* 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"
#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
/*
sv_setpv(PL_numeric_radix_sv, lc->decimal_point);
else
PL_numeric_radix_sv = newSVpv(lc->decimal_point, 0);
- if (! is_ascii_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)
- ? lc->decimal_point
- : "NULL"));
+ ? SvPVX(PL_numeric_radix_sv)
+ : "NULL",
+ (PL_numeric_radix_sv)
+ ? cBOOL(SvUTF8(PL_numeric_radix_sv))
+ : 0);
+ }
+#endif
# endif /* HAS_LOCALECONV */
#endif /* USE_LOCALE_NUMERIC */
* punct: 21-2F 3A-40 5B-60 7B-7E A1-A3 A5 A7-AB B0-B3 B5-B7 B9-BD BF-CF D1-DD DF-EF F1-FD
* Oddly, none there are listed as alphas, though some represent alphabetics
* http://www.nntp.perl.org/group/perl.perl5.porters/2013/02/msg198753.html */
-#define isNAME_C_OR_POSIX(name) ((name) != NULL \
- && ((*(name) == 'C' && (*(name) + 1) == '\0') \
+#define isNAME_C_OR_POSIX(name) ((name) != NULL \
+ && ((*(name) == 'C' && (*(name + 1)) == '\0') \
|| strEQ((name), "POSIX")))
void
* dot.
*
* This sets several interpreter-level variables:
- * PL_numeric_name The default locale's name: a copy of 'newnum'
+ * PL_numeric_name The underlying locale's name: a copy of 'newnum'
* PL_numeric_local A boolean indicating if the toggled state is such
* that the current locale is the program's underlying
* locale
}
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 */
}
PERL_ARGS_ASSERT_NEW_CTYPE;
+ /* We will replace any bad locale warning with 1) nothing if the new one is
+ * ok; or 2) a new warning for the bad new locale */
+ if (PL_warn_locale) {
+ SvREFCNT_dec_NN(PL_warn_locale);
+ PL_warn_locale = NULL;
+ }
+
PL_in_utf8_CTYPE_locale = _is_cur_LC_category_utf8(LC_CTYPE);
/* A UTF-8 locale gets standard rules. But note that code still has to
Copy(PL_fold_latin1, PL_fold_locale, 256, U8);
}
else {
+ /* Assume enough space for every character being bad. 4 spaces each
+ * for the 94 printable characters that are output like "'x' "; and 5
+ * spaces each for "'\\' ", "'\t' ", and "'\n' "; plus a terminating
+ * NUL */
+ char bad_chars_list[ (94 * 4) + (3 * 5) + 1 ];
+
+ /* 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 */
+
for (i = 0; i < 256; i++) {
if (isUPPER_LC((U8) i))
PL_fold_locale[i] = (U8) toLOWER_LC((U8) i);
PL_fold_locale[i] = (U8) toUPPER_LC((U8) i);
else
PL_fold_locale[i] = (U8) i;
+
+ /* If checking for locale problems, see if the native ASCII-range
+ * printables plus \n and \t are in their expected categories in
+ * the new locale. If not, this could mean big trouble, upending
+ * Perl's and most programs' assumptions, like having a
+ * metacharacter with special meaning become a \w. Fortunately,
+ * it's very rare to find locales that aren't supersets of ASCII
+ * nowadays. It isn't a problem for most controls to be changed
+ * into something else; we check only \n and \t, though perhaps \r
+ * could be an issue as well. */
+ if (check_for_problems
+ && (isGRAPH_A(i) || isBLANK_A(i) || i == '\n'))
+ {
+ if ((isALPHANUMERIC_A(i) && ! isALPHANUMERIC_LC(i))
+ || (isPUNCT_A(i) && ! isPUNCT_LC(i))
+ || (isBLANK_A(i) && ! isBLANK_LC(i))
+ || (i == '\n' && ! isCNTRL_LC(i)))
+ {
+ if (bad_count) { /* Separate multiple entries with a
+ blank */
+ bad_chars_list[bad_count++] = ' ';
+ }
+ bad_chars_list[bad_count++] = '\'';
+ if (isPRINT_A(i)) {
+ bad_chars_list[bad_count++] = (char) i;
+ }
+ else {
+ bad_chars_list[bad_count++] = '\\';
+ if (i == '\n') {
+ bad_chars_list[bad_count++] = 'n';
+ }
+ else {
+ assert(i == '\t');
+ bad_chars_list[bad_count++] = 't';
+ }
+ }
+ bad_chars_list[bad_count++] = '\'';
+ bad_chars_list[bad_count] = '\0';
+ }
+ }
+ }
+
+#ifdef MB_CUR_MAX
+ /* We only handle single-byte locales (outside of UTF-8 ones; so if
+ * this locale requires more than one byte, there are going to be
+ * problems. */
+ if (check_for_problems && MB_CUR_MAX > 1
+
+ /* Some platforms return MB_CUR_MAX > 1 for even the "C"
+ * locale. Just assume that the implementation for them (plus
+ * for POSIX) is correct and the > 1 value is spurious. (Since
+ * these are specially handled to never be considered UTF-8
+ * locales, as long as this is the only problem, everything
+ * should work fine */
+ && strNE(newctype, "C") && strNE(newctype, "POSIX"))
+ {
+ multi_byte_locale = TRUE;
+ }
+#endif
+
+ if (bad_count || multi_byte_locale) {
+ PL_warn_locale = Perl_newSVpvf(aTHX_
+ "Locale '%s' may not work well.%s%s%s\n",
+ newctype,
+ (multi_byte_locale)
+ ? " Some characters in it are not recognized by"
+ " Perl."
