* in such scope than if not. However, various libc functions called by Perl
* are affected by the LC_NUMERIC category, so there are macros in perl.h that
* are used to toggle between the current locale and the C locale depending on
- * the desired behavior of those functions at the moment.
+ * the desired behavior of those functions at the moment. And, LC_MESSAGES is
+ * switched to the C locale for outputting the message unless within the scope
+ * of 'use locale'.
*/
#include "EXTERN.h"
#define PERL_IN_LOCALE_C
+#include "perl_langinfo.h"
#include "perl.h"
-#ifdef I_LANGINFO
-# include <langinfo.h>
-#endif
-
#include "reentr.h"
+/* If the environment says to, we can output debugging information during
+ * initialization. This is done before option parsing, and before any thread
+ * creation, so can be a file-level static */
+#ifdef DEBUGGING
+# ifdef PERL_GLOBAL_STRUCT
+ /* no global syms allowed */
+# define debug_initialization 0
+# define DEBUG_INITIALIZATION_set(v)
+# else
+static bool debug_initialization = FALSE;
+# define DEBUG_INITIALIZATION_set(v) (debug_initialization = v)
+# endif
+#endif
+
#ifdef USE_LOCALE
/*
#endif
-void
-Perl_set_numeric_radix(pTHX)
+STATIC void
+S_set_numeric_radix(pTHX)
{
#ifdef USE_LOCALE_NUMERIC
# ifdef HAS_LOCALECONV
sv_setpv(PL_numeric_radix_sv, lc->decimal_point);
else
PL_numeric_radix_sv = newSVpv(lc->decimal_point, 0);
- if (! is_invariant_string((U8 *) lc->decimal_point, 0)
+ if (! is_utf8_invariant_string((U8 *) lc->decimal_point, 0)
&& is_utf8_string((U8 *) lc->decimal_point, 0)
&& _is_cur_LC_category_utf8(LC_NUMERIC))
{
else
PL_numeric_radix_sv = NULL;
- DEBUG_L(PerlIO_printf(Perl_debug_log, "Locale radix is %s\n",
+#ifdef DEBUGGING
+ if (DEBUG_L_TEST || debug_initialization) {
+ PerlIO_printf(Perl_debug_log, "Locale radix is '%s', ?UTF-8=%d\n",
+ (PL_numeric_radix_sv)
+ ? SvPVX(PL_numeric_radix_sv)
+ : "NULL",
(PL_numeric_radix_sv)
- ? lc->decimal_point
- : "NULL"));
+ ? cBOOL(SvUTF8(PL_numeric_radix_sv))
+ : 0);
+ }
+#endif
# endif /* HAS_LOCALECONV */
#endif /* USE_LOCALE_NUMERIC */
}
save_newnum = stdize_locale(savepv(newnum));
+
+ PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum);
+ PL_numeric_local = TRUE;
+
if (! PL_numeric_name || strNE(PL_numeric_name, save_newnum)) {
Safefree(PL_numeric_name);
PL_numeric_name = save_newnum;
}
-
- PL_numeric_standard = isNAME_C_OR_POSIX(save_newnum);
- PL_numeric_local = TRUE;
+ else {
+ Safefree(save_newnum);
+ }
/* Keep LC_NUMERIC in the C locale. This is for XS modules, so they don't
* have to worry about the radix being a non-dot. (Core operations that
PL_numeric_standard = TRUE;
PL_numeric_local = isNAME_C_OR_POSIX(PL_numeric_name);
set_numeric_radix();
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Underlying LC_NUMERIC locale now is C\n"));
+#ifdef DEBUGGING
+ if (DEBUG_L_TEST || debug_initialization) {
+ PerlIO_printf(Perl_debug_log,
+ "Underlying LC_NUMERIC locale now is C\n");
+ }
+#endif
#endif /* USE_LOCALE_NUMERIC */
}
PL_numeric_standard = isNAME_C_OR_POSIX(PL_numeric_name);
PL_numeric_local = TRUE;
set_numeric_radix();
- DEBUG_L(PerlIO_printf(Perl_debug_log,
+#ifdef DEBUGGING
+ if (DEBUG_L_TEST || debug_initialization) {
+ PerlIO_printf(Perl_debug_log,
"Underlying LC_NUMERIC locale now is %s\n",
- PL_numeric_name));
+ PL_numeric_name);
+ }
+#endif
#endif /* USE_LOCALE_NUMERIC */
}
/*
* Set up for a new ctype locale.
*/
-void
-Perl_new_ctype(pTHX_ const char *newctype)
+STATIC void
+S_new_ctype(pTHX_ const char *newctype)
{
#ifdef USE_LOCALE_CTYPE
* NUL */
char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ];
- bool check_for_problems = ckWARN_d(WARN_LOCALE); /* No warnings means
- no check */
+ /* Don't check for problems if we are suppressing the warnings */
+ bool check_for_problems = ckWARN_d(WARN_LOCALE)
+ || UNLIKELY(DEBUG_L_TEST);
bool multi_byte_locale = FALSE; /* Assume is a single-byte locale
to start */
unsigned int bad_count = 0; /* Count of bad characters */
#ifdef MB_CUR_MAX
/* We only handle single-byte locales (outside of UTF-8 ones; so if
- * this locale requires than one byte, there are going to be
+ * this locale requires more than one byte, there are going to be
* problems. */
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "%s:%d: check_for_problems=%d, MB_CUR_MAX=%d\n",
+ __FILE__, __LINE__, check_for_problems, (int) MB_CUR_MAX));
+
if (check_for_problems && MB_CUR_MAX > 1
/* Some platforms return MB_CUR_MAX > 1 for even the "C"
? bad_chars_list
: ""
);
- /* If we are actually in the scope of the locale, output the
- * message now. Otherwise we save it to be output at the first
- * operation using this locale, if that actually happens. Most
- * programs don't use locales, so they are immune to bad ones */
- if (IN_LC(LC_CTYPE)) {
+ /* If we are actually in the scope of the locale or are debugging,
+ * output the message now. If not in that scope, we save the
+ * message to be output at the first operation using this locale,
+ * if that actually happens. Most programs don't use locales, so
+ * they are immune to bad ones. */
+ if (IN_LC(LC_CTYPE) || UNLIKELY(DEBUG_L_TEST)) {
/* We have to save 'newctype' because the setlocale() just
* below may destroy it. The next setlocale() further down
/* The '0' below suppresses a bogus gcc compiler warning */
Perl_warner(aTHX_ packWARN(WARN_LOCALE), SvPVX(PL_warn_locale), 0);
+
setlocale(LC_CTYPE, badlocale);
Safefree(badlocale);
- SvREFCNT_dec_NN(PL_warn_locale);
- PL_warn_locale = NULL;
+
+ if (IN_LC(LC_CTYPE)) {
+ SvREFCNT_dec_NN(PL_warn_locale);
+ PL_warn_locale = NULL;
+ }
}
}
}
}
-void
-Perl_new_collate(pTHX_ const char *newcoll)
+STATIC void
+S_new_collate(pTHX_ const char *newcoll)
{
#ifdef USE_LOCALE_COLLATE
* Any code changing the locale (outside this file) should use
* POSIX::setlocale, which calls this function. Therefore this function
* should be called directly only from this file and from
- * POSIX::setlocale() */
+ * POSIX::setlocale().
+ *
+ * The design of locale collation is that every locale change is given an
+ * index 'PL_collation_ix'. The first time a string particpates in an
+ * operation that requires collation while locale collation is active, it
+ * is given PERL_MAGIC_collxfrm magic (via sv_collxfrm_flags()). That
+ * magic includes the collation index, and the transformation of the string
+ * by strxfrm(), q.v. That transformation is used when doing comparisons,
+ * instead of the string itself. If a string changes, the magic is
+ * cleared. The next time the locale changes, the index is incremented,
+ * and so we know during a comparison that the transformation is not
+ * necessarily still valid, and so is recomputed. Note that if the locale
+ * changes enough times, the index could wrap (a U32), and it is possible
+ * that a transformation would improperly be considered valid, leading to
+ * an unlikely bug */
if (! newcoll) {
if (PL_collation_name) {
PL_collation_name = NULL;
}
PL_collation_standard = TRUE;
+ is_standard_collation:
PL_collxfrm_base = 0;
PL_collxfrm_mult = 2;
+ PL_in_utf8_COLLATE_locale = FALSE;
+ PL_strxfrm_NUL_replacement = '\0';
+ PL_strxfrm_max_cp = 0;
return;
}
+ /* If this is not the same locale as currently, set the new one up */
if (! PL_collation_name || strNE(PL_collation_name, newcoll)) {
++PL_collation_ix;
Safefree(PL_collation_name);
PL_collation_name = stdize_locale(savepv(newcoll));
PL_collation_standard = isNAME_C_OR_POSIX(newcoll);
+ if (PL_collation_standard) {
+ goto is_standard_collation;
+ }
+
+ PL_in_utf8_COLLATE_locale = _is_cur_LC_category_utf8(LC_COLLATE);
+ PL_strxfrm_NUL_replacement = '\0';
+ PL_strxfrm_max_cp = 0;
+
+ /* A locale collation definition includes primary, secondary, tertiary,
+ * etc. weights for each character. To sort, the primary weights are
+ * used, and only if they compare equal, then the secondary weights are
+ * used, and only if they compare equal, then the tertiary, etc.
+ *
+ * strxfrm() works by taking the input string, say ABC, and creating an
+ * output transformed string consisting of first the primary weights,
+ * A¹B¹C¹ followed by the secondary ones, A²B²C²; and then the
+ * tertiary, etc, yielding A¹B¹C¹ A²B²C² A³B³C³ .... Some characters
+ * may not have weights at every level. In our example, let's say B
+ * doesn't have a tertiary weight, and A doesn't have a secondary
+ * weight. The constructed string is then going to be
+ * A¹B¹C¹ B²C² A³C³ ....
