X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2Flocale.t;h=e9ba3fad2160140ad3aac1d574e24a658c2b62bd;hb=0bd43963443f227c8bd1734af34fc6e6a7d2d54e;hp=19fba597c5450d4f2404ac420a951f4542f78ddd;hpb=b695f709e8a342e35e482b0437eb6cdacdc58b6b;p=p5sagit%2Fp5-mst-13.2.git

diff --git a/lib/locale.t b/lib/locale.t
index 19fba59..e9ba3fa 100644
--- a/lib/locale.t
+++ b/lib/locale.t
@@ -311,9 +311,9 @@ if ($^O eq 'os390') {
 sub in_utf8 () { $^H & 0x08 }
 
 if (in_utf8) {
-    require "locale/utf8";
+    require "lib/locale/utf8";
 } else {
-    require "locale/latin1";
+    require "lib/locale/latin1";
 }
 
 my @Locale;
@@ -719,19 +719,50 @@ foreach $Locale (@Locale) {
             $utf8skip{117}++;
 	} else {
 	    use locale;
-	    use locale;
-	    no utf8; # so that the native 8-bit characters work
+	    no utf8;
+            my $re = qr/[\[\(\{\*\+\?\|\^\$\\]/;
 
 	    my @f = ();
 	    foreach my $x (keys %UPPER) {
 		my $y = lc $x;
 		next unless uc $y eq $x;
-		push @f, $x unless $x =~ /$y/i && $y =~ /$x/i;
+		print "# UPPER $x lc $y ",
+			$x =~ /$y/i ? 1 : 0, " ",
+			$y =~ /$x/i ? 1 : 0, "\n" if 0;
+		# If $x and $y contain regular expression characters
+		# AND THEY lowercase (/i) to regular expression characters,
+		# regcomp() will be mightily confused.  No, the \Q doesn't
+		# help here (maybe regex engine internal lowercasing
+		# is done after the \Q?)  An example of this happening is
+		# the bg_BG (Bulgarian) locale under EBCDIC (OS/390 USS):
+		# the chr(173) (the "[") is the lowercase of the chr(235).
+		# Similarly losing EBCDIC locales include cs_cz, cs_CZ,
+		# el_gr, el_GR, en_us.IBM-037 (!), en_US.IBM-037 (!),
+		# et_ee, et_EE, hr_hr, hr_HR, hu_hu, hu_HU, lt_LT,
+		# mk_mk, mk_MK, nl_nl.IBM-037, nl_NL.IBM-037,
+		# pl_pl, pl_PL, ro_ro, ro_RO, ru_ru, ru_RU,
+		# sk_sk, sk_SK, sl_si, sl_SI, tr_tr, tr_TR.
+		if ($x =~ $re || $y =~ $re) {
+		    print "# Regex characters in '$x' or '$y', skipping test 117 for locale '$Locale'\n";
+		    next;
+		}
+		# With utf8 both will fail since the locale concept
+		# of upper/lower does not work well in Unicode.
+		push @f, $x unless $x =~ /$y/i == $y =~ /$x/i;
 	    }
 	    foreach my $x (keys %lower) {
 		my $y = uc $x;
 		next unless lc $y eq $x;
-		push @f, $x unless $x =~ /$y/i && $y =~ /$x/i;
+		print "# lower $x uc $y ",
+			$x =~ /$y/i ? 1 : 0, " ",
+			$y =~ /$x/i ? 1 : 0, "\n" if 0;
+		if ($x =~ $re || $y =~ $re) { # See above.
+		    print "# Regex characters in '$x' or '$y', skipping test 117 for locale '$Locale'\n";
+		    next;
+		}
+		# With utf8 both will fail since the locale concept
+		# of upper/lower does not work well in Unicode.
+		push @f, $x unless $x =~ /$y/i == $y =~ /$x/i;
 	    }
 	    tryneoalpha($Locale, 117, @f == 0);
 	    if (@f) {