/* This file contains some common functions needed to carry out certain
* ops. For example, both pp_sprintf() and pp_prtf() call the function
- * do_printf() found in this file.
+ * do_sprintf() found in this file.
*/
#include "EXTERN.h"
#include <signal.h>
#endif
-STATIC I32
-S_do_trans_simple(pTHX_ SV * const sv)
+
+/* Helper function for do_trans().
+ * Handles cases where the search and replacement charlists aren't UTF-8,
+ * aren't identical, and neither the /d nor /s flag is present.
+ *
+ * sv may or may not be utf8. Note that no code point above 255 can possibly
+ * be in the to-translate set
+ */
+
+STATIC Size_t
+S_do_trans_simple(pTHX_ SV * const sv, const OPtrans_map * const tbl)
{
- dVAR;
- I32 matches = 0;
+ Size_t matches = 0;
STRLEN len;
U8 *s = (U8*)SvPV_nomg(sv,len);
U8 * const send = s+len;
- const short * const tbl = (short*)cPVOP->op_pv;
PERL_ARGS_ASSERT_DO_TRANS_SIMPLE;
- if (!tbl)
- Perl_croak(aTHX_ "panic: do_trans_simple line %d",__LINE__);
-
/* First, take care of non-UTF-8 input strings, because they're easy */
if (!SvUTF8(sv)) {
while (s < send) {
- const I32 ch = tbl[*s];
+ const short ch = tbl->map[*s];
if (ch >= 0) {
matches++;
*s = (U8)ch;
SvSETMAGIC(sv);
}
else {
- const I32 grows = PL_op->op_private & OPpTRANS_GROWS;
+ const bool grows = cBOOL(PL_op->op_private & OPpTRANS_GROWS);
U8 *d;
U8 *dstart;
- /* Allow for expansion: $_="a".chr(400); tr/a/\xFE/, FE needs encoding */
+ /* Allow for worst-case expansion: Each input byte can become 2. For a
+ * given input character, this happens when it occupies a single byte
+ * under UTF-8, but is to be translated to something that occupies two:
+ * $_="a".chr(400); tr/a/\xFE/, FE needs encoding. */
if (grows)
Newx(d, len*2+1, U8);
else
dstart = d;
while (s < send) {
STRLEN ulen;
- I32 ch;
+ short ch;
/* Need to check this, otherwise 128..255 won't match */
const UV c = utf8n_to_uvchr(s, send - s, &ulen, UTF8_ALLOW_DEFAULT);
- if (c < 0x100 && (ch = tbl[c]) >= 0) {
+ if (c < 0x100 && (ch = tbl->map[c]) >= 0) {
matches++;
- d = uvchr_to_utf8(d, ch);
+ d = uvchr_to_utf8(d, (UV)ch);
s += ulen;
}
else { /* No match -> copy */
return matches;
}
-STATIC I32
-S_do_trans_count(pTHX_ SV * const sv)
+
+/* Helper function for do_trans().
+ * Handles cases where the search and replacement charlists are identical and
+ * non-utf8: so the string isn't modified, and only a count of modifiable
+ * chars is needed.
+ *
+ * Note that it doesn't handle /d or /s, since these modify the string even if
+ * the replacement list is empty.
+ *
+ * sv may or may not be utf8. Note that no code point above 255 can possibly
+ * be in the to-translate set
+ */
+
+STATIC Size_t
+S_do_trans_count(pTHX_ SV * const sv, const OPtrans_map * const tbl)
{
- dVAR;
STRLEN len;
const U8 *s = (const U8*)SvPV_nomg_const(sv, len);
const U8 * const send = s + len;
- I32 matches = 0;
- const short * const tbl = (short*)cPVOP->op_pv;
+ Size_t matches = 0;
PERL_ARGS_ASSERT_DO_TRANS_COUNT;
- if (!tbl)
- Perl_croak(aTHX_ "panic: do_trans_count line %d",__LINE__);
-
if (!SvUTF8(sv)) {
while (s < send) {
- if (tbl[*s++] >= 0)
+ if (tbl->map[*s++] >= 0)
matches++;
}
}
else {
- const I32 complement = PL_op->op_private & OPpTRANS_COMPLEMENT;
+ const bool complement = cBOOL(PL_op->op_private & OPpTRANS_COMPLEMENT);
while (s < send) {
STRLEN ulen;
const UV c = utf8n_to_uvchr(s, send - s, &ulen, UTF8_ALLOW_DEFAULT);
if (c < 0x100) {
- if (tbl[c] >= 0)
+ if (tbl->map[c] >= 0)
matches++;
} else if (complement)
matches++;
return matches;
}
-STATIC I32
-S_do_trans_complex(pTHX_ SV * const sv)
+
+/* Helper function for do_trans().
+ * Handles cases where the search and replacement charlists aren't identical
+ * and both are non-utf8, and one or both of /d, /s is specified.