+ * This has the desired effect that strcmp() will look at the secondary
+ * or tertiary weights only if the strings compare equal at all higher
+ * priority weights. The spaces shown here, like in
+ * "A¹B¹C¹ A²B²C² "
+ * are not just for readability. In the general case, these must
+ * actually be bytes, which we will call here 'separator weights'; and
+ * they must be smaller than any other weight value, but since these
+ * are C strings, only the terminating one can be a NUL (some
+ * implementations may include a non-NUL separator weight just before
+ * the NUL). Implementations tend to reserve 01 for the separator
+ * weights. They are needed so that a shorter string's secondary
+ * weights won't be misconstrued as primary weights of a longer string,
+ * etc. By making them smaller than any other weight, the shorter
+ * string will sort first. (Actually, if all secondary weights are
+ * smaller than all primary ones, there is no need for a separator
+ * weight between those two levels, etc.)
+ *
+ * The length of the transformed string is roughly a linear function of
+ * the input string. It's not exactly linear because some characters
+ * don't have weights at all levels. When we call strxfrm() we have to
+ * allocate some memory to hold the transformed string. The
+ * calculations below try to find coefficients 'm' and 'b' for this
+ * locale so that m*x + b equals how much space we need, given the size
+ * of the input string in 'x'. If we calculate too small, we increase
+ * the size as needed, and call strxfrm() again, but it is better to
+ * get it right the first time to avoid wasted expensive string
+ * transformations. */
{
- /* 2: at most so many chars ('a', 'b'). */
- /* 50: surely no system expands a char more. */
-#define XFRMBUFSIZE (2 * 50)
- char xbuf[XFRMBUFSIZE];
- const Size_t fa = strxfrm(xbuf, "a", XFRMBUFSIZE);
- const Size_t fb = strxfrm(xbuf, "ab", XFRMBUFSIZE);
- const SSize_t mult = fb - fa;
- if (mult < 1 && !(fa == 0 && fb == 0))
- Perl_croak(aTHX_ "panic: strxfrm() gets absurd - a => %"UVuf", ab => %"UVuf,
- (UV) fa, (UV) fb);
- PL_collxfrm_base = (fa > (Size_t)mult) ? (fa - mult) : 0;
- PL_collxfrm_mult = mult;
+ /* We use the string below to find how long the tranformation of it
+ * is. Almost all locales are supersets of ASCII, or at least the
+ * ASCII letters. We use all of them, half upper half lower,
+ * because if we used fewer, we might hit just the ones that are
+ * outliers in a particular locale. Most of the strings being
+ * collated will contain a preponderance of letters, and even if
+ * they are above-ASCII, they are likely to have the same number of
+ * weight levels as the ASCII ones. It turns out that digits tend
+ * to have fewer levels, and some punctuation has more, but those
+ * are relatively sparse in text, and khw believes this gives a
+ * reasonable result, but it could be changed if experience so
+ * dictates. */
+ const char longer[] = "ABCDEFGHIJKLMnopqrstuvwxyz";
+ char * x_longer; /* Transformed 'longer' */
+ Size_t x_len_longer; /* Length of 'x_longer' */
+
+ char * x_shorter; /* We also transform a substring of 'longer' */
+ Size_t x_len_shorter;
+
+ /* _mem_collxfrm() is used get the transformation (though here we
+ * are interested only in its length). It is used because it has
+ * the intelligence to handle all cases, but to work, it needs some
+ * values of 'm' and 'b' to get it started. For the purposes of
+ * this calculation we use a very conservative estimate of 'm' and
+ * 'b'. This assumes a weight can be multiple bytes, enough to
+ * hold any UV on the platform, and there are 5 levels, 4 weight
+ * bytes, and a trailing NUL. */
+ PL_collxfrm_base = 5;
+ PL_collxfrm_mult = 5 * sizeof(UV);
+
+ /* Find out how long the transformation really is */
+ x_longer = _mem_collxfrm(longer,
+ sizeof(longer) - 1,
+ &x_len_longer,
+
+ /* We avoid converting to UTF-8 in the
+ * called function by telling it the
+ * string is in UTF-8 if the locale is a
+ * UTF-8 one. Since the string passed
+ * here is invariant under UTF-8, we can
+ * claim it's UTF-8 even though it isn't.
+ * */
+ PL_in_utf8_COLLATE_locale);
+ Safefree(x_longer);
+
+ /* Find out how long the transformation of a substring of 'longer'
+ * is. Together the lengths of these transformations are
+ * sufficient to calculate 'm' and 'b'. The substring is all of
+ * 'longer' except the first character. This minimizes the chances
+ * of being swayed by outliers */
+ x_shorter = _mem_collxfrm(longer + 1,
+ sizeof(longer) - 2,
+ &x_len_shorter,
+ PL_in_utf8_COLLATE_locale);
+ Safefree(x_shorter);
+
+ /* If the results are nonsensical for this simple test, the whole
+ * locale definition is suspect. Mark it so that locale collation
+ * is not active at all for it. XXX Should we warn? */
+ if ( x_len_shorter == 0
+ || x_len_longer == 0
+ || x_len_shorter >= x_len_longer)
+ {
+ PL_collxfrm_mult = 0;
+ PL_collxfrm_base = 0;
+ }
+ else {
+ SSize_t base; /* Temporary */
+
+ /* We have both: m * strlen(longer) + b = x_len_longer
+ * m * strlen(shorter) + b = x_len_shorter;
+ * subtracting yields:
+ * m * (strlen(longer) - strlen(shorter))
+ * = x_len_longer - x_len_shorter
+ * But we have set things up so that 'shorter' is 1 byte smaller
+ * than 'longer'. Hence:
+ * m = x_len_longer - x_len_shorter
+ *
+ * But if something went wrong, make sure the multiplier is at
+ * least 1.
+ */
+ if (x_len_longer > x_len_shorter) {
+ PL_collxfrm_mult = (STRLEN) x_len_longer - x_len_shorter;
+ }
+ else {
+ PL_collxfrm_mult = 1;
+ }
+
+ /* mx + b = len
+ * so: b = len - mx
+ * but in case something has gone wrong, make sure it is
+ * non-negative */
+ base = x_len_longer - PL_collxfrm_mult * (sizeof(longer) - 1);
+ if (base < 0) {
+ base = 0;
+ }
+
+ /* Add 1 for the trailing NUL */
+ PL_collxfrm_base = base + 1;
+ }
+
+#ifdef DEBUGGING
+ if (DEBUG_L_TEST || debug_initialization) {
+ PerlIO_printf(Perl_debug_log,
+ "%s:%d: ?UTF-8 locale=%d; x_len_shorter=%zu, "
+ "x_len_longer=%zu,"
+ " collate multipler=%zu, collate base=%zu\n",
+ __FILE__, __LINE__,
+ PL_in_utf8_COLLATE_locale,
+ x_len_shorter, x_len_longer,
+ PL_collxfrm_mult, PL_collxfrm_base);
+ }
+#endif
}
}
#endif /* USE_LOCALE_COLLATE */
}
-#ifdef WIN32
+#ifndef WIN32 /* No wrapper except on Windows */
-char *
-Perl_my_setlocale(pTHX_ int category, const char* locale)
+#define my_setlocale(a,b) setlocale(a,b)
+
+#else /* WIN32 */
+
+STATIC char *
+S_my_setlocale(pTHX_ int category, const char* locale)
{
/* This, for Windows, emulates POSIX setlocale() behavior. There is no
* difference unless the input locale is "", which means on Windows to get
}
result = setlocale(category, locale);
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n", __FILE__, __LINE__,
+ setlocale_debug_string(category, locale, result)));
if (! override_LC_ALL) {
return result;
result = PerlEnv_getenv("LC_TIME");
if (result && strNE(result, "")) {
setlocale(LC_TIME, result);
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
+ __FILE__, __LINE__,
+ setlocale_debug_string(LC_TIME, result, "not captured")));
}
# endif
# ifdef USE_LOCALE_CTYPE
result = PerlEnv_getenv("LC_CTYPE");
if (result && strNE(result, "")) {
setlocale(LC_CTYPE, result);
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
+ __FILE__, __LINE__,
+ setlocale_debug_string(LC_CTYPE, result, "not captured")));
}
# endif
# ifdef USE_LOCALE_COLLATE
result = PerlEnv_getenv("LC_COLLATE");
if (result && strNE(result, "")) {
setlocale(LC_COLLATE, result);
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
+ __FILE__, __LINE__,
+ setlocale_debug_string(LC_COLLATE, result, "not captured")));
}
# endif
# ifdef USE_LOCALE_MONETARY
result = PerlEnv_getenv("LC_MONETARY");
if (result && strNE(result, "")) {
setlocale(LC_MONETARY, result);
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
+ __FILE__, __LINE__,
+ setlocale_debug_string(LC_MONETARY, result, "not captured")));
}
# endif
# ifdef USE_LOCALE_NUMERIC
result = PerlEnv_getenv("LC_NUMERIC");
if (result && strNE(result, "")) {
setlocale(LC_NUMERIC, result);
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
+ __FILE__, __LINE__,
+ setlocale_debug_string(LC_NUMERIC, result, "not captured")));
}
# endif
# ifdef USE_LOCALE_MESSAGES
result = PerlEnv_getenv("LC_MESSAGES");
if (result && strNE(result, "")) {
setlocale(LC_MESSAGES, result);
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
+ __FILE__, __LINE__,
+ setlocale_debug_string(LC_MESSAGES, result, "not captured")));
}
# endif
- return setlocale(LC_ALL, NULL);
+ result = setlocale(LC_ALL, NULL);
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "%s:%d: %s\n",