+ : "",
+ (bad_count)
+ ? "\nThe following characters (and maybe others)"
+ " may not have the same meaning as the Perl"
+ " program expects:\n"
+ : "",
+ (bad_count)
+ ? bad_chars_list
+ : ""
+ );
+ /* If we are actually in the scope of the locale or are debugging,
+ * 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) || UNLIKELY(DEBUG_L_TEST)) {
+
+ /* We have to save 'newctype' because the setlocale() just
+ * below may destroy it. The next setlocale() further down
+ * should restore it properly so that the intermediate change
+ * here is transparent to this function's caller */
+ const char * const badlocale = savepv(newctype);
+
+ setlocale(LC_CTYPE, "C");
+
+ /* 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;
+ }
}
}
}
void
+Perl__warn_problematic_locale()
+{
+
+#ifdef USE_LOCALE_CTYPE
+
+ dTHX;
+
+ /* Internal-to-core function that outputs the message in PL_warn_locale,
+ * and then NULLS it. Should be called only through the macro
+ * _CHECK_AND_WARN_PROBLEMATIC_LOCALE */
+
+ if (PL_warn_locale) {
+ /*GCC_DIAG_IGNORE(-Wformat-security); Didn't work */
+ Perl_ck_warner(aTHX_ packWARN(WARN_LOCALE),
+ SvPVX(PL_warn_locale),
+ 0 /* dummy to avoid compiler warning */ );
+ /* GCC_DIAG_RESTORE; */
+ SvREFCNT_dec_NN(PL_warn_locale);
+ PL_warn_locale = NULL;
+ }
+
+#endif
+
+}
+
+void
Perl_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_min_char = '\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_min_char = '\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
}
}
* otherwise to use the particular category's variable if set; otherwise to
* use the LANG variable. */
- bool override_LC_ALL = 0;
+ bool override_LC_ALL = FALSE;
char * result;
if (locale && strEQ(locale, "")) {
}
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;
}
- /* Here the input locale was LC_ALL, and we have set it to what is in the
+ /* Here the input category was LC_ALL, and we have set it to what is in the
* LANG variable or the system default if there is no LANG. But these have
* lower priority than the other LC_foo variables, so override it for each
* one that is set. (If they are set to "", it means to use the same thing
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
* 1 = set ok or not applicable,
* 0 = fallback to a locale of lower priority
* -1 = fallback to all locales failed, not even to the C locale
- */
+ *
+ * Under -DDEBUGGING, if the environment variable PERL_DEBUG_LOCALE_INIT is
+ * set, debugging information is output.
+ *
+ * This looks more complicated than it is, mainly due to the #ifdefs.
+ *
+ * We try to set LC_ALL to the value determined by the environment. If
+ * there is no LC_ALL on this platform, we try the individual categories we
+ * know about. If this works, we are done.
+ *
+ * But if it doesn't work, we have to do something else. We search the
+ * environment variables ourselves instead of relying on the system to do
+ * it. We look at, in order, LC_ALL, LANG, a system default locale (if we
+ * think there is one), and the ultimate fallback "C". This is all done in
+ * the same loop as above to avoid duplicating code, but it makes things
+ * more complex. After the original failure, we add the fallback
+ * possibilities to the list of locales to try, and iterate the loop
+ * through them all until one succeeds.
+ *
+ * On Ultrix, the locale MUST come from the environment, so there is
+ * preliminary code to set it. I (khw) am not sure that it is necessary,
+ * and that this couldn't be folded into the loop, but barring any real
+ * platforms to test on, it's staying as-is
+ *
+ * A slight complication is that in embedded Perls, the locale may already
+ * be set-up, and we don't want to get it from the normal environment
+ * variables. This is handled by having a special environment variable
+ * indicate we're in this situation. We simply set setlocale's 2nd
+ * parameter to be a NULL instead of "". That indicates to setlocale that
+ * it is not to change anything, but to return the current value,
+ * effectively initializing perl's db to what the locale already is.
+ *
+ * We play the same trick with NULL if a LC_ALL succeeds. We call
+ * setlocale() on the individual categores with NULL to get their existing
+ * values for our db, instead of trying to change them.
+ * */
int ok = 1;
char *curnum = NULL;
#endif /* USE_LOCALE_NUMERIC */
#ifdef __GLIBC__
- char * const language = PerlEnv_getenv("LANGUAGE");
+ const char * const language = savepv(PerlEnv_getenv("LANGUAGE"));
#endif
/* NULL uses the existing already set up locale */
: "";
const char* trial_locales[5]; /* 5 = 1 each for "", LC_ALL, LANG, "", C */
unsigned int trial_locales_count;
- char * const lc_all = PerlEnv_getenv("LC_ALL");
- char * const lang = PerlEnv_getenv("LANG");
+ const char * const lc_all = savepv(PerlEnv_getenv("LC_ALL"));
+ const char * const lang = savepv(PerlEnv_getenv("LANG"));
bool setlocale_failure = FALSE;
unsigned int i;
char *p;
- const bool locwarn = (printwarn > 1 ||
- (printwarn &&
- (!(p = PerlEnv_getenv("PERL_BADLANG")) || atoi(p))));
+
+ /* A later getenv() could zap this, so only use here */
+ const char * const bad_lang_use_once = PerlEnv_getenv("PERL_BADLANG");
+
+ const bool locwarn = (printwarn > 1
+ || (printwarn
+ && (! bad_lang_use_once
+ || (
+ /* disallow with "" or "0" */
+ *bad_lang_use_once
+ && strNE("0", bad_lang_use_once)))));
bool done = FALSE;
+ char * sl_result; /* return from setlocale() */
+ char * locale_param;
#ifdef WIN32
/* In some systems you can find out the system default locale
* and use that as the fallback locale. */
const char *system_default_locale = NULL;
#endif
+#ifdef DEBUGGING
+ DEBUG_INITIALIZATION_set((PerlEnv_getenv("PERL_DEBUG_LOCALE_INIT"))
+ ? TRUE
+ : FALSE);
+# 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);
#else
/*
# ifdef LC_ALL
if (lang) {
- if (my_setlocale(LC_ALL, setlocale_init))
+ sl_result = my_setlocale(LC_ALL, setlocale_init);
