endif
4084
67306194
NC
4085 /* Overloaded values may have toggled the UTF-8 flag on source, so we need
4086 to check DO_UTF8 again here. */
4087
4088 if (DO_UTF8(source)) {
4089 const U8 *const send = s + len;
bfac13d4 4090 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
67306194 4091
78ed8e36
KW
4092#define GREEK_CAPITAL_LETTER_IOTA 0x0399
4093#define COMBINING_GREEK_YPOGEGRAMMENI 0x0345
4c8a458a
KW
4094 /* All occurrences of these are to be moved to follow any other marks.
4095 * This is context-dependent. We may not be passed enough context to
4096 * move the iota subscript beyond all of them, but we do the best we can
4097 * with what we're given. The result is always better than if we
4098 * hadn't done this. And, the problem would only arise if we are
4099 * passed a character without all its combining marks, which would be
4100 * the caller's mistake. The information this is based on comes from a
4101 * comment in Unicode SpecialCasing.txt, (and the Standard's text
4102 * itself) and so can't be checked properly to see if it ever gets
4103 * revised. But the likelihood of it changing is remote */
00f254e2 4104 bool in_iota_subscript = FALSE;
00f254e2 4105
67306194 4106 while (s < send) {
3e16b0e6
KW
4107 STRLEN u;
4108 STRLEN ulen;
4109 UV uv;
dbb3849a
KW
4110 if (UNLIKELY(in_iota_subscript)) {
4111 UV cp = utf8_to_uvchr_buf(s, send, NULL);
4112
4113 if (! _invlist_contains_cp(PL_utf8_mark, cp)) {
3e16b0e6 4114
79ba2767
KW
4115 /* A non-mark. Time to output the iota subscript */
4116 *d++ = UTF8_TWO_BYTE_HI(GREEK_CAPITAL_LETTER_IOTA);
4117 *d++ = UTF8_TWO_BYTE_LO(GREEK_CAPITAL_LETTER_IOTA);
4118 in_iota_subscript = FALSE;
dbb3849a 4119 }
8e058693 4120 }
00f254e2 4121
8e058693
KW
4122 /* Then handle the current character. Get the changed case value
4123 * and copy it to the output buffer */
00f254e2 4124
8e058693 4125 u = UTF8SKIP(s);
130c5df3 4126#ifdef USE_LOCALE_CTYPE
a1a5ec35 4127 uv = _toUPPER_utf8_flags(s, send, tmpbuf, &ulen, IN_LC_RUNTIME(LC_CTYPE));
130c5df3 4128#else
a1a5ec35 4129 uv = _toUPPER_utf8_flags(s, send, tmpbuf, &ulen, 0);
130c5df3 4130#endif
8e058693 4131 if (uv == GREEK_CAPITAL_LETTER_IOTA
4b88fb76 4132 && utf8_to_uvchr_buf(s, send, 0) == COMBINING_GREEK_YPOGEGRAMMENI)
8e058693
KW
4133 {
4134 in_iota_subscript = TRUE;
4135 }
4136 else {
4137 if (ulen > u && (SvLEN(dest) < (min += ulen - u))) {
4138 /* If the eventually required minimum size outgrows the
4139 * available space, we need to grow. */
4140 const UV o = d - (U8*)SvPVX_const(dest);
4141
4142 /* If someone uppercases one million U+03B0s we SvGROW()
4143 * one million times. Or we could try guessing how much to
a8e41ef4
KW
4144 * allocate without allocating too much. But we can't
4145 * really guess without examining the rest of the string.
4146 * Such is life. See corresponding comment in lc code for
4147 * another option */
10656159 4148 d = o + (U8*) SvGROW(dest, min);
8e058693
KW
4149 }
4150 Copy(tmpbuf, d, ulen, U8);
4151 d += ulen;
4152 }
4153 s += u;
67306194 4154 }
4c8a458a 4155 if (in_iota_subscript) {
78ed8e36
KW
4156 *d++ = UTF8_TWO_BYTE_HI(GREEK_CAPITAL_LETTER_IOTA);
4157 *d++ = UTF8_TWO_BYTE_LO(GREEK_CAPITAL_LETTER_IOTA);
4c8a458a 4158 }
67306194
NC
4159 SvUTF8_on(dest);
4160 *d = '\0';
094a2f8c 4161
67306194 4162 SvCUR_set(dest, d - (U8*)SvPVX_const(dest));
4c8a458a
KW
4163 }
4164 else { /* Not UTF-8 */
67306194
NC
4165 if (len) {
4166 const U8 *const send = s + len;
00f254e2
KW
4167
4168 /* Use locale casing if in locale; regular style if not treating
4169 * latin1 as having case; otherwise the latin1 casing. Do the
4170 * whole thing in a tight loop, for speed, */
130c5df3 4171#ifdef USE_LOCALE_CTYPE
d6ded950 4172 if (IN_LC_RUNTIME(LC_CTYPE)) {
31f05a37
KW
4173 if (IN_UTF8_CTYPE_LOCALE) {
4174 goto do_uni_rules;
4175 }
67306194 4176 for (; s < send; d++, s++)
31f05a37 4177 *d = (U8) toUPPER_LC(*s);
31351b04 4178 }
130c5df3
KW
4179 else
4180#endif
4181 if (! IN_UNI_8_BIT) {
00f254e2 4182 for (; s < send; d++, s++) {
67306194 4183 *d = toUPPER(*s);
00f254e2 4184 }
31351b04 4185 }
00f254e2 4186 else {
91191cf7 4187#ifdef USE_LOCALE_CTYPE
31f05a37 4188 do_uni_rules:
91191cf7 4189#endif
00f254e2 4190 for (; s < send; d++, s++) {
2f8f985a
KW
4191 Size_t extra;
4192
00f254e2 4193 *d = toUPPER_LATIN1_MOD(*s);
be42d347
KW
4194 if ( LIKELY(*d != LATIN_SMALL_LETTER_Y_WITH_DIAERESIS)
4195
4196#ifdef USE_LOCALE_CTYPE
4197
4198 && (LIKELY( ! PL_in_utf8_turkic_locale
4199 || ! IN_LC_RUNTIME(LC_CTYPE))
4200 || *s != 'i')
4201#endif
4202
4203 ) {
d14578b8
KW
4204 continue;
4205 }
00f254e2
KW
4206
4207 /* The mainstream case is the tight loop above. To avoid
be42d347
KW
4208 * extra tests in that, all three characters that always
4209 * require special handling are mapped by the MOD to the
4210 * one tested just above. Use the source to distinguish
4211 * between those cases */
00f254e2 4212
79e064b9
KW
4213#if UNICODE_MAJOR_VERSION > 2 \
4214 || (UNICODE_MAJOR_VERSION == 2 && UNICODE_DOT_VERSION >= 1 \
4215 && UNICODE_DOT_DOT_VERSION >= 8)
00f254e2
KW
4216 if (*s == LATIN_SMALL_LETTER_SHARP_S) {
4217
4218 /* uc() of this requires 2 characters, but they are
4219 * ASCII. If not enough room, grow the string */
a8e41ef4 4220 if (SvLEN(dest) < ++min) {
00f254e2 4221 const UV o = d - (U8*)SvPVX_const(dest);
10656159 4222 d = o + (U8*) SvGROW(dest, min);
00f254e2
KW
4223 }
4224 *d++ = 'S'; *d = 'S'; /* upper case is 'SS' */
4225 continue; /* Back to the tight loop; still in ASCII */
4226 }
79e064b9 4227#endif
00f254e2 4228
be42d347 4229 /* The other special handling characters have their
00f254e2 4230 * upper cases outside the latin1 range, hence need to be
a8e41ef4
KW
4231 * in UTF-8, so the whole result needs to be in UTF-8.
4232 *
4233 * So, here we are somewhere in the middle of processing a
4234 * non-UTF-8 string, and realize that we will have to
4235 * convert the whole thing to UTF-8. What to do? There
4236 * are several possibilities. The simplest to code is to
4237 * convert what we have so far, set a flag, and continue on
4238 * in the loop. The flag would be tested each time through
4239 * the loop, and if set, the next character would be
4240 * converted to UTF-8 and stored. But, I (khw) didn't want
4241 * to slow down the mainstream case at all for this fairly
4242 * rare case, so I didn't want to add a test that didn't
4243 * absolutely have to be there in the loop, besides the
4244 * possibility that it would get too complicated for
4245 * optimizers to deal with. Another possibility is to just
4246 * give up, convert the source to UTF-8, and restart the
4247 * function that way. Another possibility is to convert
4248 * both what has already been processed and what is yet to
4249 * come separately to UTF-8, then jump into the loop that
4250 * handles UTF-8. But the most efficient time-wise of the
4251 * ones I could think of is what follows, and turned out to
2f8f985a
KW
4252 * not require much extra code.
4253 *
4254 * First, calculate the extra space needed for the
be42d347
KW
4255 * remainder of the source needing to be in UTF-8. Except
4256 * for the 'i' in Turkic locales, in UTF-8 strings, the
2f8f985a
KW
4257 * uppercase of a character below 256 occupies the same
4258 * number of bytes as the original. Therefore, the space
4259 * needed is the that number plus the number of characters
be42d347
KW
4260 * that become two bytes when converted to UTF-8, plus, in
4261 * turkish locales, the number of 'i's. */
2f8f985a
KW
4262
4263 extra = send - s + variant_under_utf8_count(s, send);
a8e41ef4 4264
be42d347
KW
4265#ifdef USE_LOCALE_CTYPE
4266
4267 if (UNLIKELY(*s == 'i')) { /* We wouldn't get an 'i' here
4268 unless are in a Turkic
4269 locale */
4270 const U8 * s_peek = s;
4271
4272 do {
4273 extra++;
4274
4275 s_peek = (U8 *) memchr(s_peek + 1, 'i',
4276 send - (s_peek + 1));
4277 } while (s_peek != NULL);
4278 }
4279#endif
4280
a8e41ef4 4281 /* Convert what we have so far into UTF-8, telling the
00f254e2
KW
4282 * function that we know it should be converted, and to
4283 * allow extra space for what we haven't processed yet.
2f8f985a
KW
4284 *
4285 * This may cause the string pointer to move, so need to
4286 * save and re-find it. */
00f254e2
KW
4287
4288 len = d - (U8*)SvPVX_const(dest);
4289 SvCUR_set(dest, len);
4290 len = sv_utf8_upgrade_flags_grow(dest,
4291 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
56e36cbf
KW
4292 extra
4293 + 1 /* trailing NUL */ );
00f254e2
KW
4294 d = (U8*)SvPVX(dest) + len;
4295
a8e41ef4 4296 /* Now process the remainder of the source, simultaneously
be42d347
KW
4297 * converting to upper and UTF-8.
4298 *
4299 * To avoid extra tests in the loop body, and since the
4300 * loop is so simple, split out the rare Turkic case into
4301 * its own loop */
4302
4303#ifdef USE_LOCALE_CTYPE
4304 if ( UNLIKELY(PL_in_utf8_turkic_locale)
4305 && UNLIKELY(IN_LC_RUNTIME(LC_CTYPE)))
4306 {
4307 for (; s < send; s++) {
4308 if (*s == 'i') {
4309 *d++ = UTF8_TWO_BYTE_HI(LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE);
4310 *d++ = UTF8_TWO_BYTE_LO(LATIN_CAPITAL_LETTER_I_WITH_DOT_ABOVE);
4311 }
4312 else {
4313 (void) _to_upper_title_latin1(*s, d, &len, 'S');
4314 d += len;
4315 }
4316 }
4317 }
4318 else
4319#endif
d813f430
KW
4320 for (; s < send; s++) {
4321 (void) _to_upper_title_latin1(*s, d, &len, 'S');
4322 d += len;
4323 }
4324
be42d347
KW
4325 /* Here have processed the whole source; no need to
4326 * continue with the outer loop. Each character has been
4327 * converted to upper case and converted to UTF-8. */
00f254e2
KW
4328 break;
4329 } /* End of processing all latin1-style chars */
4330 } /* End of processing all chars */
4331 } /* End of source is not empty */
4332
67306194 4333 if (source != dest) {
00f254e2 4334 *d = '\0'; /* Here d points to 1 after last char, add NUL */
67306194
NC
4335 SvCUR_set(dest, d - (U8*)SvPVX_const(dest));
4336 }
00f254e2 4337 } /* End of isn't utf8 */
130c5df3 4338#ifdef USE_LOCALE_CTYPE
d6ded950 4339 if (IN_LC_RUNTIME(LC_CTYPE)) {
445bf929
KW
4340 TAINT;
4341 SvTAINTED_on(dest);
4342 }
130c5df3 4343#endif
539689e7
FC
4344 if (dest != source && SvTAINTED(source))
4345 SvTAINT(dest);
67306194 4346 SvSETMAGIC(dest);
3cb4e04f 4347 return NORMAL;
79072805
LW
4348}
4349
4350PP(pp_lc)
4351{
39644a26 4352 dSP;
ec9af7d4 4353 SV *source = TOPs;
463ee0b2 4354 STRLEN len;
ec9af7d4
NC
4355 STRLEN min;
4356 SV *dest;
4357 const U8 *s;
4358 U8 *d;
be42d347 4359 bool has_turkic_I = FALSE;
79072805 4360
ec9af7d4
NC
4361 SvGETMAGIC(source);
4362
1921e031 4363 if ( SvPADTMP(source)
5cd5e2d6 4364 && !SvREADONLY(source) && SvPOK(source)
be42d347
KW
4365 && !DO_UTF8(source)
4366
4367#ifdef USE_LOCALE_CTYPE
ec9af7d4 4368
be42d347
KW
4369 && ( LIKELY(! IN_LC_RUNTIME(LC_CTYPE))
4370 || LIKELY(! PL_in_utf8_turkic_locale))
4371
4372#endif
4373
4374 ) {
4375
4376 /* We can convert in place, as, outside of Turkic UTF-8 locales,
4377 * lowercasing anything in the latin1 range (or else DO_UTF8 would have
4378 * been on) doesn't lengthen it. */
ec9af7d4
NC
4379 dest = source;
4380 s = d = (U8*)SvPV_force_nomg(source, len);
4381 min = len + 1;
4382 } else {
a0ed51b3 4383 dTARGET;
a0ed51b3 4384
ec9af7d4
NC
4385 dest = TARG;
4386
841a5e18 4387 s = (const U8*)SvPV_nomg_const(source, len);
ec9af7d4 4388 min = len + 1;
128c9517 4389
ec9af7d4 4390 SvUPGRADE(dest, SVt_PV);
3b416f41 4391 d = (U8*)SvGROW(dest, min);
ec9af7d4
NC
4392 (void)SvPOK_only(dest);
4393
4394 SETs(dest);
4395 }
4396
8b7358b9
KW
4397#ifdef USE_LOCALE_CTYPE
4398
4399 if (IN_LC_RUNTIME(LC_CTYPE)) {
be42d347
KW
4400 const U8 * next_I;
4401
8b7358b9 4402 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
be42d347
KW
4403
4404 /* Lowercasing in a Turkic locale can cause non-UTF-8 to need to become
4405 * UTF-8 for the single case of the character 'I' */
4406 if ( UNLIKELY(PL_in_utf8_turkic_locale)
4407 && ! DO_UTF8(source)
4408 && (next_I = (U8 *) memchr(s, 'I', len)))
4409 {
4410 Size_t I_count = 0;
4411 const U8 *const send = s + len;
4412
4413 do {
4414 I_count++;
4415
4416 next_I = (U8 *) memchr(next_I + 1, 'I',
4417 send - (next_I + 1));
4418 } while (next_I != NULL);
4419
4420 /* Except for the 'I', in UTF-8 strings, the lower case of a
4421 * character below 256 occupies the same number of bytes as the
4422 * original. Therefore, the space needed is the original length
4423 * plus I_count plus the number of characters that become two bytes
4424 * when converted to UTF-8 */
4425 sv_utf8_upgrade_flags_grow(dest, 0, len
4426 + I_count
56e36cbf
KW
4427 + variant_under_utf8_count(s, send)
4428 + 1 /* Trailing NUL */ );
be42d347
KW
4429 d = (U8*)SvPVX(dest);
4430 has_turkic_I = TRUE;
4431 }
8b7358b9
KW
4432 }
4433
4434#endif
4435
ec9af7d4
NC
4436 /* Overloaded values may have toggled the UTF-8 flag on source, so we need
4437 to check DO_UTF8 again here. */
4438
4439 if (DO_UTF8(source)) {
4440 const U8 *const send = s + len;
4441 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
be42d347 4442 bool remove_dot_above = FALSE;
ec9af7d4
NC
4443
4444 while (s < send) {
06b5486a
KW
4445 const STRLEN u = UTF8SKIP(s);
4446 STRLEN ulen;
00f254e2 4447
130c5df3 4448#ifdef USE_LOCALE_CTYPE
a8e41ef4 4449
a1a5ec35 4450 _toLOWER_utf8_flags(s, send, tmpbuf, &ulen, IN_LC_RUNTIME(LC_CTYPE));
be42d347
KW
4451
4452 /* If we are in a Turkic locale, we have to do more work. As noted
4453 * in the comments for lcfirst, there is a special case if a 'I'
4454 * is in a grapheme with COMBINING DOT ABOVE UTF8. It turns into a
4455 * 'i', and the DOT must be removed. We check for that situation,
4456 * and set a flag if the DOT is there. Then each time through the
4457 * loop, we have to see if we need to remove the next DOT above,
4458 * and if so, do it. We know that there is a DOT because
4459 * _toLOWER_utf8_flags() wouldn't have returned 'i' unless there
4460 * was one in a proper position. */
4461 if ( UNLIKELY(PL_in_utf8_turkic_locale)
4462 && IN_LC_RUNTIME(LC_CTYPE))
4463 {
4464 if ( UNLIKELY(remove_dot_above)
4465 && memBEGINs(tmpbuf, sizeof(tmpbuf), COMBINING_DOT_ABOVE_UTF8))
4466 {
4467 s += u;
4468 remove_dot_above = FALSE;
4469 continue;
4470 }
4471 else if (UNLIKELY(*s == 'I' && tmpbuf[0] == 'i')) {
4472 remove_dot_above = TRUE;
4473 }
4474 }
130c5df3 4475#else
be42d347
KW
4476 PERL_UNUSED_VAR(remove_dot_above);
4477
a1a5ec35 4478 _toLOWER_utf8_flags(s, send, tmpbuf, &ulen, 0);
130c5df3 4479#endif
00f254e2 4480
a8e41ef4
KW
4481 /* Here is where we would do context-sensitive actions for the
4482 * Greek final sigma. See the commit message for 86510fb15 for why
4483 * there isn't any */
00f254e2 4484
06b5486a 4485 if (ulen > u && (SvLEN(dest) < (min += ulen - u))) {
fdb34c52 4486
06b5486a
KW
4487 /* If the eventually required minimum size outgrows the
4488 * available space, we need to grow. */
4489 const UV o = d - (U8*)SvPVX_const(dest);
fdb34c52 4490
06b5486a
KW
4491 /* If someone lowercases one million U+0130s we SvGROW() one
4492 * million times. Or we could try guessing how much to
4493 * allocate without allocating too much. Such is life.