+ __FILE__, __LINE__,
+ setlocale_debug_string(LC_ALL, NULL, result)));
+
+ return result;
+}
+
+#endif
+
+char *
+Perl_setlocale(int category, const char * locale)
+{
+ /* This wraps POSIX::setlocale() */
+
+ char * retval;
+ dTHX;
+
+
+#ifdef USE_LOCALE_NUMERIC
+
+ /* A NULL locale means only query what the current one is. We
+ * have the LC_NUMERIC name saved, because we are normally switched
+ * into the C locale for it. Switch back so an LC_ALL query will yield
+ * the correct results; all other categories don't require special
+ * handling */
+ if (locale == NULL) {
+ if (category == LC_NUMERIC) {
+ return savepv(PL_numeric_name);
+ }
+
+# ifdef LC_ALL
+
+ else if (category == LC_ALL) {
+ SET_NUMERIC_UNDERLYING();
+ }
+
+# endif
+
+ }
+
+#endif
+
+ retval = my_setlocale(category, locale);
+
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "%s:%d: %s\n", __FILE__, __LINE__,
+ setlocale_debug_string(category, locale, retval)));
+ if (! retval) {
+ /* Should never happen that a query would return an error, but be
+ * sure and reset to C locale */
+ if (locale == 0) {
+ SET_NUMERIC_STANDARD();
+ }
+ return NULL;
+ }
+
+ /* Save retval since subsequent setlocale() calls may overwrite it. */
+ retval = savepv(retval);
+
+ /* If locale == NULL, we are just querying the state, but may have switched
+ * to NUMERIC_UNDERLYING. Switch back before returning. */
+ if (locale == NULL) {
+ SET_NUMERIC_STANDARD();
+ return retval;
+ }
+ else { /* Now that have switched locales, we have to update our records to
+ correspond */
+
+#ifdef USE_LOCALE_CTYPE
+
+ if ( category == LC_CTYPE
+
+# ifdef LC_ALL
+
+ || category == LC_ALL
+
+# endif
+
+ )
+ {
+ char *newctype;
+
+# ifdef LC_ALL
+
+ if (category == LC_ALL) {
+ newctype = setlocale(LC_CTYPE, NULL);
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "%s:%d: %s\n", __FILE__, __LINE__,
+ setlocale_debug_string(LC_CTYPE, NULL, newctype)));
+ }
+ else
+
+# endif
+
+ newctype = retval;
+ new_ctype(newctype);
+ }
+
+#endif /* USE_LOCALE_CTYPE */
+
+#ifdef USE_LOCALE_COLLATE
+
+ if ( category == LC_COLLATE
+
+# ifdef LC_ALL
+
+ || category == LC_ALL
+
+# endif
+
+ )
+ {
+ char *newcoll;
+
+# ifdef LC_ALL
+
+ if (category == LC_ALL) {
+ newcoll = setlocale(LC_COLLATE, NULL);
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "%s:%d: %s\n", __FILE__, __LINE__,
+ setlocale_debug_string(LC_COLLATE, NULL, newcoll)));
+ }
+ else
+
+# endif
+
+ newcoll = retval;
+ new_collate(newcoll);
+ }
+
+#endif /* USE_LOCALE_COLLATE */
+
+#ifdef USE_LOCALE_NUMERIC
+
+ if ( category == LC_NUMERIC
+
+# ifdef LC_ALL
+
+ || category == LC_ALL
+
+# endif
+
+ )
+ {
+ char *newnum;
+
+# ifdef LC_ALL
+
+ if (category == LC_ALL) {
+ newnum = setlocale(LC_NUMERIC, NULL);
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "%s:%d: %s\n", __FILE__, __LINE__,
+ setlocale_debug_string(LC_NUMERIC, NULL, newnum)));
+ }
+ else
+
+# endif
+
+ newnum = retval;
+ new_numeric(newnum);
+ }
+
+#endif /* USE_LOCALE_NUMERIC */
+
+ }
+
+ return retval;
+
+
+}
+
+PERL_STATIC_INLINE const char *
+S_save_to_buffer(const char * string, char **buf, Size_t *buf_size, const Size_t offset)
+{
+ /* Copy the NUL-terminated 'string' to 'buf' + 'offset'. 'buf' has size 'buf_size',
+ * growing it if necessary */
+
+ const Size_t string_size = strlen(string) + offset + 1;
+
+ PERL_ARGS_ASSERT_SAVE_TO_BUFFER;
+
+ if (*buf_size == 0) {
+ Newx(*buf, string_size, char);
+ *buf_size = string_size;
+ }
+ else if (string_size > *buf_size) {
+ Renew(*buf, string_size, char);
+ *buf_size = string_size;
+ }
+
+ Copy(string, *buf + offset, string_size - offset, char);
+ return *buf;
+}
+
+/*
+
+=head1 Locale-related functions and macros
+
+=for apidoc Perl_langinfo
+
+This is an (almostª) drop-in replacement for the system C<L<nl_langinfo(3)>>,
+taking the same C<item> parameter values, and returning the same information.
+But it is more thread-safe than regular C<nl_langinfo()>, and hides the quirks
+of Perl's locale handling from your code, and can be used on systems that lack
+a native C<nl_langinfo>.
+
+Expanding on these:
+
+=over
+
+=item *
+
+It delivers the correct results for the C<RADIXCHAR> and C<THOUSESEP> items,
+without you having to write extra code. The reason for the extra code would be
+because these are from the C<LC_NUMERIC> locale category, which is normally
+kept set to the C locale by Perl, no matter what the underlying locale is
+supposed to be, and so to get the expected results, you have to temporarily
+toggle into the underlying locale, and later toggle back. (You could use
+plain C<nl_langinfo> and C<L</STORE_LC_NUMERIC_FORCE_TO_UNDERLYING>> for this
+but then you wouldn't get the other advantages of C<Perl_langinfo()>; not
+keeping C<LC_NUMERIC> in the C locale would break a lot of CPAN, which is
+expecting the radix (decimal point) character to be a dot.)
+
+=item *
+
+Depending on C<item>, it works on systems that don't have C<nl_langinfo>, hence
+makes your code more portable. Of the fifty-some possible items specified by
+the POSIX 2008 standard,
+L<http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/langinfo.h.html>,
+only two are completely unimplemented. It uses various techniques to recover
+the other items, including calling C<L<localeconv(3)>>, and C<L<strftime(3)>>,
+both of which are specified in C89, so should be always be available. Later
+C<strftime()> versions have additional capabilities; C<""> is returned for
+those not available on your system.
+
+The details for those items which may differ from what this emulation returns
+and what a native C<nl_langinfo()> would return are:
+
+=over
+
+=item C<CODESET>
+
+=item C<ERA>
+
+Unimplemented, so returns C<"">.
+
+=item C<YESEXPR>
+
+=item C<NOEXPR>
+
+Only the values for English are returned. Earlier POSIX standards also
+specified C<YESSTR> and C<NOSTR>, but these have been removed from POSIX 2008,
+and aren't supported by C<Perl_langinfo>.
+
+=item C<D_FMT>
+
+Always evaluates to C<%x>, the locale's appropriate date representation.
+
+=item C<T_FMT>
+
+Always evaluates to C<%X>, the locale's appropriate time representation.
+
+=item C<D_T_FMT>
+
+Always evaluates to C<%c>, the locale's appropriate date and time
+representation.
+
+=item C<CRNCYSTR>
+
+The return may be incorrect for those rare locales where the currency symbol
+replaces the radix character.
+Send email to L<mailto:perlbug@perl.org> if you have examples of it needing
+to work differently.
+
+=item C<ALT_DIGITS>
+
+Currently this gives the same results as Linux does.
+Send email to L<mailto:perlbug@perl.org> if you have examples of it needing
+to work differently.
+
+=item C<ERA_D_FMT>
+
+=item C<ERA_T_FMT>
+
+=item C<ERA_D_T_FMT>
+
+=item C<T_FMT_AMPM>
+
+These are derived by using C<strftime()>, and not all versions of that function
+know about them. C<""> is returned for these on such systems.
+
+=back
+
+When using C<Perl_langinfo> on systems that don't have a native
+C<nl_langinfo()>, you must
+
+ #include "perl_langinfo.h"
+
+before the C<perl.h> C<#include>. You can replace your C<langinfo.h>
+C<#include> with this one. (Doing it this way keeps out the symbols that plain
+C<langinfo.h> imports into the namespace for code that doesn't need it.)
+
+You also should not use the bare C<langinfo.h> item names, but should preface
+them with C<PERL_>, so use C<PERL_RADIXCHAR> instead of plain C<RADIXCHAR>.
+The C<PERL_I<foo>> versions will also work for this function on systems that do
+have a native C<nl_langinfo>.
+
+=item *
+
+It is thread-friendly, returning its result in a buffer that won't be
+overwritten by another thread, so you don't have to code for that possibility.
+The buffer can be overwritten by the next call to C<nl_langinfo> or
+C<Perl_langinfo> in the same thread.
+
+=item *
+
+ªIt returns S<C<const char *>>, whereas plain C<nl_langinfo()> returns S<C<char
+*>>, but you are (only by documentation) forbidden to write into the buffer.
+By declaring this C<const>, the compiler enforces this restriction. The extra
+C<const> is why this isn't an unequivocal drop-in replacement for
+C<nl_langinfo>.
+
+=back
+
+The original impetus for C<Perl_langinfo()> was so that code that needs to
+find out the current currency symbol, floating point radix character, or digit
+grouping separator can use, on all systems, the simpler and more
+thread-friendly C<nl_langinfo> API instead of C<L<localeconv(3)>> which is a
+pain to make thread-friendly. For other fields returned by C<localeconv>, it
+is better to use the methods given in L<perlcall> to call
+L<C<POSIX::localeconv()>|POSIX/localeconv>, which is thread-friendly.