+ *
+ * sv may or may not be utf8. Note that no code point above 255 can possibly
+ * be in the to-translate set
+ */
+
+STATIC Size_t
+S_do_trans_complex(pTHX_ SV * const sv, const OPtrans_map * const tbl)
{
- dVAR;
STRLEN len;
U8 *s = (U8*)SvPV_nomg(sv, len);
U8 * const send = s+len;
- I32 matches = 0;
- const short * const tbl = (short*)cPVOP->op_pv;
+ Size_t matches = 0;
PERL_ARGS_ASSERT_DO_TRANS_COMPLEX;
- if (!tbl)
- Perl_croak(aTHX_ "panic: do_trans_complex line %d",__LINE__);
-
if (!SvUTF8(sv)) {
U8 *d = s;
U8 * const dstart = d;
if (PL_op->op_private & OPpTRANS_SQUASH) {
const U8* p = send;
while (s < send) {
- const I32 ch = tbl[*s];
+ const short ch = tbl->map[*s];
if (ch >= 0) {
*d = (U8)ch;
matches++;
if (p != d - 1 || *p != *d)
p = d++;
}
- else if (ch == -1) /* -1 is unmapped character */
- *d++ = *s;
- else if (ch == -2) /* -2 is delete character */
+ else if (ch == (short) TR_UNMAPPED)
+ *d++ = *s;
+ else if (ch == (short) TR_DELETE)
matches++;
s++;
}
}
- else {
+ else { /* Not to squash */
while (s < send) {
- const I32 ch = tbl[*s];
+ const short ch = tbl->map[*s];
if (ch >= 0) {
matches++;
*d++ = (U8)ch;
}
- else if (ch == -1) /* -1 is unmapped character */
+ else if (ch == (short) TR_UNMAPPED)
*d++ = *s;
- else if (ch == -2) /* -2 is delete character */
+ else if (ch == (short) TR_DELETE)
matches++;
s++;
}
SvCUR_set(sv, d - dstart);
}
else { /* is utf8 */
- const I32 complement = PL_op->op_private & OPpTRANS_COMPLEMENT;
- const I32 grows = PL_op->op_private & OPpTRANS_GROWS;
- const I32 del = PL_op->op_private & OPpTRANS_DELETE;
+ const bool squash = cBOOL(PL_op->op_private & OPpTRANS_SQUASH);
+ const bool grows = cBOOL(PL_op->op_private & OPpTRANS_GROWS);
U8 *d;
U8 *dstart;
- STRLEN rlen = 0;
+ Size_t size = tbl->size;
+
+ /* What the mapping of the previous character was to. If the new
+ * character has the same mapping, it is squashed from the output (but
+ * still is included in the count) */
+ UV pch = TR_OOB;
if (grows)
+ /* Allow for worst-case expansion: Each input byte can become 2.
+ * For a given input character, this happens when it occupies a
+ * single byte under UTF-8, but is to be translated to something
+ * that occupies two: */
Newx(d, len*2+1, U8);
else
d = s;
dstart = d;
- if (complement && !del)
- rlen = tbl[0x100];
- if (PL_op->op_private & OPpTRANS_SQUASH) {
- UV pch = 0xfeedface;
- while (s < send) {
- STRLEN len;
- const UV comp = utf8n_to_uvchr(s, send - s, &len,
- UTF8_ALLOW_DEFAULT);
- I32 ch;
+ while (s < send) {
+ STRLEN len;
+ const UV comp = utf8n_to_uvchr(s, send - s, &len,
+ UTF8_ALLOW_DEFAULT);
+ UV ch;
+ short sch;
+
+ sch = tbl->map[comp >= size ? size : comp];
+
+ if (sch >= 0) {
+ ch = (UV)sch;
+ replace:
+ matches++;
+ if (LIKELY(!squash || ch != pch)) {
+ d = uvchr_to_utf8(d, ch);
+ pch = ch;
+ }
+ s += len;
+ continue;
+ }
+ else if (sch == (short) TR_UNMAPPED) {
+ Move(s, d, len, U8);
+ d += len;
+ }
+ else if (sch == (short) TR_DELETE)
+ matches++;
+ else {
+ assert(sch == (short) TR_R_EMPTY); /* empty replacement */
+ ch = comp;
+ goto replace;
+ }
+
+ s += len;
+ pch = TR_OOB;
+ }
- if (comp > 0xff) {
- if (!complement) {
- Move(s, d, len, U8);
- d += len;
- }
- else {
- matches++;
- if (!del) {
- ch = (rlen == 0) ? (I32)comp :
- (comp - 0x100 < rlen) ?
- tbl[comp+1] : tbl[0x100+rlen];
- if ((UV)ch != pch) {
- d = uvchr_to_utf8(d, ch);
- pch = (UV)ch;
- }
- s += len;
- continue;
- }
- }
- }
- else if ((ch = tbl[comp]) >= 0) {
- matches++;
- if ((UV)ch != pch) {
- d = uvchr_to_utf8(d, ch);
- pch = (UV)ch;
- }
- s += len;
- continue;
- }
- else if (ch == -1) { /* -1 is unmapped character */
- Move(s, d, len, U8);
- d += len;
- }
- else if (ch == -2) /* -2 is delete character */
- matches++;
- s += len;
- pch = 0xfeedface;
- }
- }
- else {
- while (s < send) {
- STRLEN len;
- const UV comp = utf8n_to_uvchr(s, send - s, &len,
- UTF8_ALLOW_DEFAULT);
- I32 ch;
- if (comp > 0xff) {
- if (!complement) {
- Move(s, d, len, U8);
- d += len;
- }
- else {
- matches++;
- if (!del) {
- if (comp - 0x100 < rlen)
- d = uvchr_to_utf8(d, tbl[comp+1]);
- else
- d = uvchr_to_utf8(d, tbl[0x100+rlen]);
- }
- }
- }
- else if ((ch = tbl[comp]) >= 0) {
- d = uvchr_to_utf8(d, ch);
- matches++;
- }
- else if (ch == -1) { /* -1 is unmapped character */
- Move(s, d, len, U8);
- d += len;
- }
- else if (ch == -2) /* -2 is delete character */
- matches++;
- s += len;
- }
- }
if (grows) {
sv_setpvn(sv, (char*)dstart, d - dstart);
Safefree(dstart);
return matches;
}
-STATIC I32
+
+/* Helper function for do_trans().
+ * Handles utf8 cases(*) not involving the /c, /d, /s flags,
+ * and where search and replacement charlists aren't identical.
+ * (*) i.e. where the search or replacement charlists are utf8. sv may
+ * or may not be utf8.