+ DEBUG_LOCALE_INIT(LC_ALL, setlocale_init, sl_result);
+ if (sl_result)
done = TRUE;
else
setlocale_failure = TRUE;
}
- if (!setlocale_failure) {
+ if (! setlocale_failure) {
# ifdef USE_LOCALE_CTYPE
- Safefree(curctype);
- if (! (curctype =
- my_setlocale(LC_CTYPE,
- (!done && (lang || PerlEnv_getenv("LC_CTYPE")))
- ? setlocale_init : NULL)))
+ locale_param = (! done && (lang || PerlEnv_getenv("LC_CTYPE")))
+ ? setlocale_init
+ : NULL;
+ curctype = my_setlocale(LC_CTYPE, locale_param);
+ DEBUG_LOCALE_INIT(LC_CTYPE, locale_param, sl_result);
+ if (! curctype)
setlocale_failure = TRUE;
else
curctype = savepv(curctype);
# endif /* USE_LOCALE_CTYPE */
# ifdef USE_LOCALE_COLLATE
- Safefree(curcoll);
- if (! (curcoll =
- my_setlocale(LC_COLLATE,
- (!done && (lang || PerlEnv_getenv("LC_COLLATE")))
- ? setlocale_init : NULL)))
+ locale_param = (! done && (lang || PerlEnv_getenv("LC_COLLATE")))
+ ? setlocale_init
+ : NULL;
+ curcoll = my_setlocale(LC_COLLATE, locale_param);
+ DEBUG_LOCALE_INIT(LC_COLLATE, locale_param, sl_result);
+ if (! curcoll)
setlocale_failure = TRUE;
else
curcoll = savepv(curcoll);
# endif /* USE_LOCALE_COLLATE */
# ifdef USE_LOCALE_NUMERIC
- Safefree(curnum);
- if (! (curnum =
- my_setlocale(LC_NUMERIC,
- (!done && (lang || PerlEnv_getenv("LC_NUMERIC")))
- ? setlocale_init : NULL)))
+ locale_param = (! done && (lang || PerlEnv_getenv("LC_NUMERIC")))
+ ? setlocale_init
+ : NULL;
+ curnum = my_setlocale(LC_NUMERIC, locale_param);
+ DEBUG_LOCALE_INIT(LC_NUMERIC, locale_param, sl_result);
+ if (! curnum)
setlocale_failure = TRUE;
else
curnum = savepv(curnum);
# endif /* USE_LOCALE_NUMERIC */
# ifdef USE_LOCALE_MESSAGES
- if (! my_setlocale(LC_MESSAGES,
- (!done && (lang || PerlEnv_getenv("LC_MESSAGES")))
- ? setlocale_init : NULL))
- {
+ locale_param = (! done && (lang || PerlEnv_getenv("LC_MESSAGES")))
+ ? setlocale_init
+ : NULL;
+ sl_result = my_setlocale(LC_MESSAGES, locale_param);
+ DEBUG_LOCALE_INIT(LC_MESSAGES, locale_param, sl_result);
+ if (! sl_result) {
setlocale_failure = TRUE;
}
# endif /* USE_LOCALE_MESSAGES */
# ifdef USE_LOCALE_MONETARY
- if (! my_setlocale(LC_MONETARY,
- (!done && (lang || PerlEnv_getenv("LC_MONETARY")))
- ? setlocale_init : NULL))
- {
+ locale_param = (! done && (lang || PerlEnv_getenv("LC_MONETARY")))
+ ? setlocale_init
+ : NULL;
+ sl_result = my_setlocale(LC_MONETARY, locale_param);
+ DEBUG_LOCALE_INIT(LC_MONETARY, locale_param, sl_result);
+ if (! sl_result) {
setlocale_failure = TRUE;
}
# endif /* USE_LOCALE_MONETARY */
#endif /* !LOCALE_ENVIRON_REQUIRED */
/* We try each locale in the list until we get one that works, or exhaust
- * the list */
+ * the list. Normally the loop is executed just once. But if setting the
+ * locale fails, inside the loop we add fallback trials to the array and so
+ * will execute the loop multiple times */
trial_locales[0] = setlocale_init;
trial_locales_count = 1;
for (i= 0; i < trial_locales_count; i++) {
/* Note that this may change the locale, but we are going to do
* that anyway just below */
system_default_locale = setlocale(LC_ALL, "");
+ DEBUG_LOCALE_INIT(LC_ALL, "", system_default_locale);
/* Skip if invalid or it's already on the list of locales to
* try */
}
#ifdef LC_ALL
- if (! my_setlocale(LC_ALL, trial_locale)) {
+ sl_result = my_setlocale(LC_ALL, trial_locale);
+ DEBUG_LOCALE_INIT(LC_ALL, trial_locale, sl_result);
+ if (! sl_result) {
setlocale_failure = TRUE;
}
else {
if (!setlocale_failure) {
#ifdef USE_LOCALE_CTYPE
Safefree(curctype);
- if (! (curctype = my_setlocale(LC_CTYPE, trial_locale)))
+ curctype = my_setlocale(LC_CTYPE, trial_locale);
+ DEBUG_LOCALE_INIT(LC_CTYPE, trial_locale, curctype);
+ if (! curctype)
setlocale_failure = TRUE;
else
curctype = savepv(curctype);
#endif /* USE_LOCALE_CTYPE */
#ifdef USE_LOCALE_COLLATE
Safefree(curcoll);
- if (! (curcoll = my_setlocale(LC_COLLATE, trial_locale)))
+ curcoll = my_setlocale(LC_COLLATE, trial_locale);
+ DEBUG_LOCALE_INIT(LC_COLLATE, trial_locale, curcoll);
+ if (! curcoll)
setlocale_failure = TRUE;
else
curcoll = savepv(curcoll);
#endif /* USE_LOCALE_COLLATE */
#ifdef USE_LOCALE_NUMERIC
Safefree(curnum);
- if (! (curnum = my_setlocale(LC_NUMERIC, trial_locale)))
+ curnum = my_setlocale(LC_NUMERIC, trial_locale);
+ DEBUG_LOCALE_INIT(LC_NUMERIC, trial_locale, curnum);
+ if (! curnum)
setlocale_failure = TRUE;
else
curnum = savepv(curnum);
#endif /* USE_LOCALE_NUMERIC */
#ifdef USE_LOCALE_MESSAGES
- if (! (my_setlocale(LC_MESSAGES, trial_locale)))
+ sl_result = my_setlocale(LC_MESSAGES, trial_locale);
+ DEBUG_LOCALE_INIT(LC_MESSAGES, trial_locale, sl_result);
+ if (! (sl_result))
setlocale_failure = TRUE;
#endif /* USE_LOCALE_MESSAGES */
#ifdef USE_LOCALE_MONETARY
- if (! (my_setlocale(LC_MONETARY, trial_locale)))
+ sl_result = my_setlocale(LC_MONETARY, trial_locale);
+ DEBUG_LOCALE_INIT(LC_MONETARY, trial_locale, sl_result);
+ if (! (sl_result))
setlocale_failure = TRUE;
#endif /* USE_LOCALE_MONETARY */
PerlIO_printf(Perl_error_log,
"perl: warning: Setting locale failed for the categories:\n\t");
-#ifdef USE_LOCALE_CTYPE
+# ifdef USE_LOCALE_CTYPE
if (! curctype)
PerlIO_printf(Perl_error_log, "LC_CTYPE ");
-#endif /* USE_LOCALE_CTYPE */
-#ifdef USE_LOCALE_COLLATE
+# endif /* USE_LOCALE_CTYPE */
+# ifdef USE_LOCALE_COLLATE
if (! curcoll)
PerlIO_printf(Perl_error_log, "LC_COLLATE ");
-#endif /* USE_LOCALE_COLLATE */
-#ifdef USE_LOCALE_NUMERIC
+# endif /* USE_LOCALE_COLLATE */
+# ifdef USE_LOCALE_NUMERIC
if (! curnum)
PerlIO_printf(Perl_error_log, "LC_NUMERIC ");
-#endif /* USE_LOCALE_NUMERIC */
+# endif /* USE_LOCALE_NUMERIC */
PerlIO_printf(Perl_error_log, "and possibly others\n");
#endif /* LC_ALL */
{
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);
}
/* Calculate what fallback locales to try. We have avoided this
- * until we have to, becuase failure is quite unlikely. This will
+ * until we have to, because failure is quite unlikely. This will
* usually change the upper bound of the loop we are in.