+
+=cut
+
+*/
+
+const char *
+#ifdef HAS_NL_LANGINFO
+Perl_langinfo(const nl_item item)
+#else
+Perl_langinfo(const int item)
+#endif
+{
+ bool toggle = TRUE;
+
+#if defined(HAS_NL_LANGINFO)
+# if ! defined(USE_ITHREADS)
+
+ /* Single-thread, and nl_langinfo() is available. Call it, switching to
+ * underlying LC_NUMERIC for those items dependent on it */
+
+ const char * retval;
+
+ if (toggle) {
+ if (item == PERL_RADIXCHAR || item == PERL_THOUSEP) {
+ setlocale(LC_NUMERIC, PL_numeric_name);
+ }
+ else {
+ toggle = FALSE;
+ }
+ }
+
+ retval = nl_langinfo(item);
+
+ if (toggle) {
+ setlocale(LC_NUMERIC, "C");
+ }
+
+ return retval;
+
+
+# else
+
+ /* Multi-threaded, with native nl_langinfo(). Use it, copying result to
+ * per-thread buffer, and toggling LC_NUMERIC if necessary, all within a
+ * crtical section */
+
+ dTHX;
+
+ LOCALE_LOCK;
+
+ if (toggle) {
+ if (item == PERL_RADIXCHAR || item == PERL_THOUSEP) {
+ setlocale(LC_NUMERIC, PL_numeric_name);
+ }
+ else {
+ toggle = FALSE;
+ }
+ }
+
+ save_to_buffer(nl_langinfo(item), &PL_langinfo_buf, &PL_langinfo_bufsize, 0);
+
+ if (toggle) {
+ setlocale(LC_NUMERIC, "C");
+ }
+
+ LOCALE_UNLOCK;
+
+ return PL_langinfo_buf;
+
+# endif
+#else /* Below, emulate nl_langinfo as best we can */
+
+ dTHX;
+
+# ifdef HAS_LOCALECONV
+
+ const struct lconv* lc;
+
+# endif
+# ifdef HAS_STRFTIME
+
+ struct tm tm;
+ bool return_format = FALSE; /* Return the %format, not the value */
+ const char * format;
+
+# endif
+
+ /* We copy the results to a per-thread buffer, even if not multi-threaded.
+ * This is in part to simplify this code, and partly because we need a
+ * buffer anyway for strftime(), and partly because a call of localeconv()
+ * could otherwise wipe out the buffer, and the programmer would not be
+ * expecting this, as this is a nl_langinfo() substitute after all, so s/he
+ * might be thinking their localeconv() is safe until another localeconv()
+ * call. */
+
+ switch (item) {
+ Size_t len;
+ const char * retval;
+
+ /* These 2 are unimplemented */
+ case PERL_CODESET:
+ case PERL_ERA: /* For use with strftime() %E modifier */
+
+ default:
+ return "";
+
+ /* We use only an English set, since we don't know any more */
+ case PERL_YESEXPR: return "^[+1yY]";
+ case PERL_NOEXPR: return "^[-0nN]";
+
+# ifdef HAS_LOCALECONV
+
+ case PERL_CRNCYSTR:
+
+ LOCALE_LOCK;
+
+ lc = localeconv();
+ if (! lc || ! lc->currency_symbol || strEQ("", lc->currency_symbol))
+ {
+ LOCALE_UNLOCK;
+ return "";
+ }
+
+ /* Leave the first spot empty to be filled in below */
+ save_to_buffer(lc->currency_symbol, &PL_langinfo_buf,
+ &PL_langinfo_bufsize, 1);
+ if (lc->mon_decimal_point && strEQ(lc->mon_decimal_point, ""))
+ { /* khw couldn't figure out how the localedef specifications
+ would show that the $ should replace the radix; this is
+ just a guess as to how it might work.*/
+ *PL_langinfo_buf = '.';
+ }
+ else if (lc->p_cs_precedes) {
+ *PL_langinfo_buf = '-';
+ }
+ else {
+ *PL_langinfo_buf = '+';
+ }
+
+ LOCALE_UNLOCK;
+ break;
+
+ case PERL_RADIXCHAR:
+ case PERL_THOUSEP:
+
+ LOCALE_LOCK;
+
+ if (toggle) {
+ setlocale(LC_NUMERIC, PL_numeric_name);
+ }
+
+ lc = localeconv();
+ if (! lc) {
+ retval = "";
+ }
+ else switch (item) {
+ case PERL_RADIXCHAR:
+ if (! lc->decimal_point) {
+ retval = "";
+ }
+ else {
+ retval = lc->decimal_point;
+ }
+ break;
+
+ case PERL_THOUSEP:
+ if (! lc->thousands_sep || strEQ("", lc->thousands_sep)) {
+ retval = "";
+ }
+ else {
+ retval = lc->thousands_sep;
+ }
+ break;
+
+ default:
+ LOCALE_UNLOCK;
+ Perl_croak(aTHX_ "panic: %s: %d: switch case: %d problem",
+ __FILE__, __LINE__, item);
+ }
+
+ save_to_buffer(retval, &PL_langinfo_buf, &PL_langinfo_bufsize, 0);
+
+ if (toggle) {
+ setlocale(LC_NUMERIC, "C");
+ }
+
+ LOCALE_UNLOCK;
+
+ break;
+
+# endif
+# ifdef HAS_STRFTIME
+
+ /* These are defined by C89, so we assume that strftime supports them,
+ * and so are returned unconditionally; they may not be what the locale
+ * actually says, but should give good enough results for someone using
+ * them as formats (as opposed to trying to parse them to figure out
+ * what the locale says). The other format ones are actually tested to
+ * verify they work on the platform */
+ case PERL_D_FMT: return "%x";
+ case PERL_T_FMT: return "%X";
+ case PERL_D_T_FMT: return "%c";
+
+ /* These formats are only available in later strfmtime's */
+ case PERL_ERA_D_FMT: case PERL_ERA_T_FMT: case PERL_ERA_D_T_FMT:
+ case PERL_T_FMT_AMPM:
+
+ /* The rest can be gotten from most versions of strftime(). */
+ case PERL_ABDAY_1: case PERL_ABDAY_2: case PERL_ABDAY_3:
+ case PERL_ABDAY_4: case PERL_ABDAY_5: case PERL_ABDAY_6:
+ case PERL_ABDAY_7:
+ case PERL_ALT_DIGITS:
+ case PERL_AM_STR: case PERL_PM_STR:
+ case PERL_ABMON_1: case PERL_ABMON_2: case PERL_ABMON_3:
+ case PERL_ABMON_4: case PERL_ABMON_5: case PERL_ABMON_6:
+ case PERL_ABMON_7: case PERL_ABMON_8: case PERL_ABMON_9:
+ case PERL_ABMON_10: case PERL_ABMON_11: case PERL_ABMON_12:
+ case PERL_DAY_1: case PERL_DAY_2: case PERL_DAY_3: case PERL_DAY_4:
+ case PERL_DAY_5: case PERL_DAY_6: case PERL_DAY_7:
+ case PERL_MON_1: case PERL_MON_2: case PERL_MON_3: case PERL_MON_4:
+ case PERL_MON_5: case PERL_MON_6: case PERL_MON_7: case PERL_MON_8:
+ case PERL_MON_9: case PERL_MON_10: case PERL_MON_11: case PERL_MON_12:
+
+ LOCALE_LOCK;
+
+ init_tm(&tm); /* Precaution against core dumps */
+ tm.tm_sec = 30;
+ tm.tm_min = 30;
+ tm.tm_hour = 6;
+ tm.tm_year = 2017 - 1900;
+ tm.tm_wday = 0;
+ tm.tm_mon = 0;
+ switch (item) {
+ default:
+ LOCALE_UNLOCK;
+ Perl_croak(aTHX_ "panic: %s: %d: switch case: %d problem",
+ __FILE__, __LINE__, item);
+ NOT_REACHED; /* NOTREACHED */
+
+ case PERL_PM_STR: tm.tm_hour = 18;
+ case PERL_AM_STR:
+ format = "%p";
+ break;
+
+ case PERL_ABDAY_7: tm.tm_wday++;
+ case PERL_ABDAY_6: tm.tm_wday++;
+ case PERL_ABDAY_5: tm.tm_wday++;
+ case PERL_ABDAY_4: tm.tm_wday++;
+ case PERL_ABDAY_3: tm.tm_wday++;
+ case PERL_ABDAY_2: tm.tm_wday++;
+ case PERL_ABDAY_1:
+ format = "%a";
+ break;
+
+ case PERL_DAY_7: tm.tm_wday++;
+ case PERL_DAY_6: tm.tm_wday++;
+ case PERL_DAY_5: tm.tm_wday++;
+ case PERL_DAY_4: tm.tm_wday++;
+ case PERL_DAY_3: tm.tm_wday++;
+ case PERL_DAY_2: tm.tm_wday++;
+ case PERL_DAY_1:
+ format = "%A";
+ break;
+
+ case PERL_ABMON_12: tm.tm_mon++;
+ case PERL_ABMON_11: tm.tm_mon++;
+ case PERL_ABMON_10: tm.tm_mon++;
+ case PERL_ABMON_9: tm.tm_mon++;
+ case PERL_ABMON_8: tm.tm_mon++;
+ case PERL_ABMON_7: tm.tm_mon++;
+ case PERL_ABMON_6: tm.tm_mon++;
+ case PERL_ABMON_5: tm.tm_mon++;
+ case PERL_ABMON_4: tm.tm_mon++;
+ case PERL_ABMON_3: tm.tm_mon++;
+ case PERL_ABMON_2: tm.tm_mon++;
+ case PERL_ABMON_1:
+ format = "%b";
+ break;
+
+ case PERL_MON_12: tm.tm_mon++;
+ case PERL_MON_11: tm.tm_mon++;
+ case PERL_MON_10: tm.tm_mon++;
+ case PERL_MON_9: tm.tm_mon++;
+ case PERL_MON_8: tm.tm_mon++;
+ case PERL_MON_7: tm.tm_mon++;
+ case PERL_MON_6: tm.tm_mon++;
+ case PERL_MON_5: tm.tm_mon++;
+ case PERL_MON_4: tm.tm_mon++;
+ case PERL_MON_3: tm.tm_mon++;
+ case PERL_MON_2: tm.tm_mon++;
+ case PERL_MON_1:
+ format = "%B";
+ break;
+
+ case PERL_T_FMT_AMPM:
+ format = "%r";
+ return_format = TRUE;
+ break;
+
+ case PERL_ERA_D_FMT:
+ format = "%Ex";
+ return_format = TRUE;
+ break;
+
+ case PERL_ERA_T_FMT:
+ format = "%EX";
+ return_format = TRUE;
+ break;
-}
+ case PERL_ERA_D_T_FMT:
+ format = "%Ec";
+ return_format = TRUE;
+ break;
+
+ case PERL_ALT_DIGITS:
+ tm.tm_wday = 0;
+ format = "%Ow"; /* Find the alternate digit for 0 */
+ break;
+ }
+
+ /* We can't use my_strftime() because it doesn't look at tm_wday */
+ while (0 == strftime(PL_langinfo_buf, PL_langinfo_bufsize,
+ format, &tm))
+ {
+ /* A zero return means one of:
+ * a) there wasn't enough space in PL_langinfo_buf
+ * b) the format, like a plain %p, returns empty
+ * c) it was an illegal format, though some implementations of
+ * strftime will just return the illegal format as a plain
+ * character sequence.