4494 * Another option would be to grow an extra byte or two more
4495 * each time we need to grow, which would cut down the million
4496 * to 500K, with little waste */
10656159 4497 d = o + (U8*) SvGROW(dest, min);
06b5486a 4498 }
86510fb1 4499
06b5486a
KW
4500 /* Copy the newly lowercased letter to the output buffer we're
4501 * building */
4502 Copy(tmpbuf, d, ulen, U8);
4503 d += ulen;
4504 s += u;
00f254e2 4505 } /* End of looping through the source string */
ec9af7d4
NC
4506 SvUTF8_on(dest);
4507 *d = '\0';
4508 SvCUR_set(dest, d - (U8*)SvPVX_const(dest));
d595c8d9 4509 } else { /* 'source' not utf8 */
31351b04 4510 if (len) {
ec9af7d4 4511 const U8 *const send = s + len;
00f254e2
KW
4512
4513 /* Use locale casing if in locale; regular style if not treating
4514 * latin1 as having case; otherwise the latin1 casing. Do the
4515 * whole thing in a tight loop, for speed, */
130c5df3 4516#ifdef USE_LOCALE_CTYPE
d6ded950 4517 if (IN_LC_RUNTIME(LC_CTYPE)) {
be42d347
KW
4518 if (LIKELY( ! has_turkic_I)) {
4519 for (; s < send; d++, s++)
4520 *d = toLOWER_LC(*s);
4521 }
4522 else { /* This is the only case where lc() converts 'dest'
4523 into UTF-8 from a non-UTF-8 'source' */
4524 for (; s < send; s++) {
4525 if (*s == 'I') {
4526 *d++ = UTF8_TWO_BYTE_HI(LATIN_SMALL_LETTER_DOTLESS_I);
4527 *d++ = UTF8_TWO_BYTE_LO(LATIN_SMALL_LETTER_DOTLESS_I);
4528 }
4529 else {
4530 append_utf8_from_native_byte(toLOWER_LATIN1(*s), &d);
4531 }
4532 }
4533 }
445bf929 4534 }
130c5df3
KW
4535 else
4536#endif
4537 if (! IN_UNI_8_BIT) {
00f254e2 4538 for (; s < send; d++, s++) {
ec9af7d4 4539 *d = toLOWER(*s);
00f254e2
KW
4540 }
4541 }
4542 else {
4543 for (; s < send; d++, s++) {
4544 *d = toLOWER_LATIN1(*s);
4545 }
31351b04 4546 }
bbce6d69 4547 }
ec9af7d4
NC
4548 if (source != dest) {
4549 *d = '\0';
4550 SvCUR_set(dest, d - (U8*)SvPVX_const(dest));
4551 }
79072805 4552 }
130c5df3 4553#ifdef USE_LOCALE_CTYPE
d6ded950 4554 if (IN_LC_RUNTIME(LC_CTYPE)) {
445bf929
KW
4555 TAINT;
4556 SvTAINTED_on(dest);
4557 }
130c5df3 4558#endif
539689e7
FC
4559 if (dest != source && SvTAINTED(source))
4560 SvTAINT(dest);
ec9af7d4 4561 SvSETMAGIC(dest);
3cb4e04f 4562 return NORMAL;
79072805
LW
4563}
4564
a0d0e21e 4565PP(pp_quotemeta)
79072805 4566{
20b7effb 4567 dSP; dTARGET;
1b6737cc 4568 SV * const sv = TOPs;
a0d0e21e 4569 STRLEN len;
eb578fdb 4570 const char *s = SvPV_const(sv,len);
79072805 4571
7e2040f0 4572 SvUTF8_off(TARG); /* decontaminate */
a0d0e21e 4573 if (len) {
eb578fdb 4574 char *d;
862a34c6 4575 SvUPGRADE(TARG, SVt_PV);
c07a80fd 4576 SvGROW(TARG, (len * 2) + 1);
a0d0e21e 4577 d = SvPVX(TARG);
7e2040f0 4578 if (DO_UTF8(sv)) {
0dd2cdef 4579 while (len) {
29050de5 4580 STRLEN ulen = UTF8SKIP(s);
2e2b2571
KW
4581 bool to_quote = FALSE;
4582
4583 if (UTF8_IS_INVARIANT(*s)) {
4584 if (_isQUOTEMETA(*s)) {
4585 to_quote = TRUE;
4586 }
4587 }
042d9e50 4588 else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(s, s + len)) {
3fea7d29 4589 if (
130c5df3 4590#ifdef USE_LOCALE_CTYPE
20adcf7c
KW
4591 /* In locale, we quote all non-ASCII Latin1 chars.
4592 * Otherwise use the quoting rules */
a8e41ef4 4593
3fea7d29
BF
4594 IN_LC_RUNTIME(LC_CTYPE)
4595 ||
4596#endif
a62b247b 4597 _isQUOTEMETA(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s + 1))))
2e2b2571
KW
4598 {
4599 to_quote = TRUE;
4600 }
4601 }
685289b5 4602 else if (is_QUOTEMETA_high(s)) {
2e2b2571
KW
4603 to_quote = TRUE;
4604 }
4605
4606 if (to_quote) {
4607 *d++ = '\\';
4608 }
29050de5
KW
4609 if (ulen > len)
4610 ulen = len;
4611 len -= ulen;
4612 while (ulen--)
4613 *d++ = *s++;
0dd2cdef 4614 }
7e2040f0 4615 SvUTF8_on(TARG);
0dd2cdef 4616 }
2e2b2571
KW
4617 else if (IN_UNI_8_BIT) {
4618 while (len--) {
4619 if (_isQUOTEMETA(*s))
4620 *d++ = '\\';
4621 *d++ = *s++;
4622 }
4623 }
0dd2cdef 4624 else {
2e2b2571
KW
4625 /* For non UNI_8_BIT (and hence in locale) just quote all \W
4626 * including everything above ASCII */
0dd2cdef 4627 while (len--) {
adfec831 4628 if (!isWORDCHAR_A(*s))
0dd2cdef
LW
4629 *d++ = '\\';
4630 *d++ = *s++;
4631 }
79072805 4632 }
a0d0e21e 4633 *d = '\0';
349d4f2f 4634 SvCUR_set(TARG, d - SvPVX_const(TARG));
3aa33fe5 4635 (void)SvPOK_only_UTF8(TARG);
79072805 4636 }
a0d0e21e
LW
4637 else
4638 sv_setpvn(TARG, s, len);
ec93b65f 4639 SETTARG;
cfe40115 4640 return NORMAL;
79072805
LW
4641}
4642
838f2281
BF
4643PP(pp_fc)
4644{
838f2281
BF
4645 dTARGET;
4646 dSP;
4647 SV *source = TOPs;
4648 STRLEN len;
4649 STRLEN min;
4650 SV *dest;
4651 const U8 *s;
4652 const U8 *send;
4653 U8 *d;
bfac13d4 4654 U8 tmpbuf[UTF8_MAXBYTES_CASE + 1];
9b63e895
KW
4655#if UNICODE_MAJOR_VERSION > 3 /* no multifolds in early Unicode */ \
4656 || (UNICODE_MAJOR_VERSION == 3 && ( UNICODE_DOT_VERSION > 0) \
4657 || UNICODE_DOT_DOT_VERSION > 0)
a4b69695
KW
4658 const bool full_folding = TRUE; /* This variable is here so we can easily
4659 move to more generality later */
9b63e895
KW
4660#else
4661 const bool full_folding = FALSE;
4662#endif
838f2281 4663 const U8 flags = ( full_folding ? FOLD_FLAGS_FULL : 0 )
130c5df3
KW
4664#ifdef USE_LOCALE_CTYPE
4665 | ( IN_LC_RUNTIME(LC_CTYPE) ? FOLD_FLAGS_LOCALE : 0 )
4666#endif
4667 ;
838f2281
BF
4668
4669 /* This is a facsimile of pp_lc, but with a thousand bugs thanks to me.
4670 * You are welcome(?) -Hugmeir
4671 */
4672
4673 SvGETMAGIC(source);
4674
4675 dest = TARG;
4676
4677 if (SvOK(source)) {
4678 s = (const U8*)SvPV_nomg_const(source, len);
4679 } else {
4680 if (ckWARN(WARN_UNINITIALIZED))
4681 report_uninit(source);
4682 s = (const U8*)"";
4683 len = 0;
4684 }
4685
4686 min = len + 1;
4687
4688 SvUPGRADE(dest, SVt_PV);
4689 d = (U8*)SvGROW(dest, min);
4690 (void)SvPOK_only(dest);
4691
4692 SETs(dest);
4693
4694 send = s + len;
8b7358b9
KW
4695
4696#ifdef USE_LOCALE_CTYPE
4697
4698 if ( IN_LC_RUNTIME(LC_CTYPE) ) { /* Under locale */
4699 _CHECK_AND_WARN_PROBLEMATIC_LOCALE;
4700 }
4701
4702#endif
4703
838f2281 4704 if (DO_UTF8(source)) { /* UTF-8 flagged string. */
838f2281
BF
4705 while (s < send) {
4706 const STRLEN u = UTF8SKIP(s);
4707 STRLEN ulen;
4708
a1a5ec35 4709 _toFOLD_utf8_flags(s, send, tmpbuf, &ulen, flags);
838f2281
BF
4710
4711 if (ulen > u && (SvLEN(dest) < (min += ulen - u))) {
4712 const UV o = d - (U8*)SvPVX_const(dest);
10656159 4713 d = o + (U8*) SvGROW(dest, min);
838f2281
BF
4714 }
4715
4716 Copy(tmpbuf, d, ulen, U8);
4717 d += ulen;
4718 s += u;
4719 }
4720 SvUTF8_on(dest);
838f2281 4721 } /* Unflagged string */
0902dd32 4722 else if (len) {
130c5df3 4723#ifdef USE_LOCALE_CTYPE
d6ded950 4724 if ( IN_LC_RUNTIME(LC_CTYPE) ) { /* Under locale */
31f05a37
KW
4725 if (IN_UTF8_CTYPE_LOCALE) {
4726 goto do_uni_folding;
4727 }
838f2281 4728 for (; s < send; d++, s++)
ea36a843 4729 *d = (U8) toFOLD_LC(*s);
838f2281 4730 }
130c5df3
KW
4731 else
4732#endif
4733 if ( !IN_UNI_8_BIT ) { /* Under nothing, or bytes */
838f2281 4734 for (; s < send; d++, s++)
d22b930b 4735 *d = toFOLD(*s);
838f2281
BF
4736 }
4737 else {
91191cf7 4738#ifdef USE_LOCALE_CTYPE
31f05a37 4739 do_uni_folding:
91191cf7 4740#endif
be42d347 4741 /* For ASCII and the Latin-1 range, there's potentially three
a8e41ef4
KW
4742 * troublesome folds:
4743 * \x{DF} (\N{LATIN SMALL LETTER SHARP S}), which under full
4744 * casefolding becomes 'ss';
4745 * \x{B5} (\N{MICRO SIGN}), which under any fold becomes
4746 * \x{3BC} (\N{GREEK SMALL LETTER MU})
be42d347
KW
4747 * I only in Turkic locales, this folds to \x{131}
4748 * \N{LATIN SMALL LETTER DOTLESS I}
d14578b8 4749 * For the rest, the casefold is their lowercase. */
838f2281 4750 for (; s < send; d++, s++) {
be42d347
KW
4751 if ( UNLIKELY(*s == MICRO_SIGN)
4752#ifdef USE_LOCALE_CTYPE
4753 || ( UNLIKELY(PL_in_utf8_turkic_locale)
4754 && UNLIKELY(IN_LC_RUNTIME(LC_CTYPE))
4755 && UNLIKELY(*s == 'I'))
4756#endif
4757 ) {
2f8f985a
KW
4758 Size_t extra = send - s
4759 + variant_under_utf8_count(s, send);
4760
d14578b8 4761 /* \N{MICRO SIGN}'s casefold is \N{GREEK SMALL LETTER MU},
be42d347
KW
4762 * and 'I' in Turkic locales is \N{LATIN SMALL LETTER
4763 * DOTLESS I} both of which are outside of the latin-1
4764 * range. There's a couple of ways to deal with this -- khw
4765 * discusses them in pp_lc/uc, so go there :) What we do
4766 * here is upgrade what we had already casefolded, then
4767 * enter an inner loop that appends the rest of the
4768 * characters as UTF-8.
2f8f985a
KW
4769 *
4770 * First we calculate the needed size of the upgraded dest
4771 * beyond what's been processed already (the upgrade
be42d347
KW
4772 * function figures that out). Except for the 'I' in
4773 * Turkic locales, in UTF-8 strings, the fold case of a
2f8f985a
KW
4774 * character below 256 occupies the same number of bytes as
4775 * the original (even the Sharp S). Therefore, the space
4776 * needed is the number of bytes remaining plus the number
4777 * of characters that become two bytes when converted to
be42d347
KW
4778 * UTF-8 plus, in turkish locales, the number of 'I's */
4779
4780 if (UNLIKELY(*s == 'I')) {
4781 const U8 * s_peek = s;
4782
4783 do {
4784 extra++;
4785
4786 s_peek = (U8 *) memchr(s_peek + 1, 'i',
4787 send - (s_peek + 1));
4788 } while (s_peek != NULL);
4789 }
2f8f985a
KW
4790
4791 /* Growing may move things, so have to save and recalculate
4792 * 'd' */
838f2281
BF
4793 len = d - (U8*)SvPVX_const(dest);
4794 SvCUR_set(dest, len);
4795 len = sv_utf8_upgrade_flags_grow(dest,
4796 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
56e36cbf
KW
4797 extra
4798 + 1 /* Trailing NUL */ );
838f2281
BF
4799 d = (U8*)SvPVX(dest) + len;
4800
93327b75
KW
4801 *d++ = UTF8_TWO_BYTE_HI(GREEK_SMALL_LETTER_MU);
4802 *d++ = UTF8_TWO_BYTE_LO(GREEK_SMALL_LETTER_MU);
838f2281 4803 s++;
a8e41ef4 4804
838f2281
BF
4805 for (; s < send; s++) {
4806 STRLEN ulen;
526f8cbf
KW
4807 _to_uni_fold_flags(*s, d, &ulen, flags);
4808 d += ulen;
838f2281
BF
4809 }
4810 break;
4811 }
ca62a7c2
KW
4812 else if ( UNLIKELY(*s == LATIN_SMALL_LETTER_SHARP_S)
4813 && full_folding)
4814 {
d14578b8
KW
4815 /* Under full casefolding, LATIN SMALL LETTER SHARP S
4816 * becomes "ss", which may require growing the SV. */
838f2281
BF
4817 if (SvLEN(dest) < ++min) {
4818 const UV o = d - (U8*)SvPVX_const(dest);
10656159 4819 d = o + (U8*) SvGROW(dest, min);
838f2281
BF
4820 }
4821 *(d)++ = 's';
4822 *d = 's';
4823 }
a8e41ef4 4824 else { /* Else, the fold is the lower case */
838f2281
BF
4825 *d = toLOWER_LATIN1(*s);
4826 }
4827 }
4828 }
4829 }
4830 *d = '\0';
4831 SvCUR_set(dest, d - (U8*)SvPVX_const(dest));
4832
130c5df3 4833#ifdef USE_LOCALE_CTYPE
d6ded950 4834 if (IN_LC_RUNTIME(LC_CTYPE)) {
445bf929
KW
4835 TAINT;
4836 SvTAINTED_on(dest);
4837 }
130c5df3 4838#endif
838f2281
BF
4839 if (SvTAINTED(source))
4840 SvTAINT(dest);
4841 SvSETMAGIC(dest);
4842 RETURN;
4843}
4844
a0d0e21e 4845/* Arrays. */
79072805 4846
a0d0e21e 4847PP(pp_aslice)
79072805 4848{
20b7effb 4849 dSP; dMARK; dORIGMARK;
eb578fdb
KW
4850 AV *const av = MUTABLE_AV(POPs);
4851 const I32 lval = (PL_op->op_flags & OPf_MOD || LVRET);
79072805 4852
a0d0e21e 4853 if (SvTYPE(av) == SVt_PVAV) {
4ad10a0b
VP
4854 const bool localizing = PL_op->op_private & OPpLVAL_INTRO;
4855 bool can_preserve = FALSE;
4856
4857 if (localizing) {
4858 MAGIC *mg;
4859 HV *stash;
4860
4861 can_preserve = SvCANEXISTDELETE(av);
4862 }
4863
4864 if (lval && localizing) {
eb578fdb 4865 SV **svp;
c70927a6 4866 SSize_t max = -1;
924508f0 4867 for (svp = MARK + 1; svp <= SP; svp++) {
c70927a6 4868 const SSize_t elem = SvIV(*svp);
748a9306
LW
4869 if (elem > max)
4870 max = elem;
4871 }
4872 if (max > AvMAX(av))
4873 av_extend(av, max);
4874 }
4ad10a0b 4875
a0d0e21e 4876 while (++MARK <= SP) {
eb578fdb 4877 SV **svp;
c70927a6 4878 SSize_t elem = SvIV(*MARK);
4ad10a0b 4879 bool preeminent = TRUE;
a0d0e21e 4880
4ad10a0b
VP
4881 if (localizing && can_preserve) {
4882 /* If we can determine whether the element exist,
4883 * Try to preserve the existenceness of a tied array
4884 * element by using EXISTS and DELETE if possible.