+ */
+
+STATIC Size_t
S_do_trans_simple_utf8(pTHX_ SV * const sv)
{
- dVAR;
U8 *s;
U8 *send;
U8 *d;
U8 *start;
U8 *dstart, *dend;
- I32 matches = 0;
- const I32 grows = PL_op->op_private & OPpTRANS_GROWS;
+ Size_t matches = 0;
+ const bool grows = cBOOL(PL_op->op_private & OPpTRANS_GROWS);
STRLEN len;
SV* const rv =
#ifdef USE_ITHREADS
s = (U8*)SvPV_nomg(sv, len);
if (!SvUTF8(sv)) {
- const U8 *t = s;
- const U8 * const e = s + len;
- while (t < e) {
- const U8 ch = *t++;
- hibit = !NATIVE_BYTE_IS_INVARIANT(ch);
- if (hibit) {
- s = bytes_to_utf8(s, &len);
- break;
- }
+ hibit = ! is_utf8_invariant_string(s, len);
+ if (hibit) {
+ s = bytes_to_utf8(s, &len);
}
}
send = s + len;
return matches;
}
-STATIC I32
+
+/* Helper function for do_trans().
+ * Handles utf8 cases(*) where search and replacement charlists are
+ * identical: so the string isn't modified, and only a count of modifiable
+ * chars is needed.
+ * Note that it doesn't handle /d or /s, since these modify the string
+ * even if the replacement charlist is empty.
+ * (*) i.e. where the search or replacement charlists are utf8. sv may
+ * or may not be utf8.
+ */
+
+STATIC Size_t
S_do_trans_count_utf8(pTHX_ SV * const sv)
{
- dVAR;
const U8 *s;
const U8 *start = NULL;
const U8 *send;
- I32 matches = 0;
+ Size_t matches = 0;
STRLEN len;
SV* const rv =
#ifdef USE_ITHREADS
s = (const U8*)SvPV_nomg_const(sv, len);
if (!SvUTF8(sv)) {
- const U8 *t = s;
- const U8 * const e = s + len;
- while (t < e) {
- const U8 ch = *t++;
- hibit = !NATIVE_BYTE_IS_INVARIANT(ch);
- if (hibit) {
- start = s = bytes_to_utf8(s, &len);
- break;
- }
+ hibit = ! is_utf8_invariant_string(s, len);
+ if (hibit) {
+ start = s = bytes_to_utf8(s, &len);
}
}
send = s + len;
return matches;
}
-STATIC I32
+
+/* Helper function for do_trans().
+ * Handles utf8 cases(*) involving the /c, /d, /s flags,
+ * and where search and replacement charlists aren't identical.
+ * (*) i.e. where the search or replacement charlists are utf8. sv may
+ * or may not be utf8.
+ */
+
+STATIC Size_t
S_do_trans_complex_utf8(pTHX_ SV * const sv)
{
- dVAR;
U8 *start, *send;
U8 *d;
- I32 matches = 0;
- const I32 squash = PL_op->op_private & OPpTRANS_SQUASH;
- const I32 del = PL_op->op_private & OPpTRANS_DELETE;
- const I32 grows = PL_op->op_private & OPpTRANS_GROWS;
+ Size_t matches = 0;
+ const bool squash = cBOOL(PL_op->op_private & OPpTRANS_SQUASH);
+ const bool del = cBOOL(PL_op->op_private & OPpTRANS_DELETE);
+ const bool grows = cBOOL(PL_op->op_private & OPpTRANS_GROWS);
SV* const rv =
#ifdef USE_ITHREADS
PAD_SVl(cPADOP->op_padix);
PERL_ARGS_ASSERT_DO_TRANS_COMPLEX_UTF8;
if (!SvUTF8(sv)) {
- const U8 *t = s;
- const U8 * const e = s + len;
- while (t < e) {
- const U8 ch = *t++;
- hibit = !NATIVE_BYTE_IS_INVARIANT(ch);
- if (hibit) {
- s = bytes_to_utf8(s, &len);
- break;
- }
+ hibit = ! is_utf8_invariant_string(s, len);
+ if (hibit) {
+ s = bytes_to_utf8(s, &len);
}
}
send = s + len;
UV puv = 0xfeedface;
while (s < send) {
UV uv = swash_fetch(rv, s, TRUE);
-
+
if (d > dend) {
const STRLEN clen = d - dstart;
const STRLEN nlen = dend - dstart + len + UTF8_MAXBYTES;
return matches;
}
-I32
+
+/* Execute a tr//. sv is the value to be translated, while PL_op
+ * should be an OP_TRANS or OP_TRANSR op, whose op_pv field contains a
+ * translation table or whose op_sv field contains a swash.
+ * Returns a count of number of characters translated
+ */
+
+Size_t
Perl_do_trans(pTHX_ SV *sv)
{
- dVAR;
STRLEN len;
- const I32 hasutf = (PL_op->op_private &
- (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF));
+ const U8 flags = PL_op->op_private;
+ const U8 hasutf = flags & (OPpTRANS_FROM_UTF | OPpTRANS_TO_UTF);
PERL_ARGS_ASSERT_DO_TRANS;
- if (SvREADONLY(sv) && !(PL_op->op_private & OPpTRANS_IDENTICAL)) {
- Perl_croak_no_modify();
+ if (SvREADONLY(sv) && !(flags & OPpTRANS_IDENTICAL)) {
+ Perl_croak_no_modify();
}
(void)SvPV_const(sv, len);
if (!len)
return 0;
- if (!(PL_op->op_private & OPpTRANS_IDENTICAL)) {
+ if (!(flags & OPpTRANS_IDENTICAL)) {
if (!SvPOKp(sv) || SvTHINKFIRST(sv))
(void)SvPV_force_nomg(sv, len);
(void)SvPOK_only_UTF8(sv);
}
- DEBUG_t( Perl_deb(aTHX_ "2.TBL\n"));
-
- switch (PL_op->op_private & ~hasutf & (
- OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF|OPpTRANS_IDENTICAL|
- OPpTRANS_SQUASH|OPpTRANS_DELETE|OPpTRANS_COMPLEMENT)) {
- case 0:
- if (hasutf)
- return do_trans_simple_utf8(sv);
- else
- return do_trans_simple(sv);
-
- case OPpTRANS_IDENTICAL:
- case OPpTRANS_IDENTICAL|OPpTRANS_COMPLEMENT:
- if (hasutf)
- return do_trans_count_utf8(sv);
- else
- return do_trans_count(sv);
-
- default:
- if (hasutf)
- return do_trans_complex_utf8(sv);
- else
- return do_trans_complex(sv);
+ /* If we use only OPpTRANS_IDENTICAL to bypass the READONLY check,
+ * we must also rely on it to choose the readonly strategy.