+ *
+ * To quickly test for case 'b)', try again but precede the
+ * format with a plain character. If that result is still
+ * empty, the problem is either 'a)' or 'c)' */
+
+ Size_t format_size = strlen(format) + 1;
+ Size_t mod_size = format_size + 1;
+ char * mod_format;
+ char * temp_result;
+
+ Newx(mod_format, mod_size, char);
+ Newx(temp_result, PL_langinfo_bufsize, char);
+ *mod_format = '\a';
+ my_strlcpy(mod_format + 1, format, mod_size);
+ len = strftime(temp_result,
+ PL_langinfo_bufsize,
+ mod_format, &tm);
+ Safefree(mod_format);
+ Safefree(temp_result);
+
+ /* If 'len' is non-zero, it means that we had a case like %p
+ * which means the current locale doesn't use a.m. or p.m., and
+ * that is valid */
+ if (len == 0) {
+
+ /* Here, still didn't work. If we get well beyond a
+ * reasonable size, bail out to prevent an infinite loop. */
+
+ if (PL_langinfo_bufsize > 100 * format_size) {
+ *PL_langinfo_buf = '\0';
+ }
+ else { /* Double the buffer size to retry; Add 1 in case
+ original was 0, so we aren't stuck at 0. */
+ PL_langinfo_bufsize *= 2;
+ PL_langinfo_bufsize++;
+ Renew(PL_langinfo_buf, PL_langinfo_bufsize, char);
+ continue;
+ }
+ }
+
+ break;
+ }
+
+ /* Here, we got a result.
+ *
+ * If the item is 'ALT_DIGITS', PL_langinfo_buf contains the
+ * alternate format for wday 0. If the value is the same as the
+ * normal 0, there isn't an alternate, so clear the buffer. */
+ if ( item == PERL_ALT_DIGITS
+ && strEQ(PL_langinfo_buf, "0"))
+ {
+ *PL_langinfo_buf = '\0';
+ }
+
+ /* ALT_DIGITS is problematic. Experiments on it showed that
+ * strftime() did not always work properly when going from alt-9 to
+ * alt-10. Only a few locales have this item defined, and in all
+ * of them on Linux that khw was able to find, nl_langinfo() merely
+ * returned the alt-0 character, possibly doubled. Most Unicode
+ * digits are in blocks of 10 consecutive code points, so that is
+ * sufficient information for those scripts, as we can infer alt-1,
+ * alt-2, .... But for a Japanese locale, a CJK ideographic 0 is
+ * returned, and the CJK digits are not in code point order, so you
+ * can't really infer anything. The localedef for this locale did
+ * specify the succeeding digits, so that strftime() works properly
+ * on them, without needing to infer anything. But the
+ * nl_langinfo() return did not give sufficient information for the
+ * caller to understand what's going on. So until there is
+ * evidence that it should work differently, this returns the alt-0
+ * string for ALT_DIGITS.
+ *
+ * wday was chosen because its range is all a single digit. Things
+ * like tm_sec have two digits as the minimum: '00' */
+
+ LOCALE_UNLOCK;
+
+ /* If to return the format, not the value, overwrite the buffer
+ * with it. But some strftime()s will keep the original format if
+ * illegal, so change those to "" */
+ if (return_format) {
+ if (strEQ(PL_langinfo_buf, format)) {
+ *PL_langinfo_buf = '\0';
+ }
+ else {
+ save_to_buffer(format, &PL_langinfo_buf,
+ &PL_langinfo_bufsize, 0);
+ }
+ }
+
+ break;
+
+# endif
+
+ }
+
+ return PL_langinfo_buf;
#endif
+}
/*
* Initialize locale awareness.
const char * const setlocale_init = (PerlEnv_getenv("PERL_SKIP_LOCALE_INIT"))
? NULL
: "";
-#ifdef DEBUGGING
- const bool debug = (PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT"))
- ? TRUE
- : FALSE;
-# define DEBUG_LOCALE_INIT(category, locale, result) \
- STMT_START { \
- if (debug) { \
- PerlIO_printf(Perl_debug_log, \
- "%s:%d: %s\n", \
- __FILE__, __LINE__, \
- _setlocale_debug_string(category, \
- locale, \
- result)); \
- } \
- } STMT_END
-#else
-# define DEBUG_LOCALE_INIT(a,b,c)
-#endif
const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
unsigned int trial_locales_count;
const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
const char *system_default_locale = NULL;
#endif
+#ifdef DEBUGGING
+ DEBUG_INITIALIZATION_set(cBOOL(PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT")));
+# define DEBUG_LOCALE_INIT(category, locale, result) \
+ STMT_START { \
+ if (debug_initialization) { \
+ PerlIO_printf(Perl_debug_log, \
+ "%s:%d: %s\n", \
+ __FILE__, __LINE__, \
+ setlocale_debug_string(category, \
+ locale, \
+ result)); \
+ } \
+ } STMT_END
+#else
+# define DEBUG_LOCALE_INIT(a,b,c)
+#endif
+
#ifndef LOCALE_ENVIRON_REQUIRED
PERL_UNUSED_VAR(done);
PERL_UNUSED_VAR(locale_param);
: NULL;
sl_result = my_setlocale(LC_MESSAGES, locale_param);
DEBUG_LOCALE_INIT(LC_MESSAGES, locale_param, sl_result);
- if (! sl_result)
+ if (! sl_result) {
setlocale_failure = TRUE;
}
# endif /* USE_LOCALE_MESSAGES */
{
char **e;
for (e = environ; *e; e++) {
- if (strnEQ(*e, "LC_", 3)
- && strnNE(*e, "LC_ALL=", 7)
+ if (strEQs(*e, "LC_")
+ && strNEs(*e, "LC_ALL=")
&& (p = strchr(*e, '=')))
PerlIO_printf(Perl_error_log, "\t%.*s = \"%s\",\n",
(int)(p - *e), *e, p + 1);
PERL_UNUSED_ARG(printwarn);
#endif /* USE_LOCALE */
+#ifdef DEBUGGING
+ /* So won't continue to output stuff */
+ DEBUG_INITIALIZATION_set(FALSE);
+#endif
+
return ok;
}
-
#ifdef USE_LOCALE_COLLATE
-/*
- * mem_collxfrm() is a bit like strxfrm() but with two important
- * differences. First, it handles embedded NULs. Second, it allocates
- * a bit more memory than needed for the transformed data itself.
- * The real transformed data begins at offset sizeof(collationix).
- * Please see sv_collxfrm() to see how this is used.
- */
-
char *
-Perl_mem_collxfrm(pTHX_ const char *s, STRLEN len, STRLEN *xlen)
+Perl__mem_collxfrm(pTHX_ const char *input_string,
+ STRLEN len, /* Length of 'input_string' */
+ STRLEN *xlen, /* Set to length of returned string
+ (not including the collation index
+ prefix) */
+ bool utf8 /* Is the input in UTF-8? */
+ )
{
- char *xbuf;
- STRLEN xAlloc, xin, xout; /* xalloc is a reserved word in VC */
- PERL_ARGS_ASSERT_MEM_COLLXFRM;
+ /* _mem_collxfrm() is a bit like strxfrm() but with two important
+ * differences. First, it handles embedded NULs. Second, it allocates a bit
+ * more memory than needed for the transformed data itself. The real
+ * transformed data begins at offset COLLXFRM_HDR_LEN. *xlen is set to
+ * the length of that, and doesn't include the collation index size.
+ * Please see sv_collxfrm() to see how this is used. */
+
+#define COLLXFRM_HDR_LEN sizeof(PL_collation_ix)
+
+ char * s = (char *) input_string;
+ STRLEN s_strlen = strlen(input_string);
+ char *xbuf = NULL;
+ STRLEN xAlloc; /* xalloc is a reserved word in VC */
+ STRLEN length_in_chars;
+ bool first_time = TRUE; /* Cleared after first loop iteration */
+
+ PERL_ARGS_ASSERT__MEM_COLLXFRM;
+
+ /* Must be NUL-terminated */
+ assert(*(input_string + len) == '\0');
+
+ /* If this locale has defective collation, skip */
+ if (PL_collxfrm_base == 0 && PL_collxfrm_mult == 0) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "_mem_collxfrm: locale's collation is defective\n"));
+ goto bad;
+ }
+
+ /* Replace any embedded NULs with the control that sorts before any others.
+ * This will give as good as possible results on strings that don't
+ * otherwise contain that character, but otherwise there may be
+ * less-than-perfect results with that character and NUL. This is
+ * unavoidable unless we replace strxfrm with our own implementation. */
+ if (UNLIKELY(s_strlen < len)) { /* Only execute if there is an embedded
+ NUL */
+ char * e = s + len;
+ char * sans_nuls;
+ STRLEN sans_nuls_len;
+ int try_non_controls;
+ char this_replacement_char[] = "?\0"; /* Room for a two-byte string,
+ making sure 2nd byte is NUL.