4885 * Fallback to FETCH and STORE otherwise. */
4886 preeminent = av_exists(av, elem);
4887 }
4888
a0d0e21e
LW
4889 svp = av_fetch(av, elem, lval);
4890 if (lval) {
ce0d59fd 4891 if (!svp || !*svp)
cea2e8a9 4892 DIE(aTHX_ PL_no_aelem, elem);
4ad10a0b
VP
4893 if (localizing) {
4894 if (preeminent)
4895 save_aelem(av, elem, svp);
4896 else
4897 SAVEADELETE(av, elem);
4898 }
79072805 4899 }
3280af22 4900 *MARK = svp ? *svp : &PL_sv_undef;
79072805
LW
4901 }
4902 }
82334630 4903 if (GIMME_V != G_ARRAY) {
a0d0e21e 4904 MARK = ORIGMARK;
04ab2c87 4905 *++MARK = SP > ORIGMARK ? *SP : &PL_sv_undef;
a0d0e21e
LW
4906 SP = MARK;
4907 }
79072805
LW
4908 RETURN;
4909}
4910
6dd3e0f2
RZ
4911PP(pp_kvaslice)
4912{
20b7effb 4913 dSP; dMARK;
6dd3e0f2
RZ
4914 AV *const av = MUTABLE_AV(POPs);
4915 I32 lval = (PL_op->op_flags & OPf_MOD);
adad97db 4916 SSize_t items = SP - MARK;
6dd3e0f2
RZ
4917
4918 if (PL_op->op_private & OPpMAYBE_LVSUB) {
4919 const I32 flags = is_lvalue_sub();
4920 if (flags) {
4921 if (!(flags & OPpENTERSUB_INARGS))
7aae0299 4922 /* diag_listed_as: Can't modify %s in %s */
6dd3e0f2
RZ
4923 Perl_croak(aTHX_ "Can't modify index/value array slice in list assignment");
4924 lval = flags;
4925 }
4926 }
4927
4928 MEXTEND(SP,items);
4929 while (items > 1) {
4930 *(MARK+items*2-1) = *(MARK+items);
4931 items--;
4932 }
4933 items = SP-MARK;
4934 SP += items;
4935
4936 while (++MARK <= SP) {
4937 SV **svp;
4938
4939 svp = av_fetch(av, SvIV(*MARK), lval);
4940 if (lval) {
4941 if (!svp || !*svp || *svp == &PL_sv_undef) {
4942 DIE(aTHX_ PL_no_aelem, SvIV(*MARK));
4943 }
4944 *MARK = sv_mortalcopy(*MARK);
4945 }
4946 *++MARK = svp ? *svp : &PL_sv_undef;
4947 }
82334630 4948 if (GIMME_V != G_ARRAY) {
6dd3e0f2
RZ
4949 MARK = SP - items*2;
4950 *++MARK = items > 0 ? *SP : &PL_sv_undef;
4951 SP = MARK;
4952 }
4953 RETURN;
4954}
4955
b1c05ba5 4956
878d132a
NC
4957PP(pp_aeach)
4958{
878d132a 4959 dSP;
502c6561 4960 AV *array = MUTABLE_AV(POPs);
1c23e2bd 4961 const U8 gimme = GIMME_V;
453d94a9 4962 IV *iterp = Perl_av_iter_p(aTHX_ array);
878d132a
NC
4963 const IV current = (*iterp)++;
4964
b9f2b683 4965 if (current > av_tindex(array)) {
878d132a
NC
4966 *iterp = 0;
4967 if (gimme == G_SCALAR)
4968 RETPUSHUNDEF;
4969 else
4970 RETURN;
4971 }
4972
4973 EXTEND(SP, 2);
e1dccc0d 4974 mPUSHi(current);
878d132a
NC
4975 if (gimme == G_ARRAY) {
4976 SV **const element = av_fetch(array, current, 0);
4977 PUSHs(element ? *element : &PL_sv_undef);
4978 }
4979 RETURN;
4980}
4981
b1c05ba5 4982/* also used for: pp_avalues()*/
878d132a
NC
4983PP(pp_akeys)
4984{
878d132a 4985 dSP;
502c6561 4986 AV *array = MUTABLE_AV(POPs);
1c23e2bd 4987 const U8 gimme = GIMME_V;
878d132a
NC
4988
4989 *Perl_av_iter_p(aTHX_ array) = 0;
4990
4991 if (gimme == G_SCALAR) {
4992 dTARGET;
b9f2b683 4993 PUSHi(av_tindex(array) + 1);
878d132a
NC
4994 }
4995 else if (gimme == G_ARRAY) {
738155d2
FC
4996 if (UNLIKELY(PL_op->op_private & OPpMAYBE_LVSUB)) {
4997 const I32 flags = is_lvalue_sub();
4998 if (flags && !(flags & OPpENTERSUB_INARGS))
4999 /* diag_listed_as: Can't modify %s in %s */
5000 Perl_croak(aTHX_
5001 "Can't modify keys on array in list assignment");
5002 }
5003 {
878d132a 5004 IV n = Perl_av_len(aTHX_ array);
e1dccc0d 5005 IV i;
878d132a
NC
5006
5007 EXTEND(SP, n + 1);
5008
73665bc4
FC
5009 if ( PL_op->op_type == OP_AKEYS
5010 || ( PL_op->op_type == OP_AVHVSWITCH
cd642408 5011 && (PL_op->op_private & 3) + OP_AEACH == OP_AKEYS ))
73665bc4 5012 {
e1dccc0d 5013 for (i = 0; i <= n; i++) {
878d132a
NC
5014 mPUSHi(i);
5015 }
5016 }
5017 else {
5018 for (i = 0; i <= n; i++) {
5019 SV *const *const elem = Perl_av_fetch(aTHX_ array, i, 0);
5020 PUSHs(elem ? *elem : &PL_sv_undef);
5021 }
5022 }
738155d2 5023 }
878d132a
NC
5024 }
5025 RETURN;
5026}
5027
79072805
LW
5028/* Associative arrays. */
5029
5030PP(pp_each)
5031{
39644a26 5032 dSP;
85fbaab2 5033 HV * hash = MUTABLE_HV(POPs);
c07a80fd 5034 HE *entry;
1c23e2bd 5035 const U8 gimme = GIMME_V;
8ec5e241 5036
6d822dc4 5037 entry = hv_iternext(hash);
79072805 5038
79072805
LW
5039 EXTEND(SP, 2);
5040 if (entry) {
1b6737cc 5041 SV* const sv = hv_iterkeysv(entry);
2b32fed8 5042 PUSHs(sv);
54310121 5043 if (gimme == G_ARRAY) {
59af0135 5044 SV *val;
6d822dc4 5045 val = hv_iterval(hash, entry);
59af0135 5046 PUSHs(val);
79072805 5047 }
79072805 5048 }
54310121 5049 else if (gimme == G_SCALAR)
79072805
LW
5050 RETPUSHUNDEF;
5051
5052 RETURN;
5053}
5054
7332a6c4
VP
5055STATIC OP *
5056S_do_delete_local(pTHX)
79072805 5057{
39644a26 5058 dSP;
1c23e2bd 5059 const U8 gimme = GIMME_V;
7332a6c4
VP
5060 const MAGIC *mg;
5061 HV *stash;
ca3f996a 5062 const bool sliced = !!(PL_op->op_private & OPpSLICE);
626040f7 5063 SV **unsliced_keysv = sliced ? NULL : sp--;
ca3f996a 5064 SV * const osv = POPs;
626040f7 5065 SV **mark = sliced ? PL_stack_base + POPMARK : unsliced_keysv-1;
ca3f996a
FC
5066 dORIGMARK;
5067 const bool tied = SvRMAGICAL(osv)
7332a6c4 5068 && mg_find((const SV *)osv, PERL_MAGIC_tied);
ca3f996a
FC
5069 const bool can_preserve = SvCANEXISTDELETE(osv);
5070 const U32 type = SvTYPE(osv);
626040f7 5071 SV ** const end = sliced ? SP : unsliced_keysv;
ca3f996a
FC
5072
5073 if (type == SVt_PVHV) { /* hash element */
7332a6c4 5074 HV * const hv = MUTABLE_HV(osv);
ca3f996a 5075 while (++MARK <= end) {
7332a6c4
VP
5076 SV * const keysv = *MARK;
5077 SV *sv = NULL;
5078 bool preeminent = TRUE;
5079 if (can_preserve)
5080 preeminent = hv_exists_ent(hv, keysv, 0);
5081 if (tied) {
5082 HE *he = hv_fetch_ent(hv, keysv, 1, 0);
5083 if (he)
5084 sv = HeVAL(he);
5085 else
5086 preeminent = FALSE;
5087 }
5088 else {
5089 sv = hv_delete_ent(hv, keysv, 0, 0);
9332b95f
FC
5090 if (preeminent)
5091 SvREFCNT_inc_simple_void(sv); /* De-mortalize */
7332a6c4
VP
5092 }
5093 if (preeminent) {
be6064fd 5094 if (!sv) DIE(aTHX_ PL_no_helem_sv, SVfARG(keysv));
7332a6c4
VP
5095 save_helem_flags(hv, keysv, &sv, SAVEf_KEEPOLDELEM);
5096 if (tied) {
5097 *MARK = sv_mortalcopy(sv);
5098 mg_clear(sv);
5099 } else
5100 *MARK = sv;
5101 }
5102 else {
5103 SAVEHDELETE(hv, keysv);
5104 *MARK = &PL_sv_undef;
5105 }
5106 }
ca3f996a
FC
5107 }
5108 else if (type == SVt_PVAV) { /* array element */
7332a6c4
VP
5109 if (PL_op->op_flags & OPf_SPECIAL) {
5110 AV * const av = MUTABLE_AV(osv);
ca3f996a 5111 while (++MARK <= end) {
c70927a6 5112 SSize_t idx = SvIV(*MARK);
7332a6c4
VP
5113 SV *sv = NULL;
5114 bool preeminent = TRUE;
5115 if (can_preserve)
5116 preeminent = av_exists(av, idx);
5117 if (tied) {
5118 SV **svp = av_fetch(av, idx, 1);
5119 if (svp)
5120 sv = *svp;
5121 else
5122 preeminent = FALSE;
5123 }
5124 else {
5125 sv = av_delete(av, idx, 0);
9332b95f
FC
5126 if (preeminent)
5127 SvREFCNT_inc_simple_void(sv); /* De-mortalize */
7332a6c4
VP
5128 }
5129 if (preeminent) {
5130 save_aelem_flags(av, idx, &sv, SAVEf_KEEPOLDELEM);
5131 if (tied) {
5132 *MARK = sv_mortalcopy(sv);
5133 mg_clear(sv);
5134 } else
5135 *MARK = sv;
5136 }
5137 else {
5138 SAVEADELETE(av, idx);
5139 *MARK = &PL_sv_undef;
5140 }
5141 }
5142 }
ca3f996a
FC
5143 else
5144 DIE(aTHX_ "panic: avhv_delete no longer supported");
5145 }
5146 else
7332a6c4 5147 DIE(aTHX_ "Not a HASH reference");
ca3f996a 5148 if (sliced) {
7332a6c4
VP
5149 if (gimme == G_VOID)
5150 SP = ORIGMARK;
5151 else if (gimme == G_SCALAR) {
5152 MARK = ORIGMARK;
5153 if (SP > MARK)
5154 *++MARK = *SP;
5155 else
5156 *++MARK = &PL_sv_undef;
5157 SP = MARK;
5158 }
5159 }
ca3f996a 5160 else if (gimme != G_VOID)
626040f7 5161 PUSHs(*unsliced_keysv);
7332a6c4
VP
5162
5163 RETURN;
5164}
5165
5166PP(pp_delete)
5167{
7332a6c4 5168 dSP;
1c23e2bd 5169 U8 gimme;
7332a6c4
VP
5170 I32 discard;
5171
5172 if (PL_op->op_private & OPpLVAL_INTRO)
5173 return do_delete_local();
5174
5175 gimme = GIMME_V;
5176 discard = (gimme == G_VOID) ? G_DISCARD : 0;
5f05dabc 5177
cc0776d6 5178 if (PL_op->op_private & (OPpSLICE|OPpKVSLICE)) {
5f05dabc 5179 dMARK; dORIGMARK;
85fbaab2 5180 HV * const hv = MUTABLE_HV(POPs);
1b6737cc 5181 const U32 hvtype = SvTYPE(hv);
cc0776d6
DIM
5182 int skip = 0;
5183 if (PL_op->op_private & OPpKVSLICE) {
5184 SSize_t items = SP - MARK;
5185
5186 MEXTEND(SP,items);
5187 while (items > 1) {
5188 *(MARK+items*2-1) = *(MARK+items);
5189 items--;
5190 }
5191 items = SP - MARK;
5192 SP += items;
5193 skip = 1;
5194 }
01020589 5195 if (hvtype == SVt_PVHV) { /* hash element */
cc0776d6
DIM
5196 while ((MARK += (1+skip)) <= SP) {
5197 SV * const sv = hv_delete_ent(hv, *(MARK-skip), discard, 0);
01020589
GS
5198 *MARK = sv ? sv : &PL_sv_undef;
5199 }
5f05dabc 5200 }
6d822dc4
MS
5201 else if (hvtype == SVt_PVAV) { /* array element */
5202 if (PL_op->op_flags & OPf_SPECIAL) {
cc0776d6
DIM
5203 while ((MARK += (1+skip)) <= SP) {
5204 SV * const sv = av_delete(MUTABLE_AV(hv), SvIV(*(MARK-skip)), discard);
6d822dc4
MS
5205 *MARK = sv ? sv : &PL_sv_undef;
5206 }
5207 }
01020589
GS
5208 }
5209 else
5210 DIE(aTHX_ "Not a HASH reference");
54310121 5211 if (discard)
5212 SP = ORIGMARK;
5213 else if (gimme == G_SCALAR) {
5f05dabc 5214 MARK = ORIGMARK;
9111c9c0
DM
5215 if (SP > MARK)
5216 *++MARK = *SP;
5217 else
5218 *++MARK = &PL_sv_undef;
5f05dabc 5219 SP = MARK;
5220 }
5221 }
5222 else {
5223 SV *keysv = POPs;
85fbaab2 5224 HV * const hv = MUTABLE_HV(POPs);
295d248e 5225 SV *sv = NULL;
97fcbf96
MB
5226 if (SvTYPE(hv) == SVt_PVHV)
5227 sv = hv_delete_ent(hv, keysv, discard, 0);
01020589
GS
5228 else if (SvTYPE(hv) == SVt_PVAV) {
5229 if (PL_op->op_flags & OPf_SPECIAL)
502c6561 5230 sv = av_delete(MUTABLE_AV(hv), SvIV(keysv), discard);
af288a60
HS
5231 else
5232 DIE(aTHX_ "panic: avhv_delete no longer supported");
01020589 5233 }
97fcbf96 5234 else
cea2e8a9 5235 DIE(aTHX_ "Not a HASH reference");
5f05dabc 5236 if (!sv)
3280af22 5237 sv = &PL_sv_undef;
54310121 5238 if (!discard)
5239 PUSHs(sv);
79072805 5240 }
79072805
LW
5241 RETURN;
5242}
5243
a0d0e21e 5244PP(pp_exists)
79072805 5245{
39644a26 5246 dSP;
afebc493
GS
5247 SV *tmpsv;
5248 HV *hv;
5249
c7e88ff3 5250 if (UNLIKELY( PL_op->op_private & OPpEXISTS_SUB )) {
afebc493 5251 GV *gv;
0bd48802 5252 SV * const sv = POPs;
f2c0649b 5253 CV * const cv = sv_2cv(sv, &hv, &gv, 0);
afebc493
GS
5254 if (cv)
5255 RETPUSHYES;
5256 if (gv && isGV(gv) && GvCV(gv) && !GvCVGEN(gv))
5257 RETPUSHYES;
5258 RETPUSHNO;
5259 }
5260 tmpsv = POPs;
85fbaab2 5261 hv = MUTABLE_HV(POPs);
c7e88ff3 5262 if (LIKELY( SvTYPE(hv) == SVt_PVHV )) {
ae77835f 5263 if (hv_exists_ent(hv, tmpsv, 0))
c750a3ec 5264 RETPUSHYES;
ef54e1a4
JH
5265 }
5266 else if (SvTYPE(hv) == SVt_PVAV) {
01020589 5267 if (PL_op->op_flags & OPf_SPECIAL) { /* array element */
502c6561 5268 if (av_exists(MUTABLE_AV(hv), SvIV(tmpsv)))
01020589
GS
5269 RETPUSHYES;
5270 }
ef54e1a4
JH
5271 }
5272 else {
cea2e8a9 5273 DIE(aTHX_ "Not a HASH reference");
a0d0e21e 5274 }
a0d0e21e
LW
5275 RETPUSHNO;
5276}
79072805 5277
a0d0e21e
LW
5278PP(pp_hslice)
5279{
20b7effb 5280 dSP; dMARK; dORIGMARK;
eb578fdb
KW
5281 HV * const hv = MUTABLE_HV(POPs);
5282 const I32 lval = (PL_op->op_flags & OPf_MOD || LVRET);
1b6737cc 5283 const bool localizing = PL_op->op_private & OPpLVAL_INTRO;
d30e492c 5284 bool can_preserve = FALSE;
79072805 5285
eb85dfd3
DM
5286 if (localizing) {
5287 MAGIC *mg;
5288 HV *stash;
5289
2c5f48c2 5290 if (SvCANEXISTDELETE(hv))
d30e492c 5291 can_preserve = TRUE;
eb85dfd3
DM
5292 }
5293
6d822dc4 5294 while (++MARK <= SP) {
1b6737cc 5295 SV * const keysv = *MARK;
6d822dc4
MS
5296 SV **svp;
5297 HE *he;
d30e492c
VP
5298 bool preeminent = TRUE;
5299
5300 if (localizing && can_preserve) {
5301 /* If we can determine whether the element exist,
5302 * try to preserve the existenceness of a tied hash
5303 * element by using EXISTS and DELETE if possible.