+ */
+ if (flags & OPpTRANS_IDENTICAL) {
+ return hasutf ? do_trans_count_utf8(sv) : do_trans_count(sv, (OPtrans_map*)cPVOP->op_pv);
+ } else if (flags & (OPpTRANS_SQUASH|OPpTRANS_DELETE|OPpTRANS_COMPLEMENT)) {
+ return hasutf ? do_trans_complex_utf8(sv) : do_trans_complex(sv, (OPtrans_map*)cPVOP->op_pv);
+ } else {
+ return hasutf ? do_trans_simple_utf8(sv) : do_trans_simple(sv, (OPtrans_map*)cPVOP->op_pv);
}
}
void
Perl_do_join(pTHX_ SV *sv, SV *delim, SV **mark, SV **sp)
{
- dVAR;
SV ** const oldmark = mark;
I32 items = sp - mark;
STRLEN len;
STRLEN delimlen;
+ const char * const delims = SvPV_const(delim, delimlen);
PERL_ARGS_ASSERT_DO_JOIN;
- (void) SvPV_const(delim, delimlen); /* stringify and get the delimlen */
- /* SvCUR assumes it's SvPOK() and woe betide you if it's not. */
-
mark++;
len = (items > 0 ? (delimlen * (items - 1) ) : 0);
SvUPGRADE(sv, SVt_PV);
++mark;
}
- sv_setpvs(sv, "");
+ SvPVCLEAR(sv);
/* sv_setpv retains old UTF8ness [perl #24846] */
SvUTF8_off(sv);
}
if (delimlen) {
+ const U32 delimflag = DO_UTF8(delim) ? SV_CATUTF8 : SV_CATBYTES;
for (; items > 0; items--,mark++) {
- sv_catsv_nomg(sv,delim);
- sv_catsv(sv,*mark);
+ STRLEN len;
+ const char *s;
+ sv_catpvn_flags(sv,delims,delimlen,delimflag);
+ s = SvPV_const(*mark,len);
+ sv_catpvn_flags(sv,s,len,
+ DO_UTF8(*mark) ? SV_CATUTF8 : SV_CATBYTES);
}
}
else {
for (; items > 0; items--,mark++)
- sv_catsv(sv,*mark);
+ {
+ STRLEN len;
+ const char *s = SvPV_const(*mark,len);
+ sv_catpvn_flags(sv,s,len,
+ DO_UTF8(*mark) ? SV_CATUTF8 : SV_CATBYTES);
+ }
}
SvSETMAGIC(sv);
}
void
-Perl_do_sprintf(pTHX_ SV *sv, I32 len, SV **sarg)
+Perl_do_sprintf(pTHX_ SV *sv, SSize_t len, SV **sarg)
{
- dVAR;
STRLEN patlen;
const char * const pat = SvPV_const(*sarg, patlen);
bool do_taint = FALSE;
PERL_ARGS_ASSERT_DO_SPRINTF;
+ assert(len >= 1);
if (SvTAINTED(*sarg))
TAINT_PROPER(
SvUTF8_off(sv);
if (DO_UTF8(*sarg))
SvUTF8_on(sv);
- sv_vsetpvfn(sv, pat, patlen, NULL, sarg + 1, len - 1, &do_taint);
+ sv_vsetpvfn(sv, pat, patlen, NULL, sarg + 1, (Size_t)(len - 1), &do_taint);
SvSETMAGIC(sv);
if (do_taint)
SvTAINTED_on(sv);
/* currently converts input to bytes if possible, but doesn't sweat failure */
UV
-Perl_do_vecget(pTHX_ SV *sv, SSize_t offset, int size)
+Perl_do_vecget(pTHX_ SV *sv, STRLEN offset, int size)
{
- dVAR;
- STRLEN srclen, len, uoffset, bitoffs = 0;
- const unsigned char *s = (const unsigned char *) SvPV_flags_const(sv, srclen,
- SV_GMAGIC | ((PL_op->op_flags & OPf_MOD || LVRET)
- ? SV_UNDEF_RETURNS_NULL : 0));
+ STRLEN srclen, len, avail, uoffset, bitoffs = 0;
+ const I32 svpv_flags = ((PL_op->op_flags & OPf_MOD || LVRET)
+ ? SV_UNDEF_RETURNS_NULL : 0);
+ unsigned char *s = (unsigned char *)
+ SvPV_flags(sv, srclen, (svpv_flags|SV_GMAGIC));
UV retnum = 0;
if (!s) {
- s = (const unsigned char *)"";
+ s = (unsigned char *)"";
}
-
+
PERL_ARGS_ASSERT_DO_VECGET;
- if (offset < 0)
- return 0;
if (size < 1 || (size & (size-1))) /* size < 1 or not a power of two */
Perl_croak(aTHX_ "Illegal number of bits in vec");
- if (SvUTF8(sv))
- (void) Perl_sv_utf8_downgrade(aTHX_ sv, TRUE);
+ if (SvUTF8(sv)) {
+ if (Perl_sv_utf8_downgrade_flags(aTHX_ sv, TRUE, 0)) {
+ /* PVX may have changed */
+ s = (unsigned char *) SvPV_flags(sv, srclen, svpv_flags);
+ }
+ else {