+ */
+ STRLEN this_replacement_len;
+
+ /* If we don't know what non-NUL control character sorts lowest for
+ * this locale, find it */
+ if (PL_strxfrm_NUL_replacement == '\0') {
+ int j;
+ char * cur_min_x = NULL; /* The min_char's xfrm, (except it also
+ includes the collation index
+ prefixed. */
+
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log, "Looking to replace NUL\n"));
+
+ /* Unlikely, but it may be that no control will work to replace
+ * NUL, in which case we instead look for any character. Controls
+ * are preferred because collation order is, in general, context
+ * sensitive, with adjoining characters affecting the order, and
+ * controls are less likely to have such interactions, allowing the
+ * NUL-replacement to stand on its own. (Another way to look at it
+ * is to imagine what would happen if the NUL were replaced by a
+ * combining character; it wouldn't work out all that well.) */
+ for (try_non_controls = 0;
+ try_non_controls < 2;
+ try_non_controls++)
+ {
+ /* Look through all legal code points (NUL isn't) */
+ for (j = 1; j < 256; j++) {
+ char * x; /* j's xfrm plus collation index */
+ STRLEN x_len; /* length of 'x' */
+ STRLEN trial_len = 1;
+ char cur_source[] = { '\0', '\0' };
+
+ /* Skip non-controls the first time through the loop. The
+ * controls in a UTF-8 locale are the L1 ones */
+ if (! try_non_controls && (PL_in_utf8_COLLATE_locale)
+ ? ! isCNTRL_L1(j)
+ : ! isCNTRL_LC(j))
+ {
+ continue;
+ }
+
+ /* Create a 1-char string of the current code point */
+ cur_source[0] = (char) j;
+
+ /* Then transform it */
+ x = _mem_collxfrm(cur_source, trial_len, &x_len,
+ 0 /* The string is not in UTF-8 */);
+
+ /* Ignore any character that didn't successfully transform.
+ * */
+ if (! x) {
+ continue;
+ }
+
+ /* If this character's transformation is lower than
+ * the current lowest, this one becomes the lowest */
+ if ( cur_min_x == NULL
+ || strLT(x + COLLXFRM_HDR_LEN,
+ cur_min_x + COLLXFRM_HDR_LEN))
+ {
+ PL_strxfrm_NUL_replacement = j;
+ cur_min_x = x;
+ }
+ else {
+ Safefree(x);
+ }
+ } /* end of loop through all 255 characters */
+
+ /* Stop looking if found */
+ if (cur_min_x) {
+ break;
+ }
+
+ /* Unlikely, but possible, if there aren't any controls that
+ * work in the locale, repeat the loop, looking for any
+ * character that works */
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "_mem_collxfrm: No control worked. Trying non-controls\n"));
+ } /* End of loop to try first the controls, then any char */
+
+ if (! cur_min_x) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "_mem_collxfrm: Couldn't find any character to replace"
+ " embedded NULs in locale %s with", PL_collation_name));
+ goto bad;
+ }
+
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "_mem_collxfrm: Replacing embedded NULs in locale %s with "
+ "0x%02X\n", PL_collation_name, PL_strxfrm_NUL_replacement));
+
+ Safefree(cur_min_x);
+ } /* End of determining the character that is to replace NULs */
+
+ /* If the replacement is variant under UTF-8, it must match the
+ * UTF8-ness as the original */
+ if ( ! UVCHR_IS_INVARIANT(PL_strxfrm_NUL_replacement) && utf8) {
+ this_replacement_char[0] =
+ UTF8_EIGHT_BIT_HI(PL_strxfrm_NUL_replacement);
+ this_replacement_char[1] =
+ UTF8_EIGHT_BIT_LO(PL_strxfrm_NUL_replacement);
+ this_replacement_len = 2;
+ }
+ else {
+ this_replacement_char[0] = PL_strxfrm_NUL_replacement;
+ /* this_replacement_char[1] = '\0' was done at initialization */
+ this_replacement_len = 1;
+ }
+
+ /* The worst case length for the replaced string would be if every
+ * character in it is NUL. Multiply that by the length of each
+ * replacement, and allow for a trailing NUL */
+ sans_nuls_len = (len * this_replacement_len) + 1;
+ Newx(sans_nuls, sans_nuls_len, char);
+ *sans_nuls = '\0';
+
+ /* Replace each NUL with the lowest collating control. Loop until have
+ * exhausted all the NULs */
+ while (s + s_strlen < e) {
+ my_strlcat(sans_nuls, s, sans_nuls_len);
+
+ /* Do the actual replacement */
+ my_strlcat(sans_nuls, this_replacement_char, sans_nuls_len);
+
+ /* Move past the input NUL */
+ s += s_strlen + 1;
+ s_strlen = strlen(s);
+ }
+
+ /* And add anything that trails the final NUL */
+ my_strlcat(sans_nuls, s, sans_nuls_len);
+
+ /* Switch so below we transform this modified string */
+ s = sans_nuls;
+ len = strlen(s);
+ } /* End of replacing NULs */
+
+ /* Make sure the UTF8ness of the string and locale match */
+ if (utf8 != PL_in_utf8_COLLATE_locale) {
+ const char * const t = s; /* Temporary so we can later find where the
+ input was */
+
+ /* Here they don't match. Change the string's to be what the locale is
+ * expecting */
+
+ if (! utf8) { /* locale is UTF-8, but input isn't; upgrade the input */
+ s = (char *) bytes_to_utf8((const U8 *) s, &len);
+ utf8 = TRUE;
+ }
+ else { /* locale is not UTF-8; but input is; downgrade the input */
+
+ s = (char *) bytes_from_utf8((const U8 *) s, &len, &utf8);
+
+ /* If the downgrade was successful we are done, but if the input
+ * contains things that require UTF-8 to represent, have to do
+ * damage control ... */
+ if (UNLIKELY(utf8)) {
+
+ /* What we do is construct a non-UTF-8 string with
+ * 1) the characters representable by a single byte converted
+ * to be so (if necessary);
+ * 2) and the rest converted to collate the same as the
+ * highest collating representable character. That makes
+ * them collate at the end. This is similar to how we
+ * handle embedded NULs, but we use the highest collating
+ * code point instead of the smallest. Like the NUL case,
+ * this isn't perfect, but is the best we can reasonably
+ * do. Every above-255 code point will sort the same as
+ * the highest-sorting 0-255 code point. If that code
+ * point can combine in a sequence with some other code
+ * points for weight calculations, us changing something to
+ * be it can adversely affect the results. But in most
+ * cases, it should work reasonably. And note that this is
+ * really an illegal situation: using code points above 255
+ * on a locale where only 0-255 are valid. If two strings
+ * sort entirely equal, then the sort order for the
+ * above-255 code points will be in code point order. */
+
+ utf8 = FALSE;
+
+ /* If we haven't calculated the code point with the maximum
+ * collating order for this locale, do so now */
+ if (! PL_strxfrm_max_cp) {
+ int j;
+
+ /* The current transformed string that collates the
+ * highest (except it also includes the prefixed collation
+ * index. */
+ char * cur_max_x = NULL;
+
+ /* Look through all legal code points (NUL isn't) */
+ for (j = 1; j < 256; j++) {
+ char * x;
+ STRLEN x_len;
+ char cur_source[] = { '\0', '\0' };
+
+ /* Create a 1-char string of the current code point */
+ cur_source[0] = (char) j;
+
+ /* Then transform it */
+ x = _mem_collxfrm(cur_source, 1, &x_len, FALSE);
+
+ /* If something went wrong (which it shouldn't), just
+ * ignore this code point */
+ if (! x) {
+ continue;
+ }
+
+ /* If this character's transformation is higher than
+ * the current highest, this one becomes the highest */
+ if ( cur_max_x == NULL
+ || strGT(x + COLLXFRM_HDR_LEN,
+ cur_max_x + COLLXFRM_HDR_LEN))
+ {
+ PL_strxfrm_max_cp = j;
+ cur_max_x = x;
+ }
+ else {
+ Safefree(x);
+ }
+ }
+
+ if (! cur_max_x) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "_mem_collxfrm: Couldn't find any character to"
+ " replace above-Latin1 chars in locale %s with",
+ PL_collation_name));
+ goto bad;
+ }
+
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "_mem_collxfrm: highest 1-byte collating character"
+ " in locale %s is 0x%02X\n",
+ PL_collation_name,
+ PL_strxfrm_max_cp));
+
+ Safefree(cur_max_x);
+ }
+
+ /* Here we know which legal code point collates the highest.
+ * We are ready to construct the non-UTF-8 string. The length
+ * will be at least 1 byte smaller than the input string
+ * (because we changed at least one 2-byte character into a
+ * single byte), but that is eaten up by the trailing NUL */
+ Newx(s, len, char);
+
+ {
+ STRLEN i;
+ STRLEN d= 0;
+ char * e = (char *) t + len;
+
+ for (i = 0; i < len; i+= UTF8SKIP(t + i)) {
+ U8 cur_char = t[i];
+ if (UTF8_IS_INVARIANT(cur_char)) {
+ s[d++] = cur_char;
+ }
+ else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(t + i, e)) {
+ s[d++] = EIGHT_BIT_UTF8_TO_NATIVE(cur_char, t[i+1]);
+ }
+ else { /* Replace illegal cp with highest collating
+ one */
+ s[d++] = PL_strxfrm_max_cp;
+ }
+ }
+ s[d++] = '\0';
+ Renew(s, d, char); /* Free up unused space */
+ }
+ }
+ }
+
+ /* Here, we have constructed a modified version of the input. It could
+ * be that we already had a modified copy before we did this version.