5304 * Fallback to FETCH and STORE otherwise. */
5305 preeminent = hv_exists_ent(hv, keysv, 0);
6d822dc4 5306 }
eb85dfd3 5307
6d822dc4 5308 he = hv_fetch_ent(hv, keysv, lval, 0);
fe5bfecd 5309 svp = he ? &HeVAL(he) : NULL;
eb85dfd3 5310
6d822dc4 5311 if (lval) {
746f6409 5312 if (!svp || !*svp || *svp == &PL_sv_undef) {
be2597df 5313 DIE(aTHX_ PL_no_helem_sv, SVfARG(keysv));
6d822dc4
MS
5314 }
5315 if (localizing) {
6881372e 5316 if (HvNAME_get(hv) && isGV_or_RVCV(*svp))
159b6efe 5317 save_gp(MUTABLE_GV(*svp), !(PL_op->op_flags & OPf_SPECIAL));
47cfc530
VP
5318 else if (preeminent)
5319 save_helem_flags(hv, keysv, svp,
5320 (PL_op->op_flags & OPf_SPECIAL) ? 0 : SAVEf_SETMAGIC);
5321 else
5322 SAVEHDELETE(hv, keysv);
6d822dc4
MS
5323 }
5324 }
746f6409 5325 *MARK = svp && *svp ? *svp : &PL_sv_undef;
79072805 5326 }
82334630 5327 if (GIMME_V != G_ARRAY) {
a0d0e21e 5328 MARK = ORIGMARK;
04ab2c87 5329 *++MARK = SP > ORIGMARK ? *SP : &PL_sv_undef;
a0d0e21e 5330 SP = MARK;
79072805 5331 }
a0d0e21e
LW
5332 RETURN;
5333}
5334
5cae3edb
RZ
5335PP(pp_kvhslice)
5336{
20b7effb 5337 dSP; dMARK;
5cae3edb
RZ
5338 HV * const hv = MUTABLE_HV(POPs);
5339 I32 lval = (PL_op->op_flags & OPf_MOD);
adad97db 5340 SSize_t items = SP - MARK;
5cae3edb
RZ
5341
5342 if (PL_op->op_private & OPpMAYBE_LVSUB) {
5343 const I32 flags = is_lvalue_sub();
5344 if (flags) {
5345 if (!(flags & OPpENTERSUB_INARGS))
7aae0299 5346 /* diag_listed_as: Can't modify %s in %s */
cc5f9b8a
FC
5347 Perl_croak(aTHX_ "Can't modify key/value hash slice in %s assignment",
5348 GIMME_V == G_ARRAY ? "list" : "scalar");
5cae3edb
RZ
5349 lval = flags;
5350 }
5351 }
5352
5353 MEXTEND(SP,items);
5354 while (items > 1) {
5355 *(MARK+items*2-1) = *(MARK+items);
5356 items--;
5357 }
5358 items = SP-MARK;
5359 SP += items;
5360
5361 while (++MARK <= SP) {
5362 SV * const keysv = *MARK;
5363 SV **svp;
5364 HE *he;
5365
5366 he = hv_fetch_ent(hv, keysv, lval, 0);
5367 svp = he ? &HeVAL(he) : NULL;
5368
5369 if (lval) {
5370 if (!svp || !*svp || *svp == &PL_sv_undef) {
5371 DIE(aTHX_ PL_no_helem_sv, SVfARG(keysv));
5372 }
5373 *MARK = sv_mortalcopy(*MARK);
5374 }
5375 *++MARK = svp && *svp ? *svp : &PL_sv_undef;
5376 }
82334630 5377 if (GIMME_V != G_ARRAY) {
5cae3edb
RZ
5378 MARK = SP - items*2;
5379 *++MARK = items > 0 ? *SP : &PL_sv_undef;
5380 SP = MARK;
5381 }
5382 RETURN;
5383}
5384
a0d0e21e
LW
5385/* List operators. */
5386
5387PP(pp_list)
5388{
4fa715fa 5389 I32 markidx = POPMARK;
82334630 5390 if (GIMME_V != G_ARRAY) {
57bd6600
TC
5391 /* don't initialize mark here, EXTEND() may move the stack */
5392 SV **mark;
4fa715fa 5393 dSP;
b54564c3 5394 EXTEND(SP, 1); /* in case no arguments, as in @empty */
57bd6600 5395 mark = PL_stack_base + markidx;
a0d0e21e
LW
5396 if (++MARK <= SP)
5397 *MARK = *SP; /* unwanted list, return last item */
8990e307 5398 else
3280af22 5399 *MARK = &PL_sv_undef;
a0d0e21e 5400 SP = MARK;
4fa715fa 5401 PUTBACK;
79072805 5402 }
4fa715fa 5403 return NORMAL;
79072805
LW
5404}
5405
a0d0e21e 5406PP(pp_lslice)
79072805 5407{
39644a26 5408 dSP;
1b6737cc
AL
5409 SV ** const lastrelem = PL_stack_sp;
5410 SV ** const lastlelem = PL_stack_base + POPMARK;
5411 SV ** const firstlelem = PL_stack_base + POPMARK + 1;
eb578fdb 5412 SV ** const firstrelem = lastlelem + 1;
706a6ebc 5413 const U8 mod = PL_op->op_flags & OPf_MOD;
1b6737cc 5414
eb578fdb
KW
5415 const I32 max = lastrelem - lastlelem;
5416 SV **lelem;
a0d0e21e 5417
82334630 5418 if (GIMME_V != G_ARRAY) {
9e59c36b 5419 if (lastlelem < firstlelem) {
7da51ead 5420 EXTEND(SP, 1);
9e59c36b
TC
5421 *firstlelem = &PL_sv_undef;
5422 }
5423 else {
5424 I32 ix = SvIV(*lastlelem);
5425 if (ix < 0)
5426 ix += max;
5427 if (ix < 0 || ix >= max)
5428 *firstlelem = &PL_sv_undef;
5429 else
5430 *firstlelem = firstrelem[ix];
5431 }
5432 SP = firstlelem;
5433 RETURN;
a0d0e21e
LW
5434 }
5435
5436 if (max == 0) {
5437 SP = firstlelem - 1;
5438 RETURN;
5439 }
5440
5441 for (lelem = firstlelem; lelem <= lastlelem; lelem++) {
4ea561bc 5442 I32 ix = SvIV(*lelem);
c73bf8e3 5443 if (ix < 0)
a0d0e21e 5444 ix += max;
c73bf8e3
HS
5445 if (ix < 0 || ix >= max)
5446 *lelem = &PL_sv_undef;
5447 else {
c73bf8e3 5448 if (!(*lelem = firstrelem[ix]))
3280af22 5449 *lelem = &PL_sv_undef;
60779a30 5450 else if (mod && SvPADTMP(*lelem)) {
706a6ebc 5451 *lelem = firstrelem[ix] = sv_mortalcopy(*lelem);
60779a30 5452 }
748a9306 5453 }
79072805 5454 }
cbce292e 5455 SP = lastlelem;
79072805
LW
5456 RETURN;
5457}
5458
a0d0e21e
LW
5459PP(pp_anonlist)
5460{
20b7effb 5461 dSP; dMARK;
1b6737cc 5462 const I32 items = SP - MARK;
ad64d0ec 5463 SV * const av = MUTABLE_SV(av_make(items, MARK+1));
31476221 5464 SP = MARK;
6e449a3a
MHM
5465 mXPUSHs((PL_op->op_flags & OPf_SPECIAL)
5466 ? newRV_noinc(av) : av);
a0d0e21e
LW
5467 RETURN;
5468}
5469
5470PP(pp_anonhash)
79072805 5471{
20b7effb 5472 dSP; dMARK; dORIGMARK;
67e67fd7 5473 HV* const hv = newHV();
8d455b9f 5474 SV* const retval = sv_2mortal( PL_op->op_flags & OPf_SPECIAL
67e67fd7 5475 ? newRV_noinc(MUTABLE_SV(hv))
8d455b9f 5476 : MUTABLE_SV(hv) );
a0d0e21e
LW
5477
5478 while (MARK < SP) {
3ed356df
FC
5479 SV * const key =
5480 (MARK++, SvGMAGICAL(*MARK) ? sv_mortalcopy(*MARK) : *MARK);
5481 SV *val;
a0d0e21e 5482 if (MARK < SP)
3ed356df
FC
5483 {
5484 MARK++;
5485 SvGETMAGIC(*MARK);
5486 val = newSV(0);
d187b712 5487 sv_setsv_nomg(val, *MARK);
3ed356df 5488 }
a2a5de95 5489 else
3ed356df 5490 {
a2a5de95 5491 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Odd number of elements in anonymous hash");
3ed356df
FC
5492 val = newSV(0);
5493 }
f12c7020 5494 (void)hv_store_ent(hv,key,val,0);
79072805 5495 }
a0d0e21e 5496 SP = ORIGMARK;
8d455b9f 5497 XPUSHs(retval);
79072805
LW
5498 RETURN;
5499}
5500
a0d0e21e 5501PP(pp_splice)
79072805 5502{
20b7effb 5503 dSP; dMARK; dORIGMARK;
5cd408a2 5504 int num_args = (SP - MARK);
00576728 5505 AV *ary = MUTABLE_AV(*++MARK);
eb578fdb
KW
5506 SV **src;
5507 SV **dst;
c70927a6
FC
5508 SSize_t i;
5509 SSize_t offset;
5510 SSize_t length;
5511 SSize_t newlen;
5512 SSize_t after;
5513 SSize_t diff;
ad64d0ec 5514 const MAGIC * const mg = SvTIED_mg((const SV *)ary, PERL_MAGIC_tied);
93965878 5515
1b6737cc 5516 if (mg) {
3e0cb5de 5517 return Perl_tied_method(aTHX_ SV_CONST(SPLICE), mark - 1, MUTABLE_SV(ary), mg,
af71faff
NC
5518 GIMME_V | TIED_METHOD_ARGUMENTS_ON_STACK,
5519 sp - mark);
93965878 5520 }
79072805 5521
3275d25a
AC
5522 if (SvREADONLY(ary))
5523 Perl_croak_no_modify();
5524
a0d0e21e 5525 SP++;
79072805 5526
a0d0e21e 5527 if (++MARK < SP) {
4ea561bc 5528 offset = i = SvIV(*MARK);
a0d0e21e 5529 if (offset < 0)
93965878 5530 offset += AvFILLp(ary) + 1;
84902520 5531 if (offset < 0)
cea2e8a9 5532 DIE(aTHX_ PL_no_aelem, i);
a0d0e21e
LW
5533 if (++MARK < SP) {
5534 length = SvIVx(*MARK++);
48cdf507
GA
5535 if (length < 0) {
5536 length += AvFILLp(ary) - offset + 1;
5537 if (length < 0)
5538 length = 0;
5539 }
79072805
LW
5540 }
5541 else
a0d0e21e 5542 length = AvMAX(ary) + 1; /* close enough to infinity */
79072805 5543 }
a0d0e21e
LW
5544 else {
5545 offset = 0;
5546 length = AvMAX(ary) + 1;
5547 }
8cbc2e3b 5548 if (offset > AvFILLp(ary) + 1) {
5cd408a2
EB
5549 if (num_args > 2)
5550 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "splice() offset past end of array" );
93965878 5551 offset = AvFILLp(ary) + 1;
8cbc2e3b 5552 }
93965878 5553 after = AvFILLp(ary) + 1 - (offset + length);
a0d0e21e
LW
5554 if (after < 0) { /* not that much array */
5555 length += after; /* offset+length now in array */
5556 after = 0;
5557 if (!AvALLOC(ary))
5558 av_extend(ary, 0);
5559 }
5560
5561 /* At this point, MARK .. SP-1 is our new LIST */
5562
5563 newlen = SP - MARK;
5564 diff = newlen - length;
13d7cbc1
GS
5565 if (newlen && !AvREAL(ary) && AvREIFY(ary))
5566 av_reify(ary);
a0d0e21e 5567
50528de0
WL
5568 /* make new elements SVs now: avoid problems if they're from the array */
5569 for (dst = MARK, i = newlen; i; i--) {
1b6737cc 5570 SV * const h = *dst;
f2b990bf 5571 *dst++ = newSVsv(h);
50528de0
WL
5572 }
5573
a0d0e21e 5574 if (diff < 0) { /* shrinking the area */
95b63a38 5575 SV **tmparyval = NULL;
a0d0e21e 5576 if (newlen) {
a02a5408 5577 Newx(tmparyval, newlen, SV*); /* so remember insertion */
a0d0e21e 5578 Copy(MARK, tmparyval, newlen, SV*);
79072805 5579 }
a0d0e21e
LW
5580
5581 MARK = ORIGMARK + 1;
82334630 5582 if (GIMME_V == G_ARRAY) { /* copy return vals to stack */
31c61add 5583 const bool real = cBOOL(AvREAL(ary));
a0d0e21e 5584 MEXTEND(MARK, length);
31c61add 5585 if (real)
bbce6d69 5586 EXTEND_MORTAL(length);
31c61add
FC
5587 for (i = 0, dst = MARK; i < length; i++) {
5588 if ((*dst = AvARRAY(ary)[i+offset])) {
5589 if (real)
486ec47a 5590 sv_2mortal(*dst); /* free them eventually */
36477c24 5591 }
31c61add
FC
5592 else
5593 *dst = &PL_sv_undef;
5594 dst++;
a0d0e21e
LW
5595 }
5596 MARK += length - 1;
79072805 5597 }
a0d0e21e
LW
5598 else {
5599 *MARK = AvARRAY(ary)[offset+length-1];
5600 if (AvREAL(ary)) {
d689ffdd 5601 sv_2mortal(*MARK);
a0d0e21e
LW
5602 for (i = length - 1, dst = &AvARRAY(ary)[offset]; i > 0; i--)
5603 SvREFCNT_dec(*dst++); /* free them now */
79072805 5604 }
92b69f65
FC
5605 if (!*MARK)
5606 *MARK = &PL_sv_undef;
a0d0e21e 5607 }
93965878 5608 AvFILLp(ary) += diff;
a0d0e21e
LW
5609
5610 /* pull up or down? */
5611
5612 if (offset < after) { /* easier to pull up */
5613 if (offset) { /* esp. if nothing to pull */
5614 src = &AvARRAY(ary)[offset-1];
5615 dst = src - diff; /* diff is negative */
5616 for (i = offset; i > 0; i--) /* can't trust Copy */
5617 *dst-- = *src--;
79072805 5618 }
a0d0e21e 5619 dst = AvARRAY(ary);
9c6bc640 5620 AvARRAY(ary) = AvARRAY(ary) - diff; /* diff is negative */
a0d0e21e
LW
5621 AvMAX(ary) += diff;
5622 }
5623 else {
5624 if (after) { /* anything to pull down? */
5625 src = AvARRAY(ary) + offset + length;
5626 dst = src + diff; /* diff is negative */
5627 Move(src, dst, after, SV*);
79072805 5628 }
93965878 5629 dst = &AvARRAY(ary)[AvFILLp(ary)+1];
a0d0e21e
LW
5630 /* avoid later double free */
5631 }
5632 i = -diff;
5633 while (i)
ce0d59fd 5634 dst[--i] = NULL;
a8e41ef4 5635
a0d0e21e 5636 if (newlen) {
50528de0 5637 Copy( tmparyval, AvARRAY(ary) + offset, newlen, SV* );
a0d0e21e
LW
5638 Safefree(tmparyval);
5639 }
5640 }
5641 else { /* no, expanding (or same) */
d3961450 5642 SV** tmparyval = NULL;
a0d0e21e 5643 if (length) {
a02a5408 5644 Newx(tmparyval, length, SV*); /* so remember deletion */
a0d0e21e
LW
5645 Copy(AvARRAY(ary)+offset, tmparyval, length, SV*);
5646 }
5647
5648 if (diff > 0) { /* expanding */
a0d0e21e 5649 /* push up or down? */
a0d0e21e
LW
5650 if (offset < after && diff <= AvARRAY(ary) - AvALLOC(ary)) {
5651 if (offset) {
5652 src = AvARRAY(ary);
5653 dst = src - diff;
5654 Move(src, dst, offset, SV*);
79072805 5655 }
9c6bc640 5656 AvARRAY(ary) = AvARRAY(ary) - diff;/* diff is positive */
a0d0e21e 5657 AvMAX(ary) += diff;
93965878 5658 AvFILLp(ary) += diff;
79072805
LW
5659 }
5660 else {
93965878
NIS
5661 if (AvFILLp(ary) + diff >= AvMAX(ary)) /* oh, well */
5662 av_extend(ary, AvFILLp(ary) + diff);
5663 AvFILLp(ary) += diff;
a0d0e21e
LW
5664
5665 if (after) {
93965878 5666 dst = AvARRAY(ary) + AvFILLp(ary);
a0d0e21e
LW
5667 src = dst - diff;
5668 for (i = after; i; i--) {
5669 *dst-- = *src--;
5670 }
79072805
LW
5671 }
5672 }
a0d0e21e
LW
5673 }
5674
50528de0
WL
5675 if (newlen) {
5676 Copy( MARK, AvARRAY(ary) + offset, newlen, SV* );
a0d0e21e 5677 }