+ Perl_croak(aTHX_ "Use of strings with code points over 0xFF as arguments to vec is forbidden");
+ }
+ }
if (size < 8) {
bitoffs = ((offset%8)*size)%8;
uoffset = offset/(8/size);
}
- else if (size > 8)
- uoffset = offset*(size/8);
+ else if (size > 8) {
+ int n = size/8;
+ if (offset > Size_t_MAX / n - 1) /* would overflow */
+ return 0;
+ uoffset = offset*n;
+ }
else
uoffset = offset;
- len = uoffset + (bitoffs + size + 7)/8; /* required number of bytes */
- if (len > srclen) {
+ if (uoffset >= srclen)
+ return 0;
+
+ len = (bitoffs + size + 7)/8; /* required number of bytes */
+ avail = srclen - uoffset; /* available number of bytes */
+
+ /* Does the byte range overlap the end of the string? If so,
+ * handle specially. */
+ if (avail < len) {
if (size <= 8)
retnum = 0;
else {
if (size == 16) {
- if (uoffset >= srclen)
- retnum = 0;
- else
- retnum = (UV) s[uoffset] << 8;
+ assert(avail == 1);
+ retnum = (UV) s[uoffset] << 8;
}
else if (size == 32) {
- if (uoffset >= srclen)
- retnum = 0;
- else if (uoffset + 1 >= srclen)
+ assert(avail >= 1 && avail <= 3);
+ if (avail == 1)
retnum =
((UV) s[uoffset ] << 24);
- else if (uoffset + 2 >= srclen)
+ else if (avail == 2)
retnum =
((UV) s[uoffset ] << 24) +
((UV) s[uoffset + 1] << 16);
else if (size == 64) {
Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
"Bit vector size > 32 non-portable");
- if (uoffset >= srclen)
- retnum = 0;
- else if (uoffset + 1 >= srclen)
+ assert(avail >= 1 && avail <= 7);
+ if (avail == 1)
retnum =
(UV) s[uoffset ] << 56;
- else if (uoffset + 2 >= srclen)
+ else if (avail == 2)
retnum =
((UV) s[uoffset ] << 56) +
((UV) s[uoffset + 1] << 48);
- else if (uoffset + 3 >= srclen)
+ else if (avail == 3)
retnum =
((UV) s[uoffset ] << 56) +
((UV) s[uoffset + 1] << 48) +
((UV) s[uoffset + 2] << 40);
- else if (uoffset + 4 >= srclen)
+ else if (avail == 4)
retnum =
((UV) s[uoffset ] << 56) +
((UV) s[uoffset + 1] << 48) +
((UV) s[uoffset + 2] << 40) +
((UV) s[uoffset + 3] << 32);
- else if (uoffset + 5 >= srclen)
+ else if (avail == 5)
retnum =
((UV) s[uoffset ] << 56) +
((UV) s[uoffset + 1] << 48) +
((UV) s[uoffset + 2] << 40) +
((UV) s[uoffset + 3] << 32) +
- ( s[uoffset + 4] << 24);
- else if (uoffset + 6 >= srclen)
+ ((UV) s[uoffset + 4] << 24);
+ else if (avail == 6)
retnum =
((UV) s[uoffset ] << 56) +
((UV) s[uoffset + 1] << 48) +
((UV) s[uoffset + 3] << 32) +
((UV) s[uoffset + 4] << 24) +
((UV) s[uoffset + 5] << 16) +
- ( s[uoffset + 6] << 8);
+ ((UV) s[uoffset + 6] << 8);
}
#endif
}
void
Perl_do_vecset(pTHX_ SV *sv)
{
- dVAR;
- SSize_t offset, bitoffs = 0;
+ STRLEN offset, bitoffs = 0;
int size;
unsigned char *s;
UV lval;
STRLEN targlen;
STRLEN len;
SV * const targ = LvTARG(sv);
+ char errflags = LvFLAGS(sv);
PERL_ARGS_ASSERT_DO_VECSET;
+ /* some out-of-range errors have been deferred if/until the LV is
+ * actually written to: f(vec($s,-1,8)) is not always fatal */
+ if (errflags) {
+ assert(!(errflags & ~(LVf_NEG_OFF|LVf_OUT_OF_RANGE)));
+ if (errflags & LVf_NEG_OFF)
+ Perl_croak_nocontext("Negative offset to vec in lvalue context");
+ Perl_croak_nocontext("Out of memory!");
+ }
+
if (!targ)
return;
s = (unsigned char*)SvPV_force_flags(targ, targlen,
SV_GMAGIC | SV_UNDEF_RETURNS_NULL);
if (SvUTF8(targ)) {
/* This is handled by the SvPOK_only below...
