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_invariant_string((U8 *) lc->decimal_point, 0)
&& is_utf8_string((U8 *) lc->decimal_point, 0)
&& _is_cur_LC_category_utf8(LC_NUMERIC))
{
* 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
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 ];
+
+ bool check_for_problems = ckWARN_d(WARN_LOCALE); /* No warnings means
+ no check */
+ 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 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, 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)) {
+
+ /* 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
* 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, "")) {
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
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;
#ifdef WIN32
/* In some systems you can find out the system default locale
}
if (!setlocale_failure) {
# ifdef USE_LOCALE_CTYPE
- Safefree(curctype);
if (! (curctype =
my_setlocale(LC_CTYPE,
(!done && (lang || PerlEnv_getenv("LC_CTYPE")))
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")))
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")))
#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++) {
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 */
}
/* 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
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 */
/* 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;
* 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));
* 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. */
+ * too, as the odds of a non-UTF8 string being valid UTF-8 are quite small
+ * */
#ifdef HAS_LOCALECONV
# ifdef USE_LOCALE_MONETARY
lc = localeconv();
if (! lc
|| ! lc->currency_symbol
- || is_ascii_string((U8 *) lc->currency_symbol, 0))
+ || is_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;
/* 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 non-ASCII
+ * 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_ascii_string((U8 *) formatted_time, 0)) {
+ if (! formatted_time || is_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
#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_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 /* 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
while ((name += strcspn(name, "Uu") + 1)
<= save_input_locale + final_pos - 2)
{
- if (toFOLD(*(name)) != 't'
- || toFOLD(*(name + 1)) != 'f')
+ if (!isALPHA_FOLD_NE(*name, 't')
+ || isALPHA_FOLD_NE(*(name + 1), 'f'))
{
continue;
}
"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 */
}
/*
- * 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 */
+
+}
+
+
+
+/*
* ex: set ts=8 sts=4 sw=4 et:
*/