50528de0 5678
a0d0e21e 5679 MARK = ORIGMARK + 1;
82334630 5680 if (GIMME_V == G_ARRAY) { /* copy return vals to stack */
a0d0e21e 5681 if (length) {
31c61add
FC
5682 const bool real = cBOOL(AvREAL(ary));
5683 if (real)
bbce6d69 5684 EXTEND_MORTAL(length);
31c61add
FC
5685 for (i = 0, dst = MARK; i < length; i++) {
5686 if ((*dst = tmparyval[i])) {
5687 if (real)
486ec47a 5688 sv_2mortal(*dst); /* free them eventually */
36477c24 5689 }
31c61add
FC
5690 else *dst = &PL_sv_undef;
5691 dst++;
79072805
LW
5692 }
5693 }
a0d0e21e
LW
5694 MARK += length - 1;
5695 }
5696 else if (length--) {
5697 *MARK = tmparyval[length];
5698 if (AvREAL(ary)) {
d689ffdd 5699 sv_2mortal(*MARK);
a0d0e21e
LW
5700 while (length-- > 0)
5701 SvREFCNT_dec(tmparyval[length]);
79072805 5702 }
92b69f65
FC
5703 if (!*MARK)
5704 *MARK = &PL_sv_undef;
79072805 5705 }
a0d0e21e 5706 else
3280af22 5707 *MARK = &PL_sv_undef;
d3961450 5708 Safefree(tmparyval);
79072805 5709 }
474af990
FR
5710
5711 if (SvMAGICAL(ary))
5712 mg_set(MUTABLE_SV(ary));
5713
a0d0e21e 5714 SP = MARK;
79072805
LW
5715 RETURN;
5716}
5717
a0d0e21e 5718PP(pp_push)
79072805 5719{
20b7effb 5720 dSP; dMARK; dORIGMARK; dTARGET;
00576728 5721 AV * const ary = MUTABLE_AV(*++MARK);
ad64d0ec 5722 const MAGIC * const mg = SvTIED_mg((const SV *)ary, PERL_MAGIC_tied);
79072805 5723
1b6737cc 5724 if (mg) {
ad64d0ec 5725 *MARK-- = SvTIED_obj(MUTABLE_SV(ary), mg);
93965878
NIS
5726 PUSHMARK(MARK);
5727 PUTBACK;
d343c3ef 5728 ENTER_with_name("call_PUSH");
3e0cb5de 5729 call_sv(SV_CONST(PUSH),G_SCALAR|G_DISCARD|G_METHOD_NAMED);
d343c3ef 5730 LEAVE_with_name("call_PUSH");
01072573 5731 /* SPAGAIN; not needed: SP is assigned to immediately below */
93965878 5732 }
a60c0954 5733 else {
a68090fe
DM
5734 /* PL_delaymagic is restored by JUMPENV_POP on dieing, so we
5735 * only need to save locally, not on the save stack */
5736 U16 old_delaymagic = PL_delaymagic;
5737
cb077ed2 5738 if (SvREADONLY(ary) && MARK < SP) Perl_croak_no_modify();
89c14e2e 5739 PL_delaymagic = DM_DELAY;
a60c0954 5740 for (++MARK; MARK <= SP; MARK++) {
3ed356df
FC
5741 SV *sv;
5742 if (*MARK) SvGETMAGIC(*MARK);
5743 sv = newSV(0);
a60c0954 5744 if (*MARK)
3ed356df 5745 sv_setsv_nomg(sv, *MARK);
0a75904b 5746 av_store(ary, AvFILLp(ary)+1, sv);
a60c0954 5747 }
354b0578 5748 if (PL_delaymagic & DM_ARRAY_ISA)
ad64d0ec 5749 mg_set(MUTABLE_SV(ary));
a68090fe 5750 PL_delaymagic = old_delaymagic;
6eeabd23
VP
5751 }
5752 SP = ORIGMARK;
5753 if (OP_GIMME(PL_op, 0) != G_VOID) {
5754 PUSHi( AvFILL(ary) + 1 );
79072805 5755 }
79072805
LW
5756 RETURN;
5757}
5758
b1c05ba5 5759/* also used for: pp_pop()*/
a0d0e21e 5760PP(pp_shift)
79072805 5761{
39644a26 5762 dSP;
538f5756 5763 AV * const av = PL_op->op_flags & OPf_SPECIAL
94f9945d 5764 ? MUTABLE_AV(GvAVn(PL_defgv)) : MUTABLE_AV(POPs);
789b4bc9 5765 SV * const sv = PL_op->op_type == OP_SHIFT ? av_shift(av) : av_pop(av);
79072805 5766 EXTEND(SP, 1);
c2b4a044 5767 assert (sv);
d689ffdd 5768 if (AvREAL(av))
a0d0e21e
LW
5769 (void)sv_2mortal(sv);
5770 PUSHs(sv);
79072805 5771 RETURN;
79072805
LW
5772}
5773
a0d0e21e 5774PP(pp_unshift)
79072805 5775{
20b7effb 5776 dSP; dMARK; dORIGMARK; dTARGET;
00576728 5777 AV *ary = MUTABLE_AV(*++MARK);
ad64d0ec 5778 const MAGIC * const mg = SvTIED_mg((const SV *)ary, PERL_MAGIC_tied);
93965878 5779
1b6737cc 5780 if (mg) {
ad64d0ec 5781 *MARK-- = SvTIED_obj(MUTABLE_SV(ary), mg);
7fd66d9d 5782 PUSHMARK(MARK);
93965878 5783 PUTBACK;
d343c3ef 5784 ENTER_with_name("call_UNSHIFT");
36925d9e 5785 call_sv(SV_CONST(UNSHIFT),G_SCALAR|G_DISCARD|G_METHOD_NAMED);
d343c3ef 5786 LEAVE_with_name("call_UNSHIFT");
01072573 5787 /* SPAGAIN; not needed: SP is assigned to immediately below */
93965878 5788 }
a60c0954 5789 else {
a68090fe
DM
5790 /* PL_delaymagic is restored by JUMPENV_POP on dieing, so we
5791 * only need to save locally, not on the save stack */
5792 U16 old_delaymagic = PL_delaymagic;
c70927a6 5793 SSize_t i = 0;
a68090fe 5794
a60c0954 5795 av_unshift(ary, SP - MARK);
39539141 5796 PL_delaymagic = DM_DELAY;
a60c0954 5797 while (MARK < SP) {
1b6737cc 5798 SV * const sv = newSVsv(*++MARK);
a60c0954
NIS
5799 (void)av_store(ary, i++, sv);
5800 }
39539141
DIM
5801 if (PL_delaymagic & DM_ARRAY_ISA)
5802 mg_set(MUTABLE_SV(ary));
a68090fe 5803 PL_delaymagic = old_delaymagic;
79072805 5804 }
a0d0e21e 5805 SP = ORIGMARK;
6eeabd23 5806 if (OP_GIMME(PL_op, 0) != G_VOID) {
5658d0a9
LR
5807 PUSHi( AvFILL(ary) + 1 );
5808 }
79072805 5809 RETURN;
79072805
LW
5810}
5811
a0d0e21e 5812PP(pp_reverse)
79072805 5813{
20b7effb 5814 dSP; dMARK;
79072805 5815
82334630 5816 if (GIMME_V == G_ARRAY) {
484c818f
VP
5817 if (PL_op->op_private & OPpREVERSE_INPLACE) {
5818 AV *av;
5819
5820 /* See pp_sort() */
5821 assert( MARK+1 == SP && *SP && SvTYPE(*SP) == SVt_PVAV);
5822 (void)POPMARK; /* remove mark associated with ex-OP_AASSIGN */
5823 av = MUTABLE_AV((*SP));
5824 /* In-place reversing only happens in void context for the array
5825 * assignment. We don't need to push anything on the stack. */
5826 SP = MARK;
5827
5828 if (SvMAGICAL(av)) {
c70927a6 5829 SSize_t i, j;
eb578fdb 5830 SV *tmp = sv_newmortal();
484c818f
VP
5831 /* For SvCANEXISTDELETE */
5832 HV *stash;
5833 const MAGIC *mg;
5834 bool can_preserve = SvCANEXISTDELETE(av);
5835
b9f2b683 5836 for (i = 0, j = av_tindex(av); i < j; ++i, --j) {
eb578fdb 5837 SV *begin, *end;
484c818f
VP
5838
5839 if (can_preserve) {
5840 if (!av_exists(av, i)) {
5841 if (av_exists(av, j)) {
eb578fdb 5842 SV *sv = av_delete(av, j, 0);
484c818f
VP
5843 begin = *av_fetch(av, i, TRUE);
5844 sv_setsv_mg(begin, sv);
5845 }
5846 continue;
5847 }
5848 else if (!av_exists(av, j)) {
eb578fdb 5849 SV *sv = av_delete(av, i, 0);
484c818f
VP
5850 end = *av_fetch(av, j, TRUE);
5851 sv_setsv_mg(end, sv);
5852 continue;
5853 }
5854 }
5855
5856 begin = *av_fetch(av, i, TRUE);
5857 end = *av_fetch(av, j, TRUE);
5858 sv_setsv(tmp, begin);
5859 sv_setsv_mg(begin, end);
5860 sv_setsv_mg(end, tmp);
5861 }
5862 }
5863 else {
5864 SV **begin = AvARRAY(av);
484c818f 5865
95a26d8e
VP
5866 if (begin) {
5867 SV **end = begin + AvFILLp(av);
5868
5869 while (begin < end) {
eb578fdb 5870 SV * const tmp = *begin;
95a26d8e
VP
5871 *begin++ = *end;
5872 *end-- = tmp;
5873 }
484c818f
VP
5874 }
5875 }
5876 }
5877 else {
5878 SV **oldsp = SP;
5879 MARK++;
5880 while (MARK < SP) {
eb578fdb 5881 SV * const tmp = *MARK;
484c818f
VP
5882 *MARK++ = *SP;
5883 *SP-- = tmp;
5884 }
5885 /* safe as long as stack cannot get extended in the above */
5886 SP = oldsp;
a0d0e21e 5887 }
79072805
LW
5888 }
5889 else {
eb578fdb 5890 char *up;
a0d0e21e
LW
5891 dTARGET;
5892 STRLEN len;
79072805 5893
7e2040f0 5894 SvUTF8_off(TARG); /* decontaminate */
47836a13 5895 if (SP - MARK > 1) {
3280af22 5896 do_join(TARG, &PL_sv_no, MARK, SP);
47836a13
Z
5897 SP = MARK + 1;
5898 SETs(TARG);
5899 } else if (SP > MARK) {
d5d91c1e 5900 sv_setsv(TARG, *SP);
47836a13
Z
5901 SETs(TARG);
5902 } else {
d5d91c1e 5903 sv_setsv(TARG, DEFSV);
47836a13 5904 XPUSHs(TARG);
1e21d011 5905 }
69d4d9c8 5906 SvSETMAGIC(TARG); /* remove any utf8 length magic */
1e21d011 5907
a0d0e21e
LW
5908 up = SvPV_force(TARG, len);
5909 if (len > 1) {
19742f39 5910 char *down;
7e2040f0 5911 if (DO_UTF8(TARG)) { /* first reverse each character */
dfe13c55 5912 U8* s = (U8*)SvPVX(TARG);
349d4f2f 5913 const U8* send = (U8*)(s + len);
a0ed51b3 5914 while (s < send) {
d742c382 5915 if (UTF8_IS_INVARIANT(*s)) {
a0ed51b3
LW
5916 s++;
5917 continue;
5918 }
5919 else {
4b88fb76 5920 if (!utf8_to_uvchr_buf(s, send, 0))
a0dbb045 5921 break;
dfe13c55 5922 up = (char*)s;
a0ed51b3 5923 s += UTF8SKIP(s);
dfe13c55 5924 down = (char*)(s - 1);
a0dbb045 5925 /* reverse this character */
a0ed51b3 5926 while (down > up) {
19742f39 5927 const char tmp = *up;
a0ed51b3 5928 *up++ = *down;
19742f39 5929 *down-- = tmp;
a0ed51b3
LW
5930 }
5931 }
5932 }
5933 up = SvPVX(TARG);
5934 }
a0d0e21e
LW
5935 down = SvPVX(TARG) + len - 1;
5936 while (down > up) {
19742f39 5937 const char tmp = *up;
a0d0e21e 5938 *up++ = *down;
19742f39 5939 *down-- = tmp;
a0d0e21e 5940 }
3aa33fe5 5941 (void)SvPOK_only_UTF8(TARG);
79072805 5942 }
79072805 5943 }
a0d0e21e 5944 RETURN;
79072805
LW
5945}
5946
a0d0e21e 5947PP(pp_split)
79072805 5948{
20b7effb 5949 dSP; dTARG;
692044df
DM
5950 AV *ary = ( (PL_op->op_private & OPpSPLIT_ASSIGN) /* @a = split */
5951 && (PL_op->op_flags & OPf_STACKED)) /* @{expr} = split */
5012eebe 5952 ? (AV *)POPs : NULL;
eb578fdb 5953 IV limit = POPi; /* note, negative is forever */
1b6737cc 5954 SV * const sv = POPs;
a0d0e21e 5955 STRLEN len;
eb578fdb 5956 const char *s = SvPV_const(sv, len);
1b6737cc 5957 const bool do_utf8 = DO_UTF8(sv);
20ae58f7 5958 const bool in_uni_8_bit = IN_UNI_8_BIT;
727b7506 5959 const char *strend = s + len;
5012eebe 5960 PMOP *pm = cPMOPx(PL_op);
eb578fdb
KW
5961 REGEXP *rx;
5962 SV *dstr;
5963 const char *m;
c70927a6 5964 SSize_t iters = 0;
d14578b8
KW
5965 const STRLEN slen = do_utf8
5966 ? utf8_length((U8*)s, (U8*)strend)
5967 : (STRLEN)(strend - s);
c70927a6 5968 SSize_t maxiters = slen + 10;
c1a7495a 5969 I32 trailing_empty = 0;
727b7506 5970 const char *orig;
052a7c76 5971 const IV origlimit = limit;
a0d0e21e
LW
5972 I32 realarray = 0;
5973 I32 base;
1c23e2bd 5974 const U8 gimme = GIMME_V;
941446f6 5975 bool gimme_scalar;
692044df 5976 I32 oldsave = PL_savestack_ix;
437d3b4e 5977 U32 make_mortal = SVs_TEMP;
7fba1cd6 5978 bool multiline = 0;
b37c2d43 5979 MAGIC *mg = NULL;
79072805 5980
aaa362c4 5981 rx = PM_GETRE(pm);
bbce6d69 5982
a62b1201 5983 TAINT_IF(get_regex_charset(RX_EXTFLAGS(rx)) == REGEX_LOCALE_CHARSET &&
dbc200c5 5984 (RX_EXTFLAGS(rx) & (RXf_WHITE | RXf_SKIPWHITE)));
bbce6d69 5985
692044df 5986 /* handle @ary = split(...) optimisation */
5012eebe
DM
5987 if (PL_op->op_private & OPpSPLIT_ASSIGN) {
5988 if (!(PL_op->op_flags & OPf_STACKED)) {
692044df
DM
5989 if (PL_op->op_private & OPpSPLIT_LEX) {
5990 if (PL_op->op_private & OPpLVAL_INTRO)
5991 SAVECLEARSV(PAD_SVl(pm->op_pmreplrootu.op_pmtargetoff));
5012eebe 5992 ary = (AV *)PAD_SVl(pm->op_pmreplrootu.op_pmtargetoff);
692044df 5993 }
5012eebe
DM
5994 else {
5995 GV *gv =
971a9dd3 5996#ifdef USE_ITHREADS
5012eebe 5997 MUTABLE_GV(PAD_SVl(pm->op_pmreplrootu.op_pmtargetoff));
971a9dd3 5998#else
5012eebe 5999 pm->op_pmreplrootu.op_pmtargetgv;
20e98b0f 6000#endif
692044df
DM
6001 if (PL_op->op_private & OPpLVAL_INTRO)
6002 ary = save_ary(gv);
6003 else
6004 ary = GvAVn(gv);
5012eebe 6005 }
692044df
DM
6006 /* skip anything pushed by OPpLVAL_INTRO above */
6007 oldsave = PL_savestack_ix;
5012eebe
DM
6008 }
6009
a0d0e21e 6010 realarray = 1;
8ec5e241 6011 PUTBACK;
a0d0e21e 6012 av_extend(ary,0);
821956c5 6013 (void)sv_2mortal(SvREFCNT_inc_simple_NN(sv));
a0d0e21e 6014 av_clear(ary);
8ec5e241 6015 SPAGAIN;
ad64d0ec 6016 if ((mg = SvTIED_mg((const SV *)ary, PERL_MAGIC_tied))) {
8ec5e241 6017 PUSHMARK(SP);
ad64d0ec 6018 XPUSHs(SvTIED_obj(MUTABLE_SV(ary), mg));
8ec5e241
NIS
6019 }
6020 else {
1c0b011c 6021 if (!AvREAL(ary)) {
1b6737cc 6022 I32 i;
1c0b011c 6023 AvREAL_on(ary);
abff13bb 6024 AvREIFY_off(ary);
1c0b011c 6025 for (i = AvFILLp(ary); i >= 0; i--)
d14578b8 6026 AvARRAY(ary)[i] = &PL_sv_undef; /* don't free mere refs */
1c0b011c
NIS
6027 }
6028 /* temporarily switch stacks */
8b7059b1 6029 SAVESWITCHSTACK(PL_curstack, ary);
8ec5e241 6030 make_mortal = 0;
1c0b011c 6031 }
79072805 6032 }
5012eebe 6033
3280af22 6034 base = SP - PL_stack_base;
a0d0e21e 6035 orig = s;