- if (!Perl_sv_utf8_downgrade(aTHX_ targ, TRUE))
+ if (!Perl_sv_utf8_downgrade_flags(aTHX_ targ, TRUE, 0))
SvUTF8_off(targ);
*/
- (void) Perl_sv_utf8_downgrade(aTHX_ targ, TRUE);
+ (void) Perl_sv_utf8_downgrade_flags(aTHX_ targ, TRUE, 0);
}
(void)SvPOK_only(targ);
lval = SvUV(sv);
offset = LvTARGOFF(sv);
- if (offset < 0)
- Perl_croak(aTHX_ "Negative offset to vec in lvalue context");
size = LvTARGLEN(sv);
+
if (size < 1 || (size & (size-1))) /* size < 1 or not a power of two */
Perl_croak(aTHX_ "Illegal number of bits in vec");
bitoffs = ((offset%8)*size)%8;
offset /= 8/size;
}
- else if (size > 8)
- offset *= size/8;
-
- len = offset + (bitoffs + size + 7)/8; /* required number of bytes */
- if (len > targlen) {
- s = (unsigned char*)SvGROW(targ, len + 1);
- (void)memzero((char *)(s + targlen), len - targlen + 1);
- SvCUR_set(targ, len);
+ else if (size > 8) {
+ int n = size/8;
+ if (offset > Size_t_MAX / n - 1) /* would overflow */
+ Perl_croak_nocontext("Out of memory!");
+ offset *= n;
+ }
+
+ len = (bitoffs + size + 7)/8; /* required number of bytes */
+ if (targlen < offset || targlen - offset < len) {
+ STRLEN newlen = offset > Size_t_MAX - len - 1 ? /* avoid overflow */
+ Size_t_MAX : offset + len + 1;
+ s = (unsigned char*)SvGROW(targ, newlen);
+ (void)memzero((char *)(s + targlen), newlen - targlen);
+ SvCUR_set(targ, newlen - 1);
}
if (size < 8) {
void
Perl_do_vop(pTHX_ I32 optype, SV *sv, SV *left, SV *right)
{
- dVAR;
-#ifdef LIBERAL
long *dl;
long *ll;
long *rl;
-#endif
char *dc;
STRLEN leftlen;
STRLEN rightlen;
const char *lc;
const char *rc;
- STRLEN len;
+ STRLEN len = 0;
STRLEN lensave;
const char *lsave;
const char *rsave;
- bool left_utf;
- bool right_utf;
STRLEN needlen = 0;
+ bool result_needs_to_be_utf8 = FALSE;
+ bool left_utf8 = FALSE;
+ bool right_utf8 = FALSE;
+ U8 * left_non_downgraded = NULL;
+ U8 * right_non_downgraded = NULL;
+ Size_t left_non_downgraded_len = 0;
+ Size_t right_non_downgraded_len = 0;
+ char * non_downgraded = NULL;
+ Size_t non_downgraded_len = 0;
PERL_ARGS_ASSERT_DO_VOP;
if (sv != left || (optype != OP_BIT_AND && !SvOK(sv)))
- sv_setpvs(sv, ""); /* avoid undef warning on |= and ^= */
+ SvPVCLEAR(sv); /* avoid undef warning on |= and ^= */
if (sv == left) {
- lsave = lc = SvPV_force_nomg(left, leftlen);
+ lc = SvPV_force_nomg(left, leftlen);
}
else {
- lsave = lc = SvPV_nomg_const(left, leftlen);
+ lc = SvPV_nomg_const(left, leftlen);
SvPV_force_nomg_nolen(sv);
}
- rsave = rc = SvPV_nomg_const(right, rightlen);
+ rc = SvPV_nomg_const(right, rightlen);
- /* This need to come after SvPV to ensure that string overloading has
+ /* This needs to come after SvPV to ensure that string overloading has
fired off. */
- left_utf = DO_UTF8(left);
- right_utf = DO_UTF8(right);
-
- if (left_utf && !right_utf) {
- /* Avoid triggering overloading again by using temporaries.
- Maybe there should be a variant of sv_utf8_upgrade that takes pvn
- */
- right = newSVpvn_flags(rsave, rightlen, SVs_TEMP);
- sv_utf8_upgrade(right);
- rsave = rc = SvPV_nomg_const(right, rightlen);
- right_utf = TRUE;
+ /* Create downgraded temporaries of any UTF-8 encoded operands */
+ if (DO_UTF8(left)) {
+ const U8 * save_lc = (U8 *) lc;
+
+ left_utf8 = TRUE;
+ result_needs_to_be_utf8 = TRUE;
+
+ left_non_downgraded_len = leftlen;
+ lc = (char *) bytes_from_utf8_loc((const U8 *) lc, &leftlen,
+ &left_utf8,
+ (const U8 **) &left_non_downgraded);
+ /* Calculate the number of trailing unconvertible bytes. This quantity
+ * is the original length minus the length of the converted portion. */
+ left_non_downgraded_len -= left_non_downgraded - save_lc;
+ SAVEFREEPV(lc);
}
- else if (!left_utf && right_utf) {
- left = newSVpvn_flags(lsave, leftlen, SVs_TEMP);
- sv_utf8_upgrade(left);
- lsave = lc = SvPV_nomg_const(left, leftlen);
- left_utf = TRUE;
+ if (DO_UTF8(right)) {
+ const U8 * save_rc = (U8 *) rc;
+
+ right_utf8 = TRUE;
+ result_needs_to_be_utf8 = TRUE;
+
+ right_non_downgraded_len = rightlen;
+ rc = (char *) bytes_from_utf8_loc((const U8 *) rc, &rightlen,
+ &right_utf8,
+ (const U8 **) &right_non_downgraded);
+ right_non_downgraded_len -= right_non_downgraded - save_rc;
+ SAVEFREEPV(rc);
+ }
+
+ /* We set 'len' to the length that the operation actually operates on. The
+ * dangling part of the longer operand doesn't actually participate in the
+ * operation. What happens is that we pretend that the shorter operand has
+ * been extended to the right by enough imaginary zeros to match the length
+ * of the longer one. But we know in advance the result of the operation
+ * on zeros without having to do it. In the case of '&', the result is
+ * zero, and the dangling portion is simply discarded. For '|' and '^', the
+ * result is the same as the other operand, so the dangling part is just
+ * appended to the final result, unchanged. As of perl-5.32, we no longer
+ * accept above-FF code points in the dangling portion.