*
* Since the system's default way of setting the locale has not
* LANG, and the C locale. We don't try the same locale twice, so
* don't add to the list if already there. (On POSIX systems, the
* LC_ALL element will likely be a repeat of the 0th element "",
- * but there's no harm done by doing it explicitly */
+ * but there's no harm done by doing it explicitly.
+ *
+ * Note that this tries the LC_ALL environment variable even on
+ * systems which have no LC_ALL locale setting. This may or may
+ * not have been originally intentional, but there's no real need
+ * to change the behavior. */
if (lc_all) {
for (j = 0; j < trial_locales_count; j++) {
if (strEQ(lc_all, trial_locales[j])) {
#ifdef USE_LOCALE_CTYPE
Safefree(curctype);
curctype = savepv(setlocale(LC_CTYPE, NULL));
+ DEBUG_LOCALE_INIT(LC_CTYPE, NULL, curctype);
#endif /* USE_LOCALE_CTYPE */
#ifdef USE_LOCALE_COLLATE
Safefree(curcoll);
curcoll = savepv(setlocale(LC_COLLATE, NULL));
+ DEBUG_LOCALE_INIT(LC_COLLATE, NULL, curcoll);
#endif /* USE_LOCALE_COLLATE */
#ifdef USE_LOCALE_NUMERIC
Safefree(curnum);
curnum = savepv(setlocale(LC_NUMERIC, NULL));
+ DEBUG_LOCALE_INIT(LC_NUMERIC, NULL, curnum);
#endif /* USE_LOCALE_NUMERIC */
}
Safefree(curnum);
#endif /* USE_LOCALE_NUMERIC */
+#ifdef __GLIBC__
+ Safefree(language);
+#endif
+
+ Safefree(lc_all);
+ Safefree(lang);
+
#else /* !USE_LOCALE */
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 */
- /* the first sizeof(collationix) bytes are used by sv_collxfrm(). */
- /* the +1 is for the terminating NUL. */
+ PERL_ARGS_ASSERT__MEM_COLLXFRM;
- xAlloc = sizeof(PL_collation_ix) + PL_collxfrm_base + (PL_collxfrm_mult * len) + 1;
+ /* 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 (s_strlen < len) { /* Only execute if there is an embedded NUL */
+ char * e = s + len;
+ char * sans_nuls;
+ STRLEN sans_nuls_len;
+ STRLEN sans_nuls_pos;
+ 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_min_char == '\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;
+
+ /* Create a 1 byte string of the current code point */
+ char cur_source[] = { (char) j, '\0' };
+
+ if (! try_non_controls && (PL_in_utf8_COLLATE_locale)
+ ? ! isCNTRL_L1(j)
+ : ! isCNTRL_LC(j))
+ {
+ continue;
+ }
+
+ /* 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_min_char = 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_min_char));
+
+ 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_min_char) && utf8) {
+ this_replacement_char[0] = UTF8_EIGHT_BIT_HI(PL_strxfrm_min_char);
+ this_replacement_char[1] = UTF8_EIGHT_BIT_LO(PL_strxfrm_min_char);
+ this_replacement_len = 2;
+ }
+ else {
+ this_replacement_char[0] = PL_strxfrm_min_char;
+ /* 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';
+ sans_nuls_pos = 0;
+
+ /* Replace each NUL with the lowest collating control. Loop until have
+ * exhausted all the NULs */
+ while (s + s_strlen < e) {
+ sans_nuls_pos = my_strlcat(sans_nuls + sans_nuls_pos,
+ s,
+ sans_nuls_len);
+
+ /* Do the actual replacement */
+ sans_nuls_pos = my_strlcat(sans_nuls + sans_nuls_pos,
+ 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 + sans_nuls_pos, 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;
+
+ /* Create a 1-char string of the current code point. */
+ char cur_source[] = { (char) j, '\0' };
+
+ /* 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;
+
+ 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_DOWNGRADEABLE_START(cur_char)) {
+ 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);
+ }
+ }
+
+ length_in_chars = (utf8)
+ ? utf8_length((U8 *) s, (U8 *) s + len)
+ : len;
+
+ /* 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);
- /* Embedded NULs are understood but silently skipped
- * because they make no sense in locale collation. */
+ /* 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) {
+ Size_t i;
+
+ print_collxfrm_input_and_return(s, s + len, xlen, utf8);
+ PerlIO_printf(Perl_debug_log, "Its xfrm is:");
+ for (i = COLLXFRM_HDR_LEN; i < *xlen + COLLXFRM_HDR_LEN; i++) {
+ PerlIO_printf(Perl_debug_log, " %02x", (U8) xbuf[i]);
+ }
+ PerlIO_printf(Perl_debug_log, "\n");
+ }
+#endif
+
+ /* 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)
+{
+ const char * t = s;
+ bool prev_was_printable = TRUE;
+ bool first_time = TRUE;
+
+ PERL_ARGS_ASSERT_PRINT_COLLXFRM_INPUT_AND_RETURN;
+
+ PerlIO_printf(Perl_debug_log, "_mem_collxfrm[%u]: returning ",
+ PL_collation_ix);
+ if (xlen) {
+ PerlIO_printf(Perl_debug_log, "%" UVuf, (UV) *xlen);
+ }
+ else {
+ PerlIO_printf(Perl_debug_log, "NULL");
+ }
+ PerlIO_printf(Perl_debug_log, " for locale '%s', string='",
+ PL_collation_name);
+
+ 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;
+ }
+
+ PerlIO_printf(Perl_debug_log, "'\n");
+}
+
+#endif /* #ifdef DEBUGGING */
+
#endif /* USE_LOCALE_COLLATE */
#ifdef USE_LOCALE
/* Returns TRUE if the current locale for 'category' is UTF-8; FALSE
* otherwise. 'category' may not be LC_ALL. If the platform doesn't have