dbc200c5 6036 if (RX_EXTFLAGS(rx) & RXf_SKIPWHITE) {
613f191e 6037 if (do_utf8) {
d720149d 6038 while (s < strend && isSPACE_utf8_safe(s, strend))
613f191e
TS
6039 s += UTF8SKIP(s);
6040 }
a62b1201 6041 else if (get_regex_charset(RX_EXTFLAGS(rx)) == REGEX_LOCALE_CHARSET) {
d720149d 6042 while (s < strend && isSPACE_LC(*s))
bbce6d69 6043 s++;
6044 }
20ae58f7
AC
6045 else if (in_uni_8_bit) {
6046 while (s < strend && isSPACE_L1(*s))
6047 s++;
6048 }
bbce6d69 6049 else {
d720149d 6050 while (s < strend && isSPACE(*s))
bbce6d69 6051 s++;
6052 }
a0d0e21e 6053 }
73134a2e 6054 if (RX_EXTFLAGS(rx) & RXf_PMf_MULTILINE) {
7fba1cd6 6055 multiline = 1;
c07a80fd 6056 }
6057
941446f6
FC
6058 gimme_scalar = gimme == G_SCALAR && !ary;
6059
a0d0e21e
LW
6060 if (!limit)
6061 limit = maxiters + 2;
dbc200c5 6062 if (RX_EXTFLAGS(rx) & RXf_WHITE) {
a0d0e21e 6063 while (--limit) {
bbce6d69 6064 m = s;
8727f688
YO
6065 /* this one uses 'm' and is a negative test */
6066 if (do_utf8) {
7a207065 6067 while (m < strend && ! isSPACE_utf8_safe(m, strend) ) {
613f191e 6068 const int t = UTF8SKIP(m);
7a207065 6069 /* isSPACE_utf8_safe returns FALSE for malform utf8 */
613f191e
TS
6070 if (strend - m < t)
6071 m = strend;
6072 else
6073 m += t;
6074 }
a62b1201 6075 }
d14578b8
KW
6076 else if (get_regex_charset(RX_EXTFLAGS(rx)) == REGEX_LOCALE_CHARSET)
6077 {
8727f688
YO
6078 while (m < strend && !isSPACE_LC(*m))
6079 ++m;
20ae58f7
AC
6080 }
6081 else if (in_uni_8_bit) {
6082 while (m < strend && !isSPACE_L1(*m))
6083 ++m;
8727f688
YO
6084 } else {
6085 while (m < strend && !isSPACE(*m))
6086 ++m;
a8e41ef4 6087 }
a0d0e21e
LW
6088 if (m >= strend)
6089 break;
bbce6d69 6090
c1a7495a
BB
6091 if (gimme_scalar) {
6092 iters++;
6093 if (m-s == 0)
6094 trailing_empty++;
6095 else
6096 trailing_empty = 0;
6097 } else {
6098 dstr = newSVpvn_flags(s, m-s,
6099 (do_utf8 ? SVf_UTF8 : 0) | make_mortal);
6100 XPUSHs(dstr);
6101 }
bbce6d69 6102
613f191e
TS
6103 /* skip the whitespace found last */
6104 if (do_utf8)
6105 s = m + UTF8SKIP(m);
6106 else
6107 s = m + 1;
6108
8727f688
YO
6109 /* this one uses 's' and is a positive test */
6110 if (do_utf8) {
7a207065 6111 while (s < strend && isSPACE_utf8_safe(s, strend) )
8727f688 6112 s += UTF8SKIP(s);
a62b1201 6113 }
d14578b8
KW
6114 else if (get_regex_charset(RX_EXTFLAGS(rx)) == REGEX_LOCALE_CHARSET)
6115 {
8727f688
YO
6116 while (s < strend && isSPACE_LC(*s))
6117 ++s;
20ae58f7
AC
6118 }
6119 else if (in_uni_8_bit) {
6120 while (s < strend && isSPACE_L1(*s))
6121 ++s;
8727f688
YO
6122 } else {
6123 while (s < strend && isSPACE(*s))
6124 ++s;
a8e41ef4 6125 }
79072805
LW
6126 }
6127 }
07bc277f 6128 else if (RX_EXTFLAGS(rx) & RXf_START_ONLY) {
a0d0e21e 6129 while (--limit) {
a6e20a40
AL
6130 for (m = s; m < strend && *m != '\n'; m++)
6131 ;
a0d0e21e
LW
6132 m++;
6133 if (m >= strend)
6134 break;
c1a7495a
BB
6135
6136 if (gimme_scalar) {
6137 iters++;
6138 if (m-s == 0)
6139 trailing_empty++;
6140 else
6141 trailing_empty = 0;
6142 } else {
6143 dstr = newSVpvn_flags(s, m-s,
6144 (do_utf8 ? SVf_UTF8 : 0) | make_mortal);
6145 XPUSHs(dstr);
6146 }
a0d0e21e
LW
6147 s = m;
6148 }
6149 }
07bc277f 6150 else if (RX_EXTFLAGS(rx) & RXf_NULL && !(s >= strend)) {
640f820d
AB
6151 /*
6152 Pre-extend the stack, either the number of bytes or
6153 characters in the string or a limited amount, triggered by:
6154
6155 my ($x, $y) = split //, $str;
6156 or
6157 split //, $str, $i;
6158 */
c1a7495a 6159 if (!gimme_scalar) {
052a7c76
DM
6160 const IV items = limit - 1;
6161 /* setting it to -1 will trigger a panic in EXTEND() */
6162 const SSize_t sslen = slen > SSize_t_MAX ? -1 : (SSize_t)slen;
6163 if (items >=0 && items < sslen)
c1a7495a
BB
6164 EXTEND(SP, items);
6165 else
052a7c76 6166 EXTEND(SP, sslen);
c1a7495a 6167 }
640f820d 6168
e9515b0f
AB
6169 if (do_utf8) {
6170 while (--limit) {
6171 /* keep track of how many bytes we skip over */
6172 m = s;
640f820d 6173 s += UTF8SKIP(s);
c1a7495a
BB
6174 if (gimme_scalar) {
6175 iters++;
6176 if (s-m == 0)
6177 trailing_empty++;
6178 else
6179 trailing_empty = 0;
6180 } else {
6181 dstr = newSVpvn_flags(m, s-m, SVf_UTF8 | make_mortal);
640f820d 6182
c1a7495a
BB
6183 PUSHs(dstr);
6184 }
640f820d 6185
e9515b0f
AB
6186 if (s >= strend)
6187 break;
6188 }
6189 } else {
6190 while (--limit) {
c1a7495a
BB
6191 if (gimme_scalar) {
6192 iters++;
6193 } else {
6194 dstr = newSVpvn(s, 1);
e9515b0f 6195
e9515b0f 6196
c1a7495a
BB
6197 if (make_mortal)
6198 sv_2mortal(dstr);
640f820d 6199
c1a7495a
BB
6200 PUSHs(dstr);
6201 }
6202
6203 s++;
e9515b0f
AB
6204
6205 if (s >= strend)
6206 break;
6207 }
640f820d
AB
6208 }
6209 }
3c8556c3 6210 else if (do_utf8 == (RX_UTF8(rx) != 0) &&
07bc277f
NC
6211 (RX_EXTFLAGS(rx) & RXf_USE_INTUIT) && !RX_NPARENS(rx)
6212 && (RX_EXTFLAGS(rx) & RXf_CHECK_ALL)
8e1490ee 6213 && !(RX_EXTFLAGS(rx) & RXf_IS_ANCHORED)) {
07bc277f 6214 const int tail = (RX_EXTFLAGS(rx) & RXf_INTUIT_TAIL);
f9f4320a 6215 SV * const csv = CALLREG_INTUIT_STRING(rx);
cf93c79d 6216
07bc277f 6217 len = RX_MINLENRET(rx);
3c8556c3 6218 if (len == 1 && !RX_UTF8(rx) && !tail) {
1b6737cc 6219 const char c = *SvPV_nolen_const(csv);
a0d0e21e 6220 while (--limit) {
a6e20a40
AL
6221 for (m = s; m < strend && *m != c; m++)
6222 ;
a0d0e21e
LW
6223 if (m >= strend)
6224 break;
c1a7495a
BB
6225 if (gimme_scalar) {
6226 iters++;
6227 if (m-s == 0)
6228 trailing_empty++;
6229 else
6230 trailing_empty = 0;
6231 } else {
6232 dstr = newSVpvn_flags(s, m-s,
d14578b8 6233 (do_utf8 ? SVf_UTF8 : 0) | make_mortal);
c1a7495a
BB
6234 XPUSHs(dstr);
6235 }
93f04dac
JH
6236 /* The rx->minlen is in characters but we want to step
6237 * s ahead by bytes. */
1aa99e6b 6238 if (do_utf8)
cf70d9e6 6239 s = (char*)utf8_hop_forward((U8*) m, len, (U8*) strend);
1aa99e6b
IH
6240 else
6241 s = m + len; /* Fake \n at the end */
a0d0e21e
LW
6242 }
6243 }
6244 else {
a0d0e21e 6245 while (s < strend && --limit &&
f722798b 6246 (m = fbm_instr((unsigned char*)s, (unsigned char*)strend,
7fba1cd6 6247 csv, multiline ? FBMrf_MULTILINE : 0)) )
a0d0e21e 6248 {
c1a7495a
BB
6249 if (gimme_scalar) {
6250 iters++;
6251 if (m-s == 0)
6252 trailing_empty++;
6253 else
6254 trailing_empty = 0;
6255 } else {
6256 dstr = newSVpvn_flags(s, m-s,
d14578b8 6257 (do_utf8 ? SVf_UTF8 : 0) | make_mortal);
c1a7495a
BB
6258 XPUSHs(dstr);
6259 }
93f04dac
JH
6260 /* The rx->minlen is in characters but we want to step
6261 * s ahead by bytes. */
1aa99e6b 6262 if (do_utf8)
cf70d9e6 6263 s = (char*)utf8_hop_forward((U8*)m, len, (U8 *) strend);
1aa99e6b
IH
6264 else
6265 s = m + len; /* Fake \n at the end */
a0d0e21e 6266 }
463ee0b2 6267 }
463ee0b2 6268 }
a0d0e21e 6269 else {
07bc277f 6270 maxiters += slen * RX_NPARENS(rx);
080c2dec 6271 while (s < strend && --limit)
bbce6d69 6272 {
1b6737cc 6273 I32 rex_return;
080c2dec 6274 PUTBACK;
d14578b8 6275 rex_return = CALLREGEXEC(rx, (char*)s, (char*)strend, (char*)orig, 1,
c33e64f0 6276 sv, NULL, 0);
080c2dec 6277 SPAGAIN;
1b6737cc 6278 if (rex_return == 0)
080c2dec 6279 break;
d9f97599 6280 TAINT_IF(RX_MATCH_TAINTED(rx));
6502e081
DM
6281 /* we never pass the REXEC_COPY_STR flag, so it should
6282 * never get copied */
6283 assert(!RX_MATCH_COPIED(rx));
07bc277f 6284 m = RX_OFFS(rx)[0].start + orig;
c1a7495a
BB
6285
6286 if (gimme_scalar) {
6287 iters++;
6288 if (m-s == 0)
6289 trailing_empty++;
6290 else
6291 trailing_empty = 0;
6292 } else {
6293 dstr = newSVpvn_flags(s, m-s,
6294 (do_utf8 ? SVf_UTF8 : 0) | make_mortal);
6295 XPUSHs(dstr);
6296 }
07bc277f 6297 if (RX_NPARENS(rx)) {
1b6737cc 6298 I32 i;
07bc277f
NC
6299 for (i = 1; i <= (I32)RX_NPARENS(rx); i++) {
6300 s = RX_OFFS(rx)[i].start + orig;
6301 m = RX_OFFS(rx)[i].end + orig;
6de67870
JP
6302
6303 /* japhy (07/27/01) -- the (m && s) test doesn't catch
6304 parens that didn't match -- they should be set to
6305 undef, not the empty string */
c1a7495a
BB
6306 if (gimme_scalar) {
6307 iters++;
6308 if (m-s == 0)
6309 trailing_empty++;
6310 else
6311 trailing_empty = 0;
6312 } else {
6313 if (m >= orig && s >= orig) {
6314 dstr = newSVpvn_flags(s, m-s,
6315 (do_utf8 ? SVf_UTF8 : 0)
6316 | make_mortal);
6317 }
6318 else
6319 dstr = &PL_sv_undef; /* undef, not "" */
6320 XPUSHs(dstr);
748a9306 6321 }
c1a7495a 6322
a0d0e21e
LW
6323 }
6324 }
07bc277f 6325 s = RX_OFFS(rx)[0].end + orig;
a0d0e21e 6326 }
79072805 6327 }
8ec5e241 6328
c1a7495a
BB
6329 if (!gimme_scalar) {
6330 iters = (SP - PL_stack_base) - base;
6331 }
a0d0e21e 6332 if (iters > maxiters)
cea2e8a9 6333 DIE(aTHX_ "Split loop");
8ec5e241 6334
a0d0e21e
LW
6335 /* keep field after final delim? */
6336 if (s < strend || (iters && origlimit)) {
c1a7495a
BB
6337 if (!gimme_scalar) {
6338 const STRLEN l = strend - s;
6339 dstr = newSVpvn_flags(s, l, (do_utf8 ? SVf_UTF8 : 0) | make_mortal);
6340 XPUSHs(dstr);
6341 }
a0d0e21e 6342 iters++;
79072805 6343 }
a0d0e21e 6344 else if (!origlimit) {
c1a7495a
BB
6345 if (gimme_scalar) {
6346 iters -= trailing_empty;
6347 } else {
6348 while (iters > 0 && (!TOPs || !SvANY(TOPs) || SvCUR(TOPs) == 0)) {
6349 if (TOPs && !make_mortal)
6350 sv_2mortal(TOPs);
71ca73e5 6351 *SP-- = NULL;
c1a7495a
BB
6352 iters--;
6353 }
89900bd3 6354 }
a0d0e21e 6355 }
8ec5e241 6356
8b7059b1
DM
6357 PUTBACK;
6358 LEAVE_SCOPE(oldsave); /* may undo an earlier SWITCHSTACK */
6359 SPAGAIN;
a0d0e21e 6360 if (realarray) {
8ec5e241 6361 if (!mg) {
1c0b011c
NIS
6362 if (SvSMAGICAL(ary)) {
6363 PUTBACK;
ad64d0ec 6364 mg_set(MUTABLE_SV(ary));
1c0b011c
NIS
6365 SPAGAIN;
6366 }
6367 if (gimme == G_ARRAY) {
6368 EXTEND(SP, iters);
6369 Copy(AvARRAY(ary), SP + 1, iters, SV*);
6370 SP += iters;
6371 RETURN;
6372 }
8ec5e241 6373 }
1c0b011c 6374 else {
fb73857a 6375 PUTBACK;
d343c3ef 6376 ENTER_with_name("call_PUSH");
36925d9e 6377 call_sv(SV_CONST(PUSH),G_SCALAR|G_DISCARD|G_METHOD_NAMED);
d343c3ef 6378 LEAVE_with_name("call_PUSH");
fb73857a 6379 SPAGAIN;
8ec5e241 6380 if (gimme == G_ARRAY) {
c70927a6 6381 SSize_t i;
8ec5e241
NIS
6382 /* EXTEND should not be needed - we just popped them */
6383 EXTEND(SP, iters);
6384 for (i=0; i < iters; i++) {
6385 SV **svp = av_fetch(ary, i, FALSE);
3280af22 6386 PUSHs((svp) ? *svp : &PL_sv_undef);
8ec5e241 6387 }
1c0b011c
NIS
6388 RETURN;
6389 }
a0d0e21e
LW
6390 }
6391 }
6392 else {
6393 if (gimme == G_ARRAY)
6394 RETURN;
6395 }
7f18b612
YST
6396
6397 GETTARGET;
02c161ef 6398 XPUSHi(iters);
7f18b612 6399 RETURN;
79072805 6400}
85e6fe83 6401
c5917253
NC
6402PP(pp_once)
6403{
6404 dSP;
6405 SV *const sv = PAD_SVl(PL_op->op_targ);
6406
6407 if (SvPADSTALE(sv)) {
6408 /* First time. */
6409 SvPADSTALE_off(sv);
6410 RETURNOP(cLOGOP->op_other);
6411 }
6412 RETURNOP(cLOGOP->op_next);
6413}
6414
c0329465
MB
6415PP(pp_lock)
6416{
39644a26 6417 dSP;
c0329465 6418 dTOPss;
e55aaa0e 6419 SV *retsv = sv;
68795e93 6420 SvLOCK(sv);
f79aa60b
FC
6421 if (SvTYPE(retsv) == SVt_PVAV || SvTYPE(retsv) == SVt_PVHV
6422 || SvTYPE(retsv) == SVt_PVCV) {
e55aaa0e
MB
6423 retsv = refto(retsv);
6424 }
6425 SETs(retsv);
c0329465
MB
6426 RETURN;
6427}
a863c7d1 6428
65bca31a 6429
10088f56 6430/* used for: pp_padany(), pp_custom(); plus any system ops
b1c05ba5
DM
6431 * that aren't implemented on a particular platform */
6432
65bca31a
NC
6433PP(unimplemented_op)
6434{
361ed549
NC
6435 const Optype op_type = PL_op->op_type;
6436 /* Using OP_NAME() isn't going to be helpful here. Firstly, it doesn't cope
6437 with out of range op numbers - it only "special" cases op_custom.