+ * If so, that copy is no longer needed */
+ if (t != input_string) {
+ Safefree(t);
+ }
+ }
- /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */
- /* the +1 is for the terminating NUL. */
+ length_in_chars = (utf8)
+ ? utf8_length((U8 *) s, (U8 *) s + len)
+ : len;
- xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1;
+ /* The first element in the output is the collation id, used by
+ * sv_collxfrm(); then comes the space for the transformed string. The
+ * equation should give us a good estimate as to how much is needed */
+ xAlloc = COLLXFRM_HDR_LEN
+ + PL_collxfrm_base
+ + (PL_collxfrm_mult * length_in_chars);
Newx(xbuf, xAlloc, char);
- if (! xbuf)
+ if (UNLIKELY(! xbuf)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "_mem_collxfrm: Couldn't malloc %zu bytes\n", xAlloc));
goto bad;
+ }
+ /* Store the collation id */
*(U32*)xbuf = PL_collation_ix;
- xout = sizeof(PL_collation_ix);
- for (xin = 0; xin < len; ) {
- Size_t xused;
-
- for (;;) {
- xused = strxfrm(xbuf + xout, s + xin, xAlloc - xout);
- if (xused >= PERL_INT_MAX)
- goto bad;
- if ((STRLEN)xused < xAlloc - xout)
- break;
- xAlloc = (2 * xAlloc) + 1;
- Renew(xbuf, xAlloc, char);
- if (! xbuf)
- goto bad;
- }
- xin += strlen(s + xin) + 1;
- xout += xused;
+ /* Then the transformation of the input. We loop until successful, or we
+ * give up */
+ for (;;) {
+
+ *xlen = strxfrm(xbuf + COLLXFRM_HDR_LEN, s, xAlloc - COLLXFRM_HDR_LEN);
+
+ /* If the transformed string occupies less space than we told strxfrm()
+ * was available, it means it successfully transformed the whole
+ * string. */
+ if (*xlen < xAlloc - COLLXFRM_HDR_LEN) {
+
+ /* Some systems include a trailing NUL in the returned length.
+ * Ignore it, using a loop in case multiple trailing NULs are
+ * returned. */
+ while ( (*xlen) > 0
+ && *(xbuf + COLLXFRM_HDR_LEN + (*xlen) - 1) == '\0')
+ {
+ (*xlen)--;
+ }
+
+ /* If the first try didn't get it, it means our prediction was low.
+ * Modify the coefficients so that we predict a larger value in any
+ * future transformations */
+ if (! first_time) {
+ STRLEN needed = *xlen + 1; /* +1 For trailing NUL */
+ STRLEN computed_guess = PL_collxfrm_base
+ + (PL_collxfrm_mult * length_in_chars);
+
+ /* On zero-length input, just keep current slope instead of
+ * dividing by 0 */
+ const STRLEN new_m = (length_in_chars != 0)
+ ? needed / length_in_chars
+ : PL_collxfrm_mult;
+
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "%s: %d: initial size of %zu bytes for a length "
+ "%zu string was insufficient, %zu needed\n",
+ __FILE__, __LINE__,
+ computed_guess, length_in_chars, needed));
+
+ /* If slope increased, use it, but discard this result for
+ * length 1 strings, as we can't be sure that it's a real slope
+ * change */
+ if (length_in_chars > 1 && new_m > PL_collxfrm_mult) {
+#ifdef DEBUGGING
+ STRLEN old_m = PL_collxfrm_mult;
+ STRLEN old_b = PL_collxfrm_base;
+#endif
+ PL_collxfrm_mult = new_m;
+ PL_collxfrm_base = 1; /* +1 For trailing NUL */
+ computed_guess = PL_collxfrm_base
+ + (PL_collxfrm_mult * length_in_chars);
+ if (computed_guess < needed) {
+ PL_collxfrm_base += needed - computed_guess;
+ }
+
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "%s: %d: slope is now %zu; was %zu, base "
+ "is now %zu; was %zu\n",
+ __FILE__, __LINE__,
+ PL_collxfrm_mult, old_m,
+ PL_collxfrm_base, old_b));
+ }
+ else { /* Slope didn't change, but 'b' did */
+ const STRLEN new_b = needed
+ - computed_guess
+ + PL_collxfrm_base;
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "%s: %d: base is now %zu; was %zu\n",
+ __FILE__, __LINE__,
+ new_b, PL_collxfrm_base));
+ PL_collxfrm_base = new_b;
+ }
+ }
+
+ break;
+ }
+
+ if (UNLIKELY(*xlen >= PERL_INT_MAX)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "_mem_collxfrm: Needed %zu bytes, max permissible is %u\n",
+ *xlen, PERL_INT_MAX));
+ goto bad;
+ }
+
+ /* A well-behaved strxfrm() returns exactly how much space it needs
+ * (usually not including the trailing NUL) when it fails due to not
+ * enough space being provided. Assume that this is the case unless
+ * it's been proven otherwise */
+ if (LIKELY(PL_strxfrm_is_behaved) && first_time) {
+ xAlloc = *xlen + COLLXFRM_HDR_LEN + 1;
+ }
+ else { /* Here, either:
+ * 1) The strxfrm() has previously shown bad behavior; or
+ * 2) It isn't the first time through the loop, which means
+ * that the strxfrm() is now showing bad behavior, because
+ * we gave it what it said was needed in the previous
+ * iteration, and it came back saying it needed still more.
+ * (Many versions of cygwin fit this. When the buffer size
+ * isn't sufficient, they return the input size instead of
+ * how much is needed.)
+ * Increase the buffer size by a fixed percentage and try again.
+ * */
+ xAlloc += (xAlloc / 4) + 1;
+ PL_strxfrm_is_behaved = FALSE;
+
+#ifdef DEBUGGING
+ if (DEBUG_Lv_TEST || debug_initialization) {
+ PerlIO_printf(Perl_debug_log,
+ "_mem_collxfrm required more space than previously calculated"
+ " for locale %s, trying again with new guess=%d+%zu\n",
+ PL_collation_name, (int) COLLXFRM_HDR_LEN,
+ xAlloc - COLLXFRM_HDR_LEN);
+ }
+#endif
+ }
+
+ Renew(xbuf, xAlloc, char);
+ if (UNLIKELY(! xbuf)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "_mem_collxfrm: Couldn't realloc %zu bytes\n", xAlloc));
+ goto bad;
+ }
+
+ first_time = FALSE;
+ }
+
+
+#ifdef DEBUGGING
+ if (DEBUG_Lv_TEST || debug_initialization) {
+
+ print_collxfrm_input_and_return(s, s + len, xlen, utf8);
+ PerlIO_printf(Perl_debug_log, "Its xfrm is:");
+ PerlIO_printf(Perl_debug_log, "%s\n",
+ _byte_dump_string((U8 *) xbuf + COLLXFRM_HDR_LEN,
+ *xlen, 1));
+ }
+#endif
- /* Embedded NULs are understood but silently skipped
- * because they make no sense in locale collation. */
+ /* Free up unneeded space; retain ehough for trailing NUL */
+ Renew(xbuf, COLLXFRM_HDR_LEN + *xlen + 1, char);
+
+ if (s != input_string) {
+ Safefree(s);
}
- xbuf[xout] = '\0';
- *xlen = xout - sizeof(PL_collation_ix);
return xbuf;
bad:
Safefree(xbuf);
+ if (s != input_string) {
+ Safefree(s);
+ }
*xlen = 0;
+#ifdef DEBUGGING
+ if (DEBUG_Lv_TEST || debug_initialization) {
+ print_collxfrm_input_and_return(s, s + len, NULL, utf8);
+ }
+#endif
return NULL;
}
+#ifdef DEBUGGING
+
+STATIC void
+S_print_collxfrm_input_and_return(pTHX_
+ const char * const s,
+ const char * const e,
+ const STRLEN * const xlen,
+ const bool is_utf8)
+{
+
+ PERL_ARGS_ASSERT_PRINT_COLLXFRM_INPUT_AND_RETURN;
+
+ PerlIO_printf(Perl_debug_log, "_mem_collxfrm[%" UVuf "]: returning ",
+ (UV)PL_collation_ix);
+ if (xlen) {
+ PerlIO_printf(Perl_debug_log, "%zu", *xlen);
+ }
+ else {
+ PerlIO_printf(Perl_debug_log, "NULL");
+ }
+ PerlIO_printf(Perl_debug_log, " for locale '%s', string='",
+ PL_collation_name);
+ print_bytes_for_locale(s, e, is_utf8);
+
+ PerlIO_printf(Perl_debug_log, "'\n");
+}
+
+STATIC void
+S_print_bytes_for_locale(pTHX_
+ const char * const s,
+ const char * const e,
+ const bool is_utf8)
+{
+ const char * t = s;
+ bool prev_was_printable = TRUE;
+ bool first_time = TRUE;
+
+ PERL_ARGS_ASSERT_PRINT_BYTES_FOR_LOCALE;
+
+ while (t < e) {
+ UV cp = (is_utf8)
+ ? utf8_to_uvchr_buf((U8 *) t, e, NULL)
+ : * (U8 *) t;
+ if (isPRINT(cp)) {
+ if (! prev_was_printable) {
+ PerlIO_printf(Perl_debug_log, " ");
+ }
+ PerlIO_printf(Perl_debug_log, "%c", (U8) cp);
+ prev_was_printable = TRUE;
+ }
+ else {
+ if (! first_time) {
+ PerlIO_printf(Perl_debug_log, " ");
+ }
+ PerlIO_printf(Perl_debug_log, "%02" UVXf, cp);
+ prev_was_printable = FALSE;
+ }
+ t += (is_utf8) ? UTF8SKIP(t) : 1;
+ first_time = FALSE;
+ }
+}
+
+#endif /* #ifdef DEBUGGING */
+
#endif /* USE_LOCALE_COLLATE */
#ifdef USE_LOCALE
lc = localeconv();
if (! lc
|| ! lc->currency_symbol
- || is_invariant_string((U8 *) lc->currency_symbol, 0))
+ || is_utf8_invariant_string((U8 *) lc->currency_symbol, 0))
{
DEBUG_L(PerlIO_printf(Perl_debug_log, "Couldn't get currency symbol for %s, or contains only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale));
only_ascii = TRUE;
for (i = 0; i < 7 + 12; i++) { /* 7 days; 12 months */
formatted_time = my_strftime("%A %B %Z %p",
0, 0, hour, dom, month, 112, 0, 0, is_dst);
- if (! formatted_time || is_invariant_string((U8 *) formatted_time, 0)) {
+ if ( ! formatted_time
+ || is_utf8_invariant_string((U8 *) formatted_time, 0))
+ {
/* Here, we didn't find a non-ASCII. Try the next time through
* with the complemented dst and am/pm, and try with the next
break;
}
errmsg = savepv(errmsg);
- if (! is_invariant_string((U8 *) errmsg, 0)) {
+ if (! is_utf8_invariant_string((U8 *) errmsg, 0)) {
non_ascii = TRUE;
is_utf8 = is_utf8_string((U8 *) errmsg, 0);
break;
}
char *
-Perl_my_strerror(pTHX_ const int errnum) {
+Perl_my_strerror(pTHX_ const int errnum)
+{
+ /* Returns a mortalized copy of the text of the error message associated
+ * with 'errnum'. It uses the current locale's text unless the platform
+ * doesn't have the LC_MESSAGES category or we are not being called from
+ * within the scope of 'use locale'. In the former case, it uses whatever
+ * strerror returns; in the latter case it uses the text from the C locale.