* nl_langinfo(), nor MB_CUR_MAX, this employs a heuristic, which hence
- * could give the wrong result. It errs on the side of not being a UTF-8
- * locale. */
+ * could give the wrong result. The result will very likely be correct for
+ * languages that have commonly used non-ASCII characters, but for notably
+ * English, it comes down to if the locale's name ends in something like
+ * "UTF-8". It errs on the side of not being a UTF-8 locale. */
char *save_input_locale = NULL;
STRLEN final_pos;
if (category != LC_CTYPE) { /* These work only on LC_CTYPE */
/* Get the current LC_CTYPE locale */
- save_ctype_locale = stdize_locale(savepv(setlocale(LC_CTYPE, NULL)));
+ save_ctype_locale = setlocale(LC_CTYPE, NULL);
if (! save_ctype_locale) {
DEBUG_L(PerlIO_printf(Perl_debug_log,
"Could not find current locale for LC_CTYPE\n"));
goto cant_use_nllanginfo;
}
+ save_ctype_locale = stdize_locale(savepv(save_ctype_locale));
/* If LC_CTYPE and the desired category use the same locale, this
* means that finding the value for LC_CTYPE is the same as finding
# if defined(HAS_NL_LANGINFO) && defined(CODESET)
{
- char *codeset = savepv(nl_langinfo(CODESET));
+ char *codeset = nl_langinfo(CODESET);
if (codeset && strNE(codeset, "")) {
+ codeset = savepv(codeset);
/* If we switched LC_CTYPE, switch back */
if (save_ctype_locale) {
Safefree(save_input_locale);
return is_utf8;
}
- Safefree(codeset);
}
# endif
* result */
if (is_utf8) {
wchar_t wc;
- GCC_DIAG_IGNORE(-Wunused-result);
- (void) mbtowc(&wc, NULL, 0); /* Reset any shift state */
- GCC_DIAG_RESTORE;
+ PERL_UNUSED_RESULT(mbtowc(&wc, NULL, 0));/* Reset any shift state */
errno = 0;
if ((size_t)mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8))
!= strlen(HYPHEN_UTF8)
|| wc != (wchar_t) 0x2010)
{
is_utf8 = FALSE;
- DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", wc));
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "\thyphen=U+%x\n", (unsigned int)wc));
DEBUG_L(PerlIO_printf(Perl_debug_log,
"\treturn from mbtowc=%d; errno=%d; ?UTF8 locale=0\n",
mbtowc(&wc, HYPHEN_UTF8, strlen(HYPHEN_UTF8)), errno));
cant_use_nllanginfo:
-#endif /* HAS_NL_LANGINFO etc */
-
- /* nl_langinfo not available or failed somehow. Look at the locale name to
- * see if it matches qr/UTF -? 8 /ix */
-
- final_pos = strlen(save_input_locale) - 1;
- if (final_pos >= 3) {
- char *name = save_input_locale;
-
- /* Find next 'U' or 'u' and look from there */
- while ((name += strcspn(name, "Uu") + 1)
- <= save_input_locale + final_pos - 2)
- {
- if (toFOLD(*(name)) != 't'
- || toFOLD(*(name + 1)) != 'f')
- {
- continue;
- }
- name += 2;
- if (*(name) == '-') {
- if ((name > save_input_locale + final_pos - 1)) {
- break;
- }
- name++;
- }
- if (*(name) == '8') {
- Safefree(save_input_locale);
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Locale %s ends with UTF-8 in name\n",
- save_input_locale));
- return TRUE;
- }
- }
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Locale %s doesn't end with UTF-8 in name\n",
- save_input_locale));
- }
-
-#ifdef WIN32
- /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
- if (final_pos >= 4
- && *(save_input_locale + final_pos - 0) == '1'
- && *(save_input_locale + final_pos - 1) == '0'
- && *(save_input_locale + final_pos - 2) == '0'
- && *(save_input_locale + final_pos - 3) == '5'
- && *(save_input_locale + final_pos - 4) == '6')
- {
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Locale %s ends with 10056 in name, is UTF-8 locale\n",
- save_input_locale));
- Safefree(save_input_locale);
- return TRUE;
- }
-#endif
+#else /* nl_langinfo should work if available, so don't bother compiling this
+ fallback code. The final fallback of looking at the name is
+ compiled, and will be executed if nl_langinfo fails */
- /* Other common encodings are the ISO 8859 series, which aren't UTF-8 */
- if (instr(save_input_locale, "8859")) {
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Locale %s has 8859 in name, not UTF-8 locale\n",
- save_input_locale));
- Safefree(save_input_locale);
- return FALSE;
- }
+ /* nl_langinfo not available or failed somehow. Next try looking at the
+ * currency symbol to see if it disambiguates things. Often that will be
+ * in the native script, and if the symbol isn't in UTF-8, we know that the
+ * locale isn't. If it is non-ASCII UTF-8, we infer that the locale is
+ * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
+ * */
#ifdef HAS_LOCALECONV
-
# ifdef USE_LOCALE_MONETARY
-
- /* Here, there is nothing in the locale name to indicate whether the locale
- * is UTF-8 or not. This "name", the return of setlocale(), is actually
- * defined to be opaque, so we can't really rely on the absence of various
- * substrings in the name to indicate its UTF-8ness. Look at the locale's
- * currency symbol. Often that will be in the native script, and if the
- * symbol isn't in UTF-8, we know that the locale isn't. If it is
- * non-ASCII UTF-8, we infer that the locale is too.