6438 Secondly, as the three ops we "panic" on are padmy, mapstart and custom,
6439 if we get here for a custom op then that means that the custom op didn't
6440 have an implementation. Given that OP_NAME() looks up the custom op
e5576b00
DIM
6441 by its op_ppaddr, likely it will return NULL, unless someone (unhelpfully)
6442 registers &Perl_unimplemented_op as the address of their custom op.
361ed549
NC
6443 NULL doesn't generate a useful error message. "custom" does. */
6444 const char *const name = op_type >= OP_max
e5576b00 6445 ? "[out of range]" : PL_op_name[op_type];
7627e6d0
NC
6446 if(OP_IS_SOCKET(op_type))
6447 DIE(aTHX_ PL_no_sock_func, name);
361ed549 6448 DIE(aTHX_ "panic: unimplemented op %s (#%d) called", name, op_type);
65bca31a
NC
6449}
6450
bea284c8
FC
6451static void
6452S_maybe_unwind_defav(pTHX)
6453{
6454 if (CX_CUR()->cx_type & CXp_HASARGS) {
6455 PERL_CONTEXT *cx = CX_CUR();
6456
6457 assert(CxHASARGS(cx));
6458 cx_popsub_args(cx);
6459 cx->cx_type &= ~CXp_HASARGS;
6460 }
6461}
6462
deb8a388
FC
6463/* For sorting out arguments passed to a &CORE:: subroutine */
6464PP(pp_coreargs)
6465{
6466 dSP;
7fa5bd9b 6467 int opnum = SvIOK(cSVOP_sv) ? (int)SvUV(cSVOP_sv) : 0;
498a02d8 6468 int defgv = PL_opargs[opnum] & OA_DEFGV ||opnum==OP_GLOB, whicharg = 0;
7fa5bd9b 6469 AV * const at_ = GvAV(PL_defgv);
0e80230d
FC
6470 SV **svp = at_ ? AvARRAY(at_) : NULL;
6471 I32 minargs = 0, maxargs = 0, numargs = at_ ? AvFILLp(at_)+1 : 0;
7fa5bd9b 6472 I32 oa = opnum ? PL_opargs[opnum] >> OASHIFT : 0;
46e00a91 6473 bool seen_question = 0;
7fa5bd9b 6474 const char *err = NULL;
3e6568b4 6475 const bool pushmark = PL_op->op_private & OPpCOREARGS_PUSHMARK;
7fa5bd9b 6476
46e00a91
FC
6477 /* Count how many args there are first, to get some idea how far to
6478 extend the stack. */
7fa5bd9b 6479 while (oa) {
bf0571fd 6480 if ((oa & 7) == OA_LIST) { maxargs = I32_MAX; break; }
7fa5bd9b 6481 maxargs++;
46e00a91
FC
6482 if (oa & OA_OPTIONAL) seen_question = 1;
6483 if (!seen_question) minargs++;
7fa5bd9b
FC
6484 oa >>= 4;
6485 }
6486
6487 if(numargs < minargs) err = "Not enough";
6488 else if(numargs > maxargs) err = "Too many";
6489 if (err)
6490 /* diag_listed_as: Too many arguments for %s */
6491 Perl_croak(aTHX_
6492 "%s arguments for %s", err,
2a90c7c6 6493 opnum ? PL_op_desc[opnum] : SvPV_nolen_const(cSVOP_sv)
7fa5bd9b
FC
6494 );
6495
6496 /* Reset the stack pointer. Without this, we end up returning our own
6497 arguments in list context, in addition to the values we are supposed
6498 to return. nextstate usually does this on sub entry, but we need
e1fa07e3 6499 to run the next op with the caller's hints, so we cannot have a
7fa5bd9b 6500 nextstate. */
4ebe6e95 6501 SP = PL_stack_base + CX_CUR()->blk_oldsp;
7fa5bd9b 6502
46e00a91
FC
6503 if(!maxargs) RETURN;
6504
bf0571fd
FC
6505 /* We do this here, rather than with a separate pushmark op, as it has
6506 to come in between two things this function does (stack reset and
6507 arg pushing). This seems the easiest way to do it. */
3e6568b4 6508 if (pushmark) {
bf0571fd
FC
6509 PUTBACK;
6510 (void)Perl_pp_pushmark(aTHX);
6511 }
6512
6513 EXTEND(SP, maxargs == I32_MAX ? numargs : maxargs);
c931b036 6514 PUTBACK; /* The code below can die in various places. */
46e00a91
FC
6515
6516 oa = PL_opargs[opnum] >> OASHIFT;
3e6568b4 6517 for (; oa&&(numargs||!pushmark); (void)(numargs&&(++svp,--numargs))) {
c931b036 6518 whicharg++;
46e00a91
FC
6519 switch (oa & 7) {
6520 case OA_SCALAR:
1efec5ed 6521 try_defsv:
d6d78e19 6522 if (!numargs && defgv && whicharg == minargs + 1) {
195eefec 6523 PUSHs(DEFSV);
d6d78e19
FC
6524 }
6525 else PUSHs(numargs ? svp && *svp ? *svp : &PL_sv_undef : NULL);
46e00a91 6526 break;
bf0571fd
FC
6527 case OA_LIST:
6528 while (numargs--) {
6529 PUSHs(svp && *svp ? *svp : &PL_sv_undef);
6530 svp++;
6531 }
6532 RETURN;
bea284c8
FC
6533 case OA_AVREF:
6534 if (!numargs) {
6535 GV *gv;
6536 if (CvUNIQUE(find_runcv_where(FIND_RUNCV_level_eq,1,NULL)))
6537 gv = PL_argvgv;
6538 else {
6539 S_maybe_unwind_defav(aTHX);
6540 gv = PL_defgv;
6541 }
6542 PUSHs((SV *)GvAVn(gv));
6543 break;
6544 }
6545 if (!svp || !*svp || !SvROK(*svp)
6546 || SvTYPE(SvRV(*svp)) != SVt_PVAV)
6547 DIE(aTHX_
6548 /* diag_listed_as: Type of arg %d to &CORE::%s must be %s*/
6549 "Type of arg %d to &CORE::%s must be array reference",
6550 whicharg, PL_op_desc[opnum]
6551 );
6552 PUSHs(SvRV(*svp));
6553 break;
19c481f4
FC
6554 case OA_HVREF:
6555 if (!svp || !*svp || !SvROK(*svp)
73665bc4
FC
6556 || ( SvTYPE(SvRV(*svp)) != SVt_PVHV
6557 && ( opnum == OP_DBMCLOSE || opnum == OP_DBMOPEN
6558 || SvTYPE(SvRV(*svp)) != SVt_PVAV )))
19c481f4
FC
6559 DIE(aTHX_
6560 /* diag_listed_as: Type of arg %d to &CORE::%s must be %s*/
73665bc4
FC
6561 "Type of arg %d to &CORE::%s must be hash%s reference",
6562 whicharg, PL_op_desc[opnum],
6563 opnum == OP_DBMCLOSE || opnum == OP_DBMOPEN
6564 ? ""
6565 : " or array"
19c481f4
FC
6566 );
6567 PUSHs(SvRV(*svp));
6568 break;
c931b036 6569 case OA_FILEREF:
30901a8a
FC
6570 if (!numargs) PUSHs(NULL);
6571 else if(svp && *svp && SvROK(*svp) && isGV_with_GP(SvRV(*svp)))
c931b036
FC
6572 /* no magic here, as the prototype will have added an extra
6573 refgen and we just want what was there before that */
6574 PUSHs(SvRV(*svp));
6575 else {
6576 const bool constr = PL_op->op_private & whicharg;
6577 PUSHs(S_rv2gv(aTHX_
6578 svp && *svp ? *svp : &PL_sv_undef,
b54f893d 6579 constr, cBOOL(CopHINTS_get(PL_curcop) & HINT_STRICT_REFS),
c931b036
FC
6580 !constr
6581 ));
6582 }
6583 break;
c72a5629 6584 case OA_SCALARREF:
1efec5ed
FC
6585 if (!numargs) goto try_defsv;
6586 else {
17008668
FC
6587 const bool wantscalar =
6588 PL_op->op_private & OPpCOREARGS_SCALARMOD;
c72a5629 6589 if (!svp || !*svp || !SvROK(*svp)
17008668
FC
6590 /* We have to permit globrefs even for the \$ proto, as
6591 *foo is indistinguishable from ${\*foo}, and the proto-
6592 type permits the latter. */
6593 || SvTYPE(SvRV(*svp)) > (
efe889ae 6594 wantscalar ? SVt_PVLV
46bef06f
FC
6595 : opnum == OP_LOCK || opnum == OP_UNDEF
6596 ? SVt_PVCV
efe889ae 6597 : SVt_PVHV
17008668 6598 )
c72a5629
FC
6599 )
6600 DIE(aTHX_
17008668 6601 "Type of arg %d to &CORE::%s must be %s",
46bef06f 6602 whicharg, PL_op_name[opnum],
17008668
FC
6603 wantscalar
6604 ? "scalar reference"
46bef06f 6605 : opnum == OP_LOCK || opnum == OP_UNDEF
efe889ae
FC
6606 ? "reference to one of [$@%&*]"
6607 : "reference to one of [$@%*]"
c72a5629
FC
6608 );
6609 PUSHs(SvRV(*svp));
bea284c8 6610 if (opnum == OP_UNDEF && SvRV(*svp) == (SV *)PL_defgv) {
88bb468b 6611 /* Undo @_ localisation, so that sub exit does not undo
04e686b8 6612 part of our undeffing. */
bea284c8 6613 S_maybe_unwind_defav(aTHX);
88bb468b 6614 }
17008668 6615 }
1efec5ed 6616 break;
46e00a91 6617 default:
46e00a91
FC
6618 DIE(aTHX_ "panic: unknown OA_*: %x", (unsigned)(oa&7));
6619 }
6620 oa = oa >> 4;
6621 }
6622
deb8a388
FC
6623 RETURN;
6624}
6625
a2232057
DM
6626/* Implement CORE::keys(),values(),each().
6627 *
6628 * We won't know until run-time whether the arg is an array or hash,
6629 * so this op calls
6630 *
6631 * pp_keys/pp_values/pp_each
6632 * or
6633 * pp_akeys/pp_avalues/pp_aeach
6634 *
6635 * as appropriate (or whatever pp function actually implements the OP_FOO
6636 * functionality for each FOO).