+ */
+ if (left_utf8 || right_utf8) {
+ Perl_croak(aTHX_ FATAL_ABOVE_FF_MSG, PL_op_desc[optype]);
+ }
+ else { /* Neither is UTF-8 */
+ len = MIN(leftlen, rightlen);
}
- len = leftlen < rightlen ? leftlen : rightlen;
lensave = len;
+ lsave = lc;
+ rsave = rc;
+
SvCUR_set(sv, len);
(void)SvPOK_only(sv);
- if ((left_utf || right_utf) && (sv == left || sv == right)) {
- needlen = optype == OP_BIT_AND ? len : leftlen + rightlen;
- Newxz(dc, needlen + 1, char);
- }
- else if (SvOK(sv) || SvTYPE(sv) > SVt_PVMG) {
+ if (SvOK(sv) || SvTYPE(sv) > SVt_PVMG) {
dc = SvPV_force_nomg_nolen(sv);
if (SvLEN(sv) < len + 1) {
dc = SvGROW(sv, len + 1);
(void)memzero(dc + SvCUR(sv), len - SvCUR(sv) + 1);
}
- if (optype != OP_BIT_AND && (left_utf || right_utf))
- dc = SvGROW(sv, leftlen + rightlen + 1);
}
else {
needlen = optype == OP_BIT_AND
sv_usepvn_flags(sv, dc, needlen, SV_HAS_TRAILING_NUL);
dc = SvPVX(sv); /* sv_usepvn() calls Renew() */
}
- if (left_utf || right_utf) {
- UV duc, luc, ruc;
- char *dcorig = dc;
- char *dcsave = NULL;
- STRLEN lulen = leftlen;
- STRLEN rulen = rightlen;
- STRLEN ulen;
- switch (optype) {
- case OP_BIT_AND:
- while (lulen && rulen) {
- luc = utf8n_to_uvchr((U8*)lc, lulen, &ulen, UTF8_ALLOW_ANYUV);
- lc += ulen;
- lulen -= ulen;
- ruc = utf8n_to_uvchr((U8*)rc, rulen, &ulen, UTF8_ALLOW_ANYUV);
- rc += ulen;
- rulen -= ulen;
- duc = luc & ruc;
- dc = (char*)uvchr_to_utf8((U8*)dc, duc);
- }
- if (sv == left || sv == right)
- (void)sv_usepvn(sv, dcorig, needlen);
- SvCUR_set(sv, dc - dcorig);
- break;
- case OP_BIT_XOR:
- while (lulen && rulen) {
- luc = utf8n_to_uvchr((U8*)lc, lulen, &ulen, UTF8_ALLOW_ANYUV);
- lc += ulen;
- lulen -= ulen;
- ruc = utf8n_to_uvchr((U8*)rc, rulen, &ulen, UTF8_ALLOW_ANYUV);
- rc += ulen;
- rulen -= ulen;
- duc = luc ^ ruc;
- dc = (char*)uvchr_to_utf8((U8*)dc, duc);
- }
- goto mop_up_utf;
- case OP_BIT_OR:
- while (lulen && rulen) {
- luc = utf8n_to_uvchr((U8*)lc, lulen, &ulen, UTF8_ALLOW_ANYUV);
- lc += ulen;
- lulen -= ulen;
- ruc = utf8n_to_uvchr((U8*)rc, rulen, &ulen, UTF8_ALLOW_ANYUV);
- rc += ulen;
- rulen -= ulen;
- duc = luc | ruc;
- dc = (char*)uvchr_to_utf8((U8*)dc, duc);
- }
- mop_up_utf:
- if (rulen)
- dcsave = savepvn(rc, rulen);
- else if (lulen)
- dcsave = savepvn(lc, lulen);
- if (sv == left || sv == right)
- (void)sv_usepvn(sv, dcorig, needlen); /* uses Renew(); defaults to nomg */
- SvCUR_set(sv, dc - dcorig);
- if (rulen)
- sv_catpvn_nomg(sv, dcsave, rulen);
- else if (lulen)
- sv_catpvn_nomg(sv, dcsave, lulen);
- else
- *SvEND(sv) = '\0';
- Safefree(dcsave);
- break;
- default:
- if (sv == left || sv == right)
- Safefree(dcorig);
- Perl_croak(aTHX_ "panic: do_vop called for op %u (%s)",
- (unsigned)optype, PL_op_name[optype]);
- }
- SvUTF8_on(sv);
- goto finish;
- }
- else
-#ifdef LIBERAL
if (len >= sizeof(long)*4 &&
- !((unsigned long)dc % sizeof(long)) &&
- !((unsigned long)lc % sizeof(long)) &&
- !((unsigned long)rc % sizeof(long))) /* It's almost always aligned... */
+ !(PTR2nat(dc) % sizeof(long)) &&
+ !(PTR2nat(lc) % sizeof(long)) &&
+ !(PTR2nat(rc) % sizeof(long))) /* It's almost always aligned... */
{
const STRLEN remainder = len % (sizeof(long)*4);
len /= (sizeof(long)*4);
len = remainder;
}
-#endif
- {
- switch (optype) {
- case OP_BIT_AND:
- while (len--)
- *dc++ = *lc++ & *rc++;
- *dc = '\0';
- break;
- case OP_BIT_XOR:
- while (len--)
- *dc++ = *lc++ ^ *rc++;
- goto mop_up;
- case OP_BIT_OR:
- while (len--)
- *dc++ = *lc++ | *rc++;
- mop_up:
- len = lensave;
- if (rightlen > len)
- sv_catpvn_nomg(sv, rsave + len, rightlen - len);
- else if (leftlen > (STRLEN)len)
- sv_catpvn_nomg(sv, lsave + len, leftlen - len);
- else
- *SvEND(sv) = '\0';
- break;
- }
+
+ switch (optype) {
+ case OP_BIT_AND:
+ while (len--)
+ *dc++ = *lc++ & *rc++;
+ *dc = '\0';
+ break;
+ case OP_BIT_XOR:
+ while (len--)
+ *dc++ = *lc++ ^ *rc++;