- * To do this, like above for LC_CTYPE, we first set LC_MONETARY to the
- * locale of the desired category, if it isn't that locale already */
-
{
char *save_monetary_locale = NULL;
- bool illegal_utf8 = FALSE;
bool only_ascii = FALSE;
- const struct lconv* const lc = localeconv();
+ bool is_utf8 = FALSE;
+ struct lconv* lc;
+
+ /* Like above for LC_CTYPE, we first set LC_MONETARY to the locale of
+ * the desired category, if it isn't that locale already */
if (category != LC_MONETARY) {
- save_monetary_locale = stdize_locale(savepv(setlocale(LC_MONETARY,
- NULL)));
+ save_monetary_locale = setlocale(LC_MONETARY, NULL);
if (! save_monetary_locale) {
DEBUG_L(PerlIO_printf(Perl_debug_log,
"Could not find current locale for LC_MONETARY\n"));
goto cant_use_monetary;
}
+ save_monetary_locale = stdize_locale(savepv(save_monetary_locale));
- if (strNE(save_monetary_locale, save_input_locale)) {
- if (! setlocale(LC_MONETARY, save_input_locale)) {
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Could not change LC_MONETARY locale to %s\n",
- save_input_locale));
- Safefree(save_monetary_locale);
- goto cant_use_monetary;
- }
+ if (strEQ(save_monetary_locale, save_input_locale)) {
+ Safefree(save_monetary_locale);
+ save_monetary_locale = NULL;
+ }
+ else if (! setlocale(LC_MONETARY, save_input_locale)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Could not change LC_MONETARY locale to %s\n",
+ save_input_locale));
+ Safefree(save_monetary_locale);
+ goto cant_use_monetary;
}
}
/* Here the current LC_MONETARY is set to the locale of the category
* whose information is desired. */
- if (lc && lc->currency_symbol) {
- if (! is_utf8_string((U8 *) lc->currency_symbol, 0)) {
- DEBUG_L(PerlIO_printf(Perl_debug_log,
- "Currency symbol for %s is not legal UTF-8\n",
- save_input_locale));
- illegal_utf8 = TRUE;
- }
- else if (is_ascii_string((U8 *) lc->currency_symbol, 0)) {
- DEBUG_L(PerlIO_printf(Perl_debug_log, "Currency symbol for %s contains only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale));
- only_ascii = TRUE;
- }
+ lc = localeconv();
+ if (! lc
+ || ! lc->currency_symbol
+ || 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;
+ }
+ else {
+ is_utf8 = is_utf8_string((U8 *) lc->currency_symbol, 0);
}
/* If we changed it, restore LC_MONETARY to its original locale */
Safefree(save_monetary_locale);
}
- Safefree(save_input_locale);
-
- /* It isn't a UTF-8 locale if the symbol is not legal UTF-8; otherwise
- * assume the locale is UTF-8 if and only if the symbol is non-ascii
- * UTF-8. (We can't really tell if the locale is UTF-8 or not if the
- * symbol is just a '$', so we err on the side of it not being UTF-8)
- * */
- DEBUG_L(PerlIO_printf(Perl_debug_log, "\tis_utf8=%d\n", (illegal_utf8)
- ? FALSE
- : ! only_ascii));
- return (illegal_utf8)
- ? FALSE
- : ! only_ascii;
+ if (! only_ascii) {
+ /* It isn't a UTF-8 locale if the symbol is not legal UTF-8;
+ * otherwise assume the locale is UTF-8 if and only if the symbol
+ * is non-ascii UTF-8. */
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?Currency symbol for %s is UTF-8=%d\n",
+ save_input_locale, is_utf8));
+ Safefree(save_input_locale);
+ return is_utf8;
+ }
}
cant_use_monetary:
# endif /* USE_LOCALE_MONETARY */
#endif /* HAS_LOCALECONV */
-#if 0 && defined(HAS_STRERROR) && defined(USE_LOCALE_MESSAGES)
+#if defined(HAS_STRFTIME) && defined(USE_LOCALE_TIME)
+
+/* Still haven't found a non-ASCII string to disambiguate UTF-8 or not. Try
+ * the names of the months and weekdays, timezone, and am/pm indicator */
+ {
+ char *save_time_locale = NULL;
+ int hour = 10;
+ bool is_dst = FALSE;
+ int dom = 1;
+ int month = 0;
+ int i;
+ char * formatted_time;
+
+
+ /* Like above for LC_MONETARY, we set LC_TIME to the locale of the
+ * desired category, if it isn't that locale already */
+
+ if (category != LC_TIME) {
+
+ save_time_locale = setlocale(LC_TIME, NULL);
+ if (! save_time_locale) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Could not find current locale for LC_TIME\n"));
+ goto cant_use_time;
+ }
+ save_time_locale = stdize_locale(savepv(save_time_locale));
+
+ if (strEQ(save_time_locale, save_input_locale)) {
+ Safefree(save_time_locale);
+ save_time_locale = NULL;
+ }
+ else if (! setlocale(LC_TIME, save_input_locale)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Could not change LC_TIME locale to %s\n",
+ save_input_locale));
+ Safefree(save_time_locale);
+ goto cant_use_time;
+ }
+ }
+
+ /* Here the current LC_TIME is set to the locale of the category
+ * whose information is desired. Look at all the days of the week and
+ * month names, and the timezone and am/pm indicator for UTF-8 variant
+ * characters. The first such a one found will tell us if the locale
+ * is UTF-8 or not */
+
+ 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_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
+ * weekday. After we have gotten all weekdays, try the next
+ * month */
+ is_dst = ! is_dst;
+ hour = (hour + 12) % 24;
+ dom++;
+ if (i > 6) {
+ month++;
+ }
+ continue;
+ }
+
+ /* Here, we have a non-ASCII. Return TRUE is it is valid UTF8;
+ * false otherwise. But first, restore LC_TIME to its original
+ * locale if we changed it */
+ if (save_time_locale) {
+ setlocale(LC_TIME, save_time_locale);
+ Safefree(save_time_locale);
+ }
+
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?time-related strings for %s are UTF-8=%d\n",
+ save_input_locale,
+ is_utf8_string((U8 *) formatted_time, 0)));
+ Safefree(save_input_locale);
+ return is_utf8_string((U8 *) formatted_time, 0);
+ }
+
+ /* Falling off the end of the loop indicates all the names were just
+ * ASCII. Go on to the next test. If we changed it, restore LC_TIME