6637 */
6638
88101882
FC
6639PP(pp_avhvswitch)
6640{
a73158aa 6641 dVAR; dSP;
73665bc4
FC
6642 return PL_ppaddr[
6643 (SvTYPE(TOPs) == SVt_PVAV ? OP_AEACH : OP_EACH)
e1e26374 6644 + (PL_op->op_private & OPpAVHVSWITCH_MASK)
73665bc4 6645 ](aTHX);
88101882
FC
6646}
6647
84ed0108
FC
6648PP(pp_runcv)
6649{
6650 dSP;
6651 CV *cv;
6652 if (PL_op->op_private & OPpOFFBYONE) {
db4cf31d 6653 cv = find_runcv_where(FIND_RUNCV_level_eq, 1, NULL);
84ed0108
FC
6654 }
6655 else cv = find_runcv(NULL);
e157a82b 6656 XPUSHs(CvEVAL(cv) ? &PL_sv_undef : sv_2mortal(newRV((SV *)cv)));
84ed0108
FC
6657 RETURN;
6658}
6659
05a34802 6660static void
2331e434 6661S_localise_aelem_lval(pTHX_ AV * const av, SV * const keysv,
05a34802
FC
6662 const bool can_preserve)
6663{
2331e434 6664 const SSize_t ix = SvIV(keysv);
05a34802
FC
6665 if (can_preserve ? av_exists(av, ix) : TRUE) {
6666 SV ** const svp = av_fetch(av, ix, 1);
6667 if (!svp || !*svp)
6668 Perl_croak(aTHX_ PL_no_aelem, ix);
6669 save_aelem(av, ix, svp);
6670 }
6671 else
6672 SAVEADELETE(av, ix);
6673}
6674
5f94141d
FC
6675static void
6676S_localise_helem_lval(pTHX_ HV * const hv, SV * const keysv,
6677 const bool can_preserve)
6678{
6679 if (can_preserve ? hv_exists_ent(hv, keysv, 0) : TRUE) {
6680 HE * const he = hv_fetch_ent(hv, keysv, 1, 0);
6681 SV ** const svp = he ? &HeVAL(he) : NULL;
6682 if (!svp || !*svp)
6683 Perl_croak(aTHX_ PL_no_helem_sv, SVfARG(keysv));
6684 save_helem_flags(hv, keysv, svp, 0);
6685 }
6686 else
6687 SAVEHDELETE(hv, keysv);
6688}
6689
9782ce69
FC
6690static void
6691S_localise_gv_slot(pTHX_ GV *gv, U8 type)
6692{
6693 if (type == OPpLVREF_SV) {
6694 save_pushptrptr(gv, SvREFCNT_inc_simple(GvSV(gv)), SAVEt_GVSV);
6695 GvSV(gv) = 0;
6696 }
6697 else if (type == OPpLVREF_AV)
6698 /* XXX Inefficient, as it creates a new AV, which we are
6699 about to clobber. */
6700 save_ary(gv);
6701 else {
6702 assert(type == OPpLVREF_HV);
6703 /* XXX Likewise inefficient. */
6704 save_hash(gv);
6705 }
6706}
6707
6708
254da51f
FC
6709PP(pp_refassign)
6710{
4fec8804 6711 dSP;
6102323a 6712 SV * const key = PL_op->op_private & OPpLVREF_ELEM ? POPs : NULL;
d8a875d9 6713 SV * const left = PL_op->op_flags & OPf_STACKED ? POPs : NULL;
4fec8804 6714 dTOPss;
3f114923 6715 const char *bad = NULL;
ac0da85a 6716 const U8 type = PL_op->op_private & OPpLVREF_TYPE;
4fec8804 6717 if (!SvROK(sv)) DIE(aTHX_ "Assigned value is not a reference");
ac0da85a 6718 switch (type) {
3f114923
FC
6719 case OPpLVREF_SV:
6720 if (SvTYPE(SvRV(sv)) > SVt_PVLV)
6721 bad = " SCALAR";
6722 break;
6723 case OPpLVREF_AV:
6724 if (SvTYPE(SvRV(sv)) != SVt_PVAV)
6725 bad = "n ARRAY";
6726 break;
6727 case OPpLVREF_HV:
6728 if (SvTYPE(SvRV(sv)) != SVt_PVHV)
6729 bad = " HASH";
6730 break;
6731 case OPpLVREF_CV:
6732 if (SvTYPE(SvRV(sv)) != SVt_PVCV)
6733 bad = " CODE";
6734 }
6735 if (bad)
1f8155a2 6736 /* diag_listed_as: Assigned value is not %s reference */
3f114923 6737 DIE(aTHX_ "Assigned value is not a%s reference", bad);
b943805e
JH
6738 {
6739 MAGIC *mg;
6740 HV *stash;
d8a875d9
FC
6741 switch (left ? SvTYPE(left) : 0) {
6742 case 0:
cf5d2d91
FC
6743 {
6744 SV * const old = PAD_SV(ARGTARG);
d8a875d9 6745 PAD_SETSV(ARGTARG, SvREFCNT_inc_NN(SvRV(sv)));
cf5d2d91 6746 SvREFCNT_dec(old);
3ad7d304
FC
6747 if ((PL_op->op_private & (OPpLVAL_INTRO|OPpPAD_STATE))
6748 == OPpLVAL_INTRO)
fc048fcf 6749 SAVECLEARSV(PAD_SVl(ARGTARG));
d8a875d9 6750 break;
cf5d2d91 6751 }
d8a875d9 6752 case SVt_PVGV:
2a57afb1 6753 if (PL_op->op_private & OPpLVAL_INTRO) {
9782ce69 6754 S_localise_gv_slot(aTHX_ (GV *)left, type);
2a57afb1 6755 }
d8a875d9
FC
6756 gv_setref(left, sv);
6757 SvSETMAGIC(left);
6102323a
FC
6758 break;
6759 case SVt_PVAV:
69a23520 6760 assert(key);
40d2b828 6761 if (UNLIKELY(PL_op->op_private & OPpLVAL_INTRO)) {
2331e434 6762 S_localise_aelem_lval(aTHX_ (AV *)left, key,
05a34802 6763 SvCANEXISTDELETE(left));
40d2b828 6764 }
6102323a
FC
6765 av_store((AV *)left, SvIV(key), SvREFCNT_inc_simple_NN(SvRV(sv)));
6766 break;
5f94141d 6767 case SVt_PVHV:
69a23520
JH
6768 if (UNLIKELY(PL_op->op_private & OPpLVAL_INTRO)) {
6769 assert(key);
5f94141d
FC
6770 S_localise_helem_lval(aTHX_ (HV *)left, key,
6771 SvCANEXISTDELETE(left));
69a23520 6772 }
7fcb36d5 6773 (void)hv_store_ent((HV *)left, key, SvREFCNT_inc_simple_NN(SvRV(sv)), 0);
d8a875d9 6774 }
4fec8804
FC
6775 if (PL_op->op_flags & OPf_MOD)
6776 SETs(sv_2mortal(newSVsv(sv)));
6777 /* XXX else can weak references go stale before they are read, e.g.,
6778 in leavesub? */
6779 RETURN;
b943805e 6780 }
254da51f
FC
6781}
6782
4c5bab50
FC
6783PP(pp_lvref)
6784{
26a50d99
FC
6785 dSP;
6786 SV * const ret = sv_2mortal(newSV_type(SVt_PVMG));
6102323a 6787 SV * const elem = PL_op->op_private & OPpLVREF_ELEM ? POPs : NULL;
2a57afb1 6788 SV * const arg = PL_op->op_flags & OPf_STACKED ? POPs : NULL;
9782ce69
FC
6789 MAGIC * const mg = sv_magicext(ret, arg, PERL_MAGIC_lvref,
6790 &PL_vtbl_lvref, (char *)elem,
23270f96 6791 elem ? HEf_SVKEY : (I32)ARGTARG);
9782ce69 6792 mg->mg_private = PL_op->op_private;
d39c26a6
FC
6793 if (PL_op->op_private & OPpLVREF_ITER)
6794 mg->mg_flags |= MGf_PERSIST;
9846cd95 6795 if (UNLIKELY(PL_op->op_private & OPpLVAL_INTRO)) {
40d2b828 6796 if (elem) {
38bb0011
JH
6797 MAGIC *mg;
6798 HV *stash;
6799 assert(arg);
6800 {
6801 const bool can_preserve = SvCANEXISTDELETE(arg);
6802 if (SvTYPE(arg) == SVt_PVAV)
6803 S_localise_aelem_lval(aTHX_ (AV *)arg, elem, can_preserve);
6804 else
6805 S_localise_helem_lval(aTHX_ (HV *)arg, elem, can_preserve);
6806 }
40d2b828
FC
6807 }
6808 else if (arg) {
a8e41ef4 6809 S_localise_gv_slot(aTHX_ (GV *)arg,
9782ce69 6810 PL_op->op_private & OPpLVREF_TYPE);
2a57afb1 6811 }
3ad7d304 6812 else if (!(PL_op->op_private & OPpPAD_STATE))
c146a62a 6813 SAVECLEARSV(PAD_SVl(ARGTARG));
1199b01a 6814 }
c146a62a
FC
6815 XPUSHs(ret);
6816 RETURN;
4c5bab50 6817}
84ed0108 6818
16b99412
FC
6819PP(pp_lvrefslice)
6820{
a95dad8a 6821 dSP; dMARK;
0ca7b7f7
FC
6822 AV * const av = (AV *)POPs;
6823 const bool localizing = PL_op->op_private & OPpLVAL_INTRO;
6824 bool can_preserve = FALSE;
6825
9846cd95 6826 if (UNLIKELY(localizing)) {
0ca7b7f7
FC
6827 MAGIC *mg;
6828 HV *stash;
6829 SV **svp;
6830
6831 can_preserve = SvCANEXISTDELETE(av);
6832
6833 if (SvTYPE(av) == SVt_PVAV) {
6834 SSize_t max = -1;
6835
6836 for (svp = MARK + 1; svp <= SP; svp++) {
6837 const SSize_t elem = SvIV(*svp);
6838 if (elem > max)
6839 max = elem;
6840 }
6841 if (max > AvMAX(av))
6842 av_extend(av, max);
6843 }
6844 }
6845
6846 while (++MARK <= SP) {
6847 SV * const elemsv = *MARK;
b97fe865
DM
6848 if (UNLIKELY(localizing)) {
6849 if (SvTYPE(av) == SVt_PVAV)
6850 S_localise_aelem_lval(aTHX_ av, elemsv, can_preserve);
6851 else
6852 S_localise_helem_lval(aTHX_ (HV *)av, elemsv, can_preserve);
6853 }
0ca7b7f7
FC
6854 *MARK = sv_2mortal(newSV_type(SVt_PVMG));
6855 sv_magic(*MARK,(SV *)av,PERL_MAGIC_lvref,(char *)elemsv,HEf_SVKEY);
6856 }
6857 RETURN;
16b99412
FC
6858}
6859
2882b3ff
FC
6860PP(pp_lvavref)
6861{
bdaf10a5
FC
6862 if (PL_op->op_flags & OPf_STACKED)
6863 Perl_pp_rv2av(aTHX);
6864 else
6865 Perl_pp_padav(aTHX);
6866 {
6867 dSP;
6868 dTOPss;
6869 SETs(0); /* special alias marker that aassign recognises */
6870 XPUSHs(sv);
6871 RETURN;
6872 }
2882b3ff
FC
6873}
6874
b77472f9
FC
6875PP(pp_anonconst)
6876{
6877 dSP;
6878 dTOPss;
6879 SETs(sv_2mortal((SV *)newCONSTSUB(SvTYPE(CopSTASH(PL_curcop))==SVt_PVHV
6880 ? CopSTASH(PL_curcop)
6881 : NULL,
6882 NULL, SvREFCNT_inc_simple_NN(sv))));
6883 RETURN;
6884}
6885
4fa06845
DM
6886
6887/* process one subroutine argument - typically when the sub has a signature:
6888 * introduce PL_curpad[op_targ] and assign to it the value
6889 * for $: (OPf_STACKED ? *sp : $_[N])
6890 * for @/%: @_[N..$#_]
6891 *
a8e41ef4 6892 * It's equivalent to
4fa06845
DM
6893 * my $foo = $_[N];
6894 * or
6895 * my $foo = (value-on-stack)
6896 * or
6897 * my @foo = @_[N..$#_]
6898 * etc
4fa06845
DM
6899 */
6900
6901PP(pp_argelem)
6902{
6903 dTARG;
6904 SV *val;
6905 SV ** padentry;
6906 OP *o = PL_op;
6907 AV *defav = GvAV(PL_defgv); /* @_ */
6daeaaa3 6908 IV ix = PTR2IV(cUNOP_AUXo->op_aux);
4fa06845 6909 IV argc;
4fa06845
DM
6910
6911 /* do 'my $var, @var or %var' action */
6912 padentry = &(PAD_SVl(o->op_targ));
6913 save_clearsv(padentry);
6914 targ = *padentry;
6915
6916 if ((o->op_private & OPpARGELEM_MASK) == OPpARGELEM_SV) {
6917 if (o->op_flags & OPf_STACKED) {
6918 dSP;
6919 val = POPs;
6920 PUTBACK;
6921 }
6922 else {
f6ca42c7 6923 SV **svp;
4fa06845 6924 /* should already have been checked */
f6ca42c7 6925 assert(ix >= 0);
6daeaaa3
DM
6926#if IVSIZE > PTRSIZE
6927 assert(ix <= SSize_t_MAX);
6928#endif
f6ca42c7
DM
6929
6930 svp = av_fetch(defav, ix, FALSE);
6931 val = svp ? *svp : &PL_sv_undef;
4fa06845
DM
6932 }
6933
6934 /* $var = $val */
6935
6936 /* cargo-culted from pp_sassign */
6937 assert(TAINTING_get || !TAINT_get);
6938 if (UNLIKELY(TAINT_get) && !SvTAINTED(val))
6939 TAINT_NOT;
6940
f6ca42c7 6941 SvSetMagicSV(targ, val);
4fa06845
DM
6942 return o->op_next;
6943 }
6944
6945 /* must be AV or HV */
6946
6947 assert(!(o->op_flags & OPf_STACKED));
f6ca42c7 6948 argc = ((IV)AvFILL(defav) + 1) - ix;
4fa06845
DM
6949
6950 /* This is a copy of the relevant parts of pp_aassign().
4fa06845
DM
6951 */
6952 if ((o->op_private & OPpARGELEM_MASK) == OPpARGELEM_AV) {
f6ca42c7
DM
6953 IV i;
6954
6955 if (AvFILL((AV*)targ) > -1) {
6956 /* target should usually be empty. If we get get
6957 * here, someone's been doing some weird closure tricks.
6958 * Make a copy of all args before clearing the array,
6959 * to avoid the equivalent of @a = ($a[0]) prematurely freeing
6960 * elements. See similar code in pp_aassign.
6961 */
6962 for (i = 0; i < argc; i++) {
6963 SV **svp = av_fetch(defav, ix + i, FALSE);
6964 SV *newsv = newSV(0);
6965 sv_setsv_flags(newsv,
6966 svp ? *svp : &PL_sv_undef,
6967 (SV_DO_COW_SVSETSV|SV_NOSTEAL));
6968 if (!av_store(defav, ix + i, newsv))
6969 SvREFCNT_dec_NN(newsv);
6970 }
6971 av_clear((AV*)targ);
6972 }
6973
6974 if (argc <= 0)
6975 return o->op_next;
4fa06845 6976
4fa06845
DM
6977 av_extend((AV*)targ, argc);
6978
f6ca42c7 6979 i = 0;
4fa06845
DM
6980 while (argc--) {
6981 SV *tmpsv;
f6ca42c7
DM
6982 SV **svp = av_fetch(defav, ix + i, FALSE);
6983 SV *val = svp ? *svp : &PL_sv_undef;
4fa06845 6984 tmpsv = newSV(0);
f6ca42c7 6985 sv_setsv(tmpsv, val);
4fa06845
DM
6986 av_store((AV*)targ, i++, tmpsv);
6987 TAINT_NOT;
6988 }
6989
6990 }
6991 else {
f6ca42c7
DM
6992 IV i;
6993
4fa06845
DM
6994 assert((o->op_private & OPpARGELEM_MASK) == OPpARGELEM_HV);
6995
f6ca42c7
DM
6996 if (SvRMAGICAL(targ) || HvUSEDKEYS((HV*)targ)) {
6997 /* see "target should usually be empty" comment above */
6998 for (i = 0; i < argc; i++) {
6999 SV **svp = av_fetch(defav, ix + i, FALSE);
7000 SV *newsv = newSV(0);
7001 sv_setsv_flags(newsv,
7002 svp ? *svp : &PL_sv_undef,
7003 (SV_DO_COW_SVSETSV|SV_NOSTEAL));
7004 if (!av_store(defav, ix + i, newsv))
7005 SvREFCNT_dec_NN(newsv);
7006 }
7007 hv_clear((HV*)targ);
7008 }
7009
7010 if (argc <= 0)
7011 return o->op_next;
4fa06845 7012 assert(argc % 2 == 0);
4fa06845 7013
f6ca42c7 7014 i = 0;
4fa06845
DM
7015 while (argc) {
7016 SV *tmpsv;
f6ca42c7
DM
7017 SV **svp;
7018 SV *key;
7019 SV *val;
7020
7021 svp = av_fetch(defav, ix + i++, FALSE);
7022 key = svp ? *svp : &PL_sv_undef;
7023 svp = av_fetch(defav, ix + i++, FALSE);
7024 val = svp ? *svp : &PL_sv_undef;
4fa06845 7025
4fa06845
DM
7026 argc -= 2;
7027 if (UNLIKELY(SvGMAGICAL(key)))
7028 key = sv_mortalcopy(key);
7029 tmpsv = newSV(0);
7030 sv_setsv(tmpsv, val);
7031 hv_store_ent((HV*)targ, key, tmpsv, 0);
7032 TAINT_NOT;
7033 }
7034 }
7035
7036 return o->op_next;
7037}
7038
7039/* Handle a default value for one subroutine argument (typically as part
7040 * of a subroutine signature).
7041 * It's equivalent to
7042 * @_ > op_targ ? $_[op_targ] : result_of(op_other)
7043 *
7044 * Intended to be used where op_next is an OP_ARGELEM
7045 *
7046 * We abuse the op_targ field slightly: it's an index into @_ rather than
7047 * into PL_curpad.
7048 */
7049
7050PP(pp_argdefelem)
7051{
7052 OP * const o = PL_op;
7053 AV *defav = GvAV(PL_defgv); /* @_ */
6daeaaa3 7054 IV ix = (IV)o->op_targ;
4fa06845 7055
f6ca42c7 7056 assert(ix >= 0);
6daeaaa3
DM
7057#if IVSIZE > PTRSIZE
7058 assert(ix <= SSize_t_MAX);
7059#endif
f6ca42c7
DM
7060
7061 if (AvFILL(defav) >= ix) {
4fa06845 7062 dSP;
f6ca42c7
DM
7063 SV **svp = av_fetch(defav, ix, FALSE);
7064 SV *val = svp ? *svp : &PL_sv_undef;
7065 XPUSHs(val);
4fa06845
DM
7066 RETURN;
7067 }
7068 return cLOGOPo->op_other;
7069}
7070
7071
ac7609e4
AC
7072static SV *
7073S_find_runcv_name(void)
7074{
7075 dTHX;
7076 CV *cv;
7077 GV *gv;
7078 SV *sv;
7079
7080 cv = find_runcv(0);
7081 if (!cv)
7082 return &PL_sv_no;
7083
7084 gv = CvGV(cv);
7085 if (!gv)
7086 return &PL_sv_no;
7087
7088 sv = sv_2mortal(newSV(0));
7089 gv_fullname4(sv, gv, NULL, TRUE);
7090 return sv;
7091}
4fa06845 7092
f417cfa9 7093/* Check a sub's arguments - i.e. that it has the correct number of args
4fa06845
DM
7094 * (and anything else we might think of in future). Typically used with
7095 * signatured subs.
7096 */
7097
7098PP(pp_argcheck)
7099{
7100 OP * const o = PL_op;
f417cfa9 7101 struct op_argcheck_aux *aux = (struct op_argcheck_aux *)cUNOP_AUXo->op_aux;
e6158756
DM
7102 UV params = aux->params;
7103 UV opt_params = aux->opt_params;
f417cfa9 7104 char slurpy = aux->slurpy;
4fa06845 7105 AV *defav = GvAV(PL_defgv); /* @_ */
7d769928 7106 UV argc;
4fa06845
DM
7107 bool too_few;
7108
7109 assert(!SvMAGICAL(defav));
7d769928 7110 argc = (UV)(AvFILLp(defav) + 1);
4fa06845
DM
7111 too_few = (argc < (params - opt_params));
7112
7113 if (UNLIKELY(too_few || (!slurpy && argc > params)))
ac7609e4
AC
7114 /* diag_listed_as: Too few arguments for subroutine '%s' */
7115 /* diag_listed_as: Too many arguments for subroutine '%s' */
7116 Perl_croak_caller("Too %s arguments for subroutine '%" SVf "'",
7117 too_few ? "few" : "many", S_find_runcv_name());
4fa06845
DM
7118
7119 if (UNLIKELY(slurpy == '%' && argc > params && (argc - params) % 2))
ac7609e4
AC
7120 /* diag_listed_as: Odd name/value argument for subroutine '%s' */
7121 Perl_croak_caller("Odd name/value argument for subroutine '%" SVf "'",
7122 S_find_runcv_name());
4fa06845
DM
7123
7124 return NORMAL;
7125}
7126
813e85a0
PE
7127PP(pp_isa)
7128{
7129 dSP;
7130 SV *left, *right;
7131
7132 right = POPs;
7133 left = TOPs;
7134
7135 SETs(boolSV(sv_isa_sv(left, right)));
7136 RETURN;
7137}
7138
02b85d3d
Z
7139PP(pp_cmpchain_and)
7140{
7141 dSP;
7142 SV *result = POPs;
7143 PUTBACK;
7144 if (SvTRUE_NN(result)) {
7145 return cLOGOP->op_other;
7146 } else {
7147 TOPs = result;
7148 return NORMAL;
7149 }
7150}
7151
7152PP(pp_cmpchain_dup)
7153{
7154 dSP;
7155 SV *right = TOPs;
7156 SV *left = TOPm1s;
7157 TOPm1s = right;
7158 TOPs = left;
7159 XPUSHs(right);
7160 RETURN;
7161}
7162
e609e586 7163/*
14d04a33 7164 * ex: set ts=8 sts=4 sw=4 et:
37442d52 7165 */