+ goto mop_up;
+ case OP_BIT_OR:
+ while (len--)
+ *dc++ = *lc++ | *rc++;
+ mop_up:
+ len = lensave;
+ if (rightlen > len) {
+ if (dc == rc)
+ SvCUR_set(sv, rightlen);
+ else
+ sv_catpvn_nomg(sv, rsave + len, rightlen - len);
+ }
+ else if (leftlen > len) {
+ if (dc == lc)
+ SvCUR_set(sv, leftlen);
+ else
+ sv_catpvn_nomg(sv, lsave + len, leftlen - len);
+ }
+ *SvEND(sv) = '\0';
+
+ /* If there is trailing stuff that couldn't be converted from UTF-8, it
+ * is appended as-is for the ^ and | operators. This preserves
+ * backwards compatibility */
+ if (right_non_downgraded) {
+ non_downgraded = (char *) right_non_downgraded;
+ non_downgraded_len = right_non_downgraded_len;
+ }
+ else if (left_non_downgraded) {
+ non_downgraded = (char *) left_non_downgraded;
+ non_downgraded_len = left_non_downgraded_len;
+ }
+
+ break;
}
-finish:
+
+ if (result_needs_to_be_utf8) {
+ sv_utf8_upgrade_nomg(sv);
+
+ /* Append any trailing UTF-8 as-is. */
+ if (non_downgraded) {
+ sv_catpvn_nomg(sv, non_downgraded, non_downgraded_len);
+ }
+ }
+
SvTAINT(sv);
}
+
+/* Perl_do_kv() may be:
+ * * called directly as the pp function for pp_keys() and pp_values();
+ * * It may also be called directly when the op is OP_AVHVSWITCH, to
+ * implement CORE::keys(), CORE::values().
+ *
+ * In all cases it expects an HV on the stack and returns a list of keys,
+ * values, or key-value pairs, depending on PL_op.
+ */
+
OP *
Perl_do_kv(pTHX)
{
- dVAR;
dSP;
HV * const keys = MUTABLE_HV(POPs);
- HE *entry;
- const I32 gimme = GIMME_V;
- const I32 dokv = (PL_op->op_type == OP_RV2HV || PL_op->op_type == OP_PADHV);
- /* op_type is OP_RKEYS/OP_RVALUES if pp_rkeys delegated to here */
- const I32 dokeys = dokv || (PL_op->op_type == OP_KEYS || PL_op->op_type == OP_RKEYS);
- const I32 dovalues = dokv || (PL_op->op_type == OP_VALUES || PL_op->op_type == OP_RVALUES);
+ const U8 gimme = GIMME_V;
+
+ const I32 dokeys = (PL_op->op_type == OP_KEYS)
+ || ( PL_op->op_type == OP_AVHVSWITCH
+ && (PL_op->op_private & OPpAVHVSWITCH_MASK)
+ + OP_EACH == OP_KEYS);
+
+ const I32 dovalues = (PL_op->op_type == OP_VALUES)
+ || ( PL_op->op_type == OP_AVHVSWITCH
+ && (PL_op->op_private & OPpAVHVSWITCH_MASK)
+ + OP_EACH == OP_VALUES);
+
+ assert( PL_op->op_type == OP_KEYS
+ || PL_op->op_type == OP_VALUES
+ || PL_op->op_type == OP_AVHVSWITCH);
+
+ assert(!( PL_op->op_type == OP_VALUES
+ && (PL_op->op_private & OPpMAYBE_LVSUB)));
(void)hv_iterinit(keys); /* always reset iterator regardless */
IV i;
dTARGET;
+ /* note that in 'scalar(keys %h)' the OP_KEYS is usually
+ * optimised away and the action is performed directly by the
+ * padhv or rv2hv op. We now only get here via OP_AVHVSWITCH
+ * and \&CORE::keys
+ */
if (! SvTIED_mg((const SV *)keys, PERL_MAGIC_tied) ) {
i = HvUSEDKEYS(keys);
}
RETURN;
}
- EXTEND(SP, HvUSEDKEYS(keys) * (dokeys + dovalues));
-
- PUTBACK; /* hv_iternext and hv_iterval might clobber stack_sp */
- while ((entry = hv_iternext(keys))) {
- SPAGAIN;
- if (dokeys) {
- SV* const sv = hv_iterkeysv(entry);
- XPUSHs(sv); /* won't clobber stack_sp */
- }
- if (dovalues) {
- SV *tmpstr;
- PUTBACK;
- tmpstr = hv_iterval(keys,entry);
- DEBUG_H(Perl_sv_setpvf(aTHX_ tmpstr, "%lu%%%d=%lu",
- (unsigned long)HeHASH(entry),
- (int)HvMAX(keys)+1,
- (unsigned long)(HeHASH(entry) & HvMAX(keys))));
- SPAGAIN;
- XPUSHs(tmpstr);
- }
- PUTBACK;
+ if (UNLIKELY(PL_op->op_private & OPpMAYBE_LVSUB)) {
+ const I32 flags = is_lvalue_sub();
+ if (flags && !(flags & OPpENTERSUB_INARGS))
+ /* diag_listed_as: Can't modify %s in %s */
+ Perl_croak(aTHX_ "Can't modify keys in list assignment");
}
+
+ PUTBACK;
+ hv_pushkv(keys, (dokeys | (dovalues << 1)));
return NORMAL;
}
/*
- * Local variables:
- * c-indentation-style: bsd
- * c-basic-offset: 4
- * indent-tabs-mode: nil
- * End:
- *
* ex: set ts=8 sts=4 sw=4 et:
*/
https://perl5.git.perl.org / perl5.git / blobdiff