+ *
+ * The function just calls strerror(), but temporarily switches, if needed,
+ * to the C locale */
+
+ char *errstr;
+ dVAR;
- /* Uses C locale for the error text unless within scope of 'use locale' for
- * LC_MESSAGES */
+#ifndef USE_LOCALE_MESSAGES
-#ifdef USE_LOCALE_MESSAGES
- if (! IN_LC(LC_MESSAGES)) {
- char * save_locale = setlocale(LC_MESSAGES, NULL);
- if (! isNAME_C_OR_POSIX(save_locale)) {
- char *errstr;
+ /* If platform doesn't have messages category, we don't do any switching to
+ * the C locale; we just use whatever strerror() returns */
+
+ errstr = savepv(Strerror(errnum));
+
+#else /* Has locale messages */
+
+ const bool within_locale_scope = IN_LC(LC_MESSAGES);
+
+# if defined(HAS_POSIX_2008_LOCALE) && defined(HAS_STRERROR_L)
+
+ /* This function is trivial if we have strerror_l() */
+
+ if (within_locale_scope) {
+ errstr = strerror(errnum);
+ }
+ else {
+ errstr = strerror_l(errnum, PL_C_locale_obj);
+ }
+
+ errstr = savepv(errstr);
+
+# else /* Doesn't have strerror_l(). */
+
+# ifdef USE_POSIX_2008_LOCALE
+
+ locale_t save_locale = NULL;
+
+# else
+
+ char * save_locale = NULL;
+ bool locale_is_C = FALSE;
+
+ /* We have a critical section to prevent another thread from changing the
+ * locale out from under us (or zapping the buffer returned from
+ * setlocale() ) */
+ LOCALE_LOCK;
+
+# endif
+
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "my_strerror called with errnum %d\n", errnum));
+ if (! within_locale_scope) {
+ errno = 0;
+
+# ifdef USE_POSIX_2008_LOCALE /* Use the thread-safe locale functions */
+
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "Not within locale scope, about to call"
+ " uselocale(0x%p)\n", PL_C_locale_obj));
+ save_locale = uselocale(PL_C_locale_obj);
+ if (! save_locale) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "uselocale failed, errno=%d\n", errno));
+ }
+ else {
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "uselocale returned 0x%p\n", save_locale));
+ }
+
+# else /* Not thread-safe build */
+
+ save_locale = setlocale(LC_MESSAGES, NULL);
+ if (! save_locale) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "setlocale failed, errno=%d\n", errno));
+ }
+ else {
+ locale_is_C = isNAME_C_OR_POSIX(save_locale);
+
+ /* Switch to the C locale if not already in it */
+ if (! locale_is_C) {
+
+ /* The setlocale() just below likely will zap 'save_locale', so
+ * create a copy. */
+ save_locale = savepv(save_locale);
+ setlocale(LC_MESSAGES, "C");
+ }
+ }
+
+# endif
+
+ } /* end of ! within_locale_scope */
+ else {
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log, "%s: %d: WITHIN locale scope\n",
+ __FILE__, __LINE__));
+ }
+
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "Any locale change has been done; about to call Strerror\n"));
+ errstr = savepv(Strerror(errnum));
- /* The next setlocale likely will zap this, so create a copy */
- save_locale = savepv(save_locale);
+ if (! within_locale_scope) {
+ errno = 0;
- setlocale(LC_MESSAGES, "C");
+# ifdef USE_POSIX_2008_LOCALE
+
+ DEBUG_Lv(PerlIO_printf(Perl_debug_log,
+ "%s: %d: not within locale scope, restoring the locale\n",
+ __FILE__, __LINE__));
+ if (save_locale && ! uselocale(save_locale)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "uselocale restore failed, errno=%d\n", errno));
+ }
+ }
- /* This points to the static space in Strerror, with all its
- * limitations */
- errstr = Strerror(errnum);
+# else
- setlocale(LC_MESSAGES, save_locale);
+ if (save_locale && ! locale_is_C) {
+ if (! setlocale(LC_MESSAGES, save_locale)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "setlocale restore failed, errno=%d\n", errno));
+ }
Safefree(save_locale);
- return errstr;
}
}
+
+ LOCALE_UNLOCK;
+
+# endif
+# endif /* End of doesn't have strerror_l */
+#endif /* End of does have locale messages */
+
+#ifdef DEBUGGING
+
+ if (DEBUG_Lv_TEST) {
+ PerlIO_printf(Perl_debug_log, "Strerror returned; saving a copy: '");
+ print_bytes_for_locale(errstr, errstr + strlen(errstr), 0);
+ PerlIO_printf(Perl_debug_log, "'\n");
+ }
+
#endif
- return Strerror(errnum);
+ SAVEFREEPV(errstr);
+ return errstr;
}
/*
-=head1 Locale-related functions and macros
-
=for apidoc sync_locale
Changing the program's locale should be avoided by XS code. Nevertheless,
#if defined(DEBUGGING) && defined(USE_LOCALE)
-char *
-Perl__setlocale_debug_string(const int category, /* category number,
+STATIC char *
+S_setlocale_debug_string(const int category, /* category number,
like LC_ALL */
const char* const locale, /* locale name */
/* Returns a pointer to a NUL-terminated string in static storage with
* added text about the info passed in. This is not thread safe and will
* be overwritten by the next call, so this should be used just to
- * formulate a string to immediately print or savepv() on.
- *
- * Buffer overflow checking is done only after the fact (via an assert),
- * because this is used only in DEBUGGING, and an attacker would have to
- * control the start up of perl with the correct environment variable or
- * command line option. */
-
- static char ret[128] = "";
+ * formulate a string to immediately print or savepv() on. */
- strcpy(ret, "setlocale(");
+ /* initialise to a non-null value to keep it out of BSS and so keep
+ * -DPERL_GLOBAL_STRUCT_PRIVATE happy */
+ static char ret[128] = "If you can read this, thank your buggy C"
+ " library strlcpy(), and change your hints file"
+ " to undef it";
+ my_strlcpy(ret, "setlocale(", sizeof(ret));
switch (category) {
default:
- sprintf(ret, "%s? %d", ret, category);
+ my_snprintf(ret, sizeof(ret), "%s? %d", ret, category);
break;
# ifdef LC_ALL
case LC_ALL:
- strcat(ret, "LC_ALL");
+ my_strlcat(ret, "LC_ALL", sizeof(ret));
break;
# endif
# ifdef LC_CTYPE
case LC_CTYPE:
- strcat(ret, "LC_CTYPE");
+ my_strlcat(ret, "LC_CTYPE", sizeof(ret));
break;
# endif
# ifdef LC_NUMERIC
case LC_NUMERIC:
- strcat(ret, "LC_NUMERIC");
+ my_strlcat(ret, "LC_NUMERIC", sizeof(ret));
break;
# endif
# ifdef LC_COLLATE
case LC_COLLATE:
- strcat(ret, "LC_COLLATE");
+ my_strlcat(ret, "LC_COLLATE", sizeof(ret));
break;
# endif
# ifdef LC_TIME
case LC_TIME:
- strcat(ret, "LC_TIME");
+ my_strlcat(ret, "LC_TIME", sizeof(ret));
break;
# endif
# ifdef LC_MONETARY
case LC_MONETARY:
- strcat(ret, "LC_MONETARY");
+ my_strlcat(ret, "LC_MONETARY", sizeof(ret));
break;
# endif
# ifdef LC_MESSAGES
case LC_MESSAGES:
- strcat(ret, "LC_MESSAGES");
+ my_strlcat(ret, "LC_MESSAGES", sizeof(ret));
break;
# endif
}
- strcat(ret, ", ");
+ my_strlcat(ret, ", ", sizeof(ret));
if (locale) {
- strcat(ret, "\"");
- strcat(ret, locale);
- strcat(ret, "\"");
+ my_strlcat(ret, "\"", sizeof(ret));
+ my_strlcat(ret, locale, sizeof(ret));
+ my_strlcat(ret, "\"", sizeof(ret));
}
else {
- strcat(ret, "NULL");
+ my_strlcat(ret, "NULL", sizeof(ret));
}
- strcat(ret, ") returned ");
+ my_strlcat(ret, ") returned ", sizeof(ret));
if (retval) {
- strcat(ret, "\"");
- strcat(ret, retval);
- strcat(ret, "\"");
+ my_strlcat(ret, "\"", sizeof(ret));
+ my_strlcat(ret, retval, sizeof(ret));
+ my_strlcat(ret, "\"", sizeof(ret));
}
else {
- strcat(ret, "NULL");
+ my_strlcat(ret, "NULL", sizeof(ret));
}
assert(strlen(ret) < sizeof(ret));