+ * to its original locale */
+ if (save_time_locale) {
+ setlocale(LC_TIME, save_time_locale);
+ Safefree(save_time_locale);
+ }
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "All time-related words for %s contain only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale));
+ }
+ cant_use_time:
+
+#endif
+
+#if 0 && defined(USE_LOCALE_MESSAGES) && defined(HAS_SYS_ERRLIST)
/* This code is ifdefd out because it was found to not be necessary in testing
- * on our dromedary test machine, which has over 700 locales. There, looking
- * at just the currency symbol gave essentially the same results as doing this
- * extra work. Executing this also caused segfaults in miniperl. I left it in
- * so as to avoid rewriting it if real-world experience indicates that
- * dromedary is an outlier. Essentially, instead of returning abpve if we
+ * on our dromedary test machine, which has over 700 locales. There, this
+ * added no value to looking at the currency symbol and the time strings. I
+ * left it in so as to avoid rewriting it if real-world experience indicates
+ * that dromedary is an outlier. Essentially, instead of returning abpve if we
* haven't found illegal utf8, we continue on and examine all the strerror()
* messages on the platform for utf8ness. If all are ASCII, we still don't
* know the answer; but otherwise we have a pretty good indication of the
- * utf8ness. The reason this doesn't necessarily help much is that the
- * messages may not have been translated into the locale. The currency symbol
- * is much more likely to have been translated. The code below would need to
- * be altered somewhat to just be a continuation of testing the currency
- * symbol. */
+ * utf8ness. The reason this doesn't help much is that the messages may not
+ * have been translated into the locale. The currency symbol and time strings
+ * are much more likely to have been translated. */
+ {
int e;
- unsigned int failures = 0, non_ascii = 0;
+ bool is_utf8 = FALSE;
+ bool non_ascii = FALSE;
char *save_messages_locale = NULL;
+ const char * errmsg = NULL;
- /* Like above for LC_CTYPE, we set LC_MESSAGES to the locale of the
- * desired category, if it isn't that locale already */
+ /* Like above, we set LC_MESSAGES to the locale of the desired
+ * category, if it isn't that locale already */
if (category != LC_MESSAGES) {
- save_messages_locale = stdize_locale(savepv(setlocale(LC_MESSAGES,
- NULL)));
+ save_messages_locale = setlocale(LC_MESSAGES, NULL);
if (! save_messages_locale) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Could not find current locale for LC_MESSAGES\n"));
goto cant_use_messages;
}
+ save_messages_locale = stdize_locale(savepv(save_messages_locale));
if (strEQ(save_messages_locale, save_input_locale)) {
- Safefree(save_input_locale);
+ Safefree(save_messages_locale);
+ save_messages_locale = NULL;
}
else if (! setlocale(LC_MESSAGES, save_input_locale)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Could not change LC_MESSAGES locale to %s\n",
+ save_input_locale));
Safefree(save_messages_locale);
goto cant_use_messages;
}
}
/* Here the current LC_MESSAGES is set to the locale of the category
- * whose information is desired. Look through all the messages */
+ * whose information is desired. Look through all the messages. We
+ * can't use Strerror() here because it may expand to code that
+ * segfaults in miniperl */
- for (e = 0;
-#ifdef HAS_SYS_ERRLIST
- e <= sys_nerr
-#endif
- ; e++)
- {
- const U8* const errmsg = (U8 *) Strerror(e) ;
- if (!errmsg)
- break;
- if (! is_utf8_string(errmsg, 0)) {
- failures++;
+ for (e = 0; e <= sys_nerr; e++) {
+ errno = 0;
+ errmsg = sys_errlist[e];
+ if (errno || !errmsg) {
break;
}
- else if (! is_ascii_string(errmsg, 0)) {
- non_ascii++;
+ errmsg = savepv(errmsg);
+ if (! is_utf8_invariant_string((U8 *) errmsg, 0)) {
+ non_ascii = TRUE;
+ is_utf8 = is_utf8_string((U8 *) errmsg, 0);
+ break;
}
}
+ Safefree(errmsg);
/* And, if we changed it, restore LC_MESSAGES to its original locale */
if (save_messages_locale) {
Safefree(save_messages_locale);
}
- /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
- * any non-ascii means it is one; otherwise we assume it isn't */
- return (failures) ? FALSE : non_ascii;
+ if (non_ascii) {
+ /* Any non-UTF-8 message means not a UTF-8 locale; if all are valid,
+ * any non-ascii means it is one; otherwise we assume it isn't */
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "\t?error messages for %s are UTF-8=%d\n",
+ save_input_locale,
+ is_utf8));
+ Safefree(save_input_locale);
+ return is_utf8;
+ }
+
+ DEBUG_L(PerlIO_printf(Perl_debug_log, "All error messages for %s contain only ASCII; can't use for determining if UTF-8 locale\n", save_input_locale));
}
cant_use_messages:
#endif
+#endif /* the code that is compiled when no nl_langinfo */
+
+#ifndef EBCDIC /* On os390, even if the name ends with "UTF-8', it isn't a
+ UTF-8 locale */
+ /* As a last resort, look at the locale name to see if it matches
+ * qr/UTF -? * 8 /ix, or some other common locale names. This "name", the
+ * return of setlocale(), is actually defined to be opaque, so we can't
+ * really rely on the absence of various substrings in the name to indicate
+ * its UTF-8ness, but if it has UTF8 in the name, it is extremely likely to
+ * be a UTF-8 locale. Similarly for the other common names */
+
+ final_pos = strlen(save_input_locale) - 1;
+ if (final_pos >= 3) {
+ char *name = save_input_locale;
+
+ /* Find next 'U' or 'u' and look from there */
+ while ((name += strcspn(name, "Uu") + 1)
+ <= save_input_locale + final_pos - 2)
+ {
+ if (!isALPHA_FOLD_NE(*name, 't')
+ || isALPHA_FOLD_NE(*(name + 1), 'f'))
+ {
+ continue;
+ }
+ name += 2;
+ if (*(name) == '-') {
+ if ((name > save_input_locale + final_pos - 1)) {
+ break;
+ }
+ name++;
+ }
+ if (*(name) == '8') {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Locale %s ends with UTF-8 in name\n",
+ save_input_locale));
+ Safefree(save_input_locale);
+ return TRUE;
+ }
+ }
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Locale %s doesn't end with UTF-8 in name\n",
+ save_input_locale));
+ }
+#endif
+
+#ifdef WIN32
+ /* http://msdn.microsoft.com/en-us/library/windows/desktop/dd317756.aspx */
+ if (final_pos >= 4
+ && *(save_input_locale + final_pos - 0) == '1'
+ && *(save_input_locale + final_pos - 1) == '0'
+ && *(save_input_locale + final_pos - 2) == '0'
+ && *(save_input_locale + final_pos - 3) == '5'
+ && *(save_input_locale + final_pos - 4) == '6')
+ {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Locale %s ends with 10056 in name, is UTF-8 locale\n",
+ save_input_locale));
+ Safefree(save_input_locale);
+ return TRUE;
+ }
+#endif
+
+ /* Other common encodings are the ISO 8859 series, which aren't UTF-8. But
+ * since we are about to return FALSE anyway, there is no point in doing
+ * this extra work */
+#if 0
+ if (instr(save_input_locale, "8859")) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "Locale %s has 8859 in name, not UTF-8 locale\n",
+ save_input_locale));
+ Safefree(save_input_locale);
+ return FALSE;
+ }
+#endif
+
DEBUG_L(PerlIO_printf(Perl_debug_log,
"Assuming locale %s is not a UTF-8 locale\n",
save_input_locale));
}
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;
- /* Uses C locale for the error text unless within scope of 'use locale' for
- * LC_MESSAGES */
+#ifdef USE_LOCALE_MESSAGES /* If platform doesn't have messages category, we
+ don't do any switching to the C locale; we just
+ use whatever strerror() returns */
+ const bool within_locale_scope = IN_LC(LC_MESSAGES);
+
+ dVAR;
+
+# ifdef USE_THREAD_SAFE_LOCALE
+ locale_t save_locale;
+# else
+ char * save_locale;
+ 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
+
+ if (! within_locale_scope) {
+ errno = 0;
+
+# ifdef USE_THREAD_SAFE_LOCALE /* Use the thread-safe locale functions */
+
+ save_locale = uselocale(PL_C_locale_obj);
+ if (! save_locale) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "uselocale failed, errno=%d\n", errno));
+ }
+
+# 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 */
+
+#endif
+
+ errstr = Strerror(errnum);
+ if (errstr) {
+ errstr = savepv(errstr);
+ SAVEFREEPV(errstr);
+ }
#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;
- /* 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_THREAD_SAFE_LOCALE
- /* This points to the static space in Strerror, with all its
- * limitations */
- errstr = Strerror(errnum);
+ if (save_locale && ! uselocale(save_locale)) {
+ DEBUG_L(PerlIO_printf(Perl_debug_log,
+ "uselocale restore failed, errno=%d\n", errno));
+ }
+ }
- setlocale(LC_MESSAGES, save_locale);
+# else
+
+ 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
- return Strerror(errnum);
+ return errstr;
}
/*
- * Local variables:
- * c-indentation-style: bsd
- * c-basic-offset: 4
- * indent-tabs-mode: nil
- * End:
- *
+
+=head1 Locale-related functions and macros
+
+=for apidoc sync_locale
+
+Changing the program's locale should be avoided by XS code. Nevertheless,
+certain non-Perl libraries called from XS, such as C<Gtk> do so. When this
+happens, Perl needs to be told that the locale has changed. Use this function
+to do so, before returning to Perl.
+
+=cut
+*/
+
+void
+Perl_sync_locale(pTHX)
+{
+
+#ifdef USE_LOCALE_CTYPE
+ new_ctype(setlocale(LC_CTYPE, NULL));
+#endif /* USE_LOCALE_CTYPE */
+
+#ifdef USE_LOCALE_COLLATE
+ new_collate(setlocale(LC_COLLATE, NULL));
+#endif
+
+#ifdef USE_LOCALE_NUMERIC
+ set_numeric_local(); /* Switch from "C" to underlying LC_NUMERIC */
+ new_numeric(setlocale(LC_NUMERIC, NULL));
+#endif /* USE_LOCALE_NUMERIC */
+
+}
+
+#if defined(DEBUGGING) && defined(USE_LOCALE)
+
+char *
+Perl__setlocale_debug_string(const int category, /* category number,
+ like LC_ALL */
+ const char* const locale, /* locale name */
+
+ /* return value from setlocale() when attempting to
+ * set 'category' to 'locale' */
+ const char* const retval)
+{
+ /* 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. */
+
+ /* 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:
+ my_snprintf(ret, sizeof(ret), "%s? %d", ret, category);
+ break;
+# ifdef LC_ALL
+ case LC_ALL:
+ my_strlcat(ret, "LC_ALL", sizeof(ret));
+ break;
+# endif
+# ifdef LC_CTYPE
+ case LC_CTYPE:
+ my_strlcat(ret, "LC_CTYPE", sizeof(ret));
+ break;
+# endif
+# ifdef LC_NUMERIC
+ case LC_NUMERIC:
+ my_strlcat(ret, "LC_NUMERIC", sizeof(ret));
+ break;
+# endif
+# ifdef LC_COLLATE
+ case LC_COLLATE:
+ my_strlcat(ret, "LC_COLLATE", sizeof(ret));
+ break;
+# endif
+# ifdef LC_TIME
+ case LC_TIME:
+ my_strlcat(ret, "LC_TIME", sizeof(ret));
+ break;
+# endif
+# ifdef LC_MONETARY
+ case LC_MONETARY:
+ my_strlcat(ret, "LC_MONETARY", sizeof(ret));
+ break;
+# endif
+# ifdef LC_MESSAGES
+ case LC_MESSAGES:
+ my_strlcat(ret, "LC_MESSAGES", sizeof(ret));
+ break;
+# endif
+ }
+
+ my_strlcat(ret, ", ", sizeof(ret));
+
+ if (locale) {
+ my_strlcat(ret, "\"", sizeof(ret));
+ my_strlcat(ret, locale, sizeof(ret));
+ my_strlcat(ret, "\"", sizeof(ret));
+ }
+ else {
+ my_strlcat(ret, "NULL", sizeof(ret));
+ }
+
+ my_strlcat(ret, ") returned ", sizeof(ret));
+
+ if (retval) {
+ my_strlcat(ret, "\"", sizeof(ret));
+ my_strlcat(ret, retval, sizeof(ret));
+ my_strlcat(ret, "\"", sizeof(ret));
+ }
+ else {
+ my_strlcat(ret, "NULL", sizeof(ret));
+ }
+
+ assert(strlen(ret) < sizeof(ret));
+
+ return ret;
+}
+
+#endif
+
+
+/*
* ex: set ts=8 sts=4 sw=4 et:
*/
https://perl5.git.perl.org / perl5.git / blobdiff