* encoding.c (rb_enc_compatible): change the rule for empty strings:

remove the special treatment of the US-ASCII encoded empty string.
  Now Encoding.compatible? usually respect the encoding of the
  receiver.

git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@30877 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

ChangeLog

@@ -1,3 +1,10 @@
+Tue Feb 15 15:43:29 2011  NARUSE, Yui  <naruse@ruby-lang.org>
+
+	* encoding.c (rb_enc_compatible): change the rule for empty strings:
+	  remove the special treatment of the US-ASCII encoded empty string.
+	  Now Encoding.compatible? usually respect the encoding of the
+	  receiver.
+
 Tue Feb 15 15:39:37 2011  NARUSE, Yui  <naruse@ruby-lang.org>
 
 	* string.c (rb_enc_cr_str_buf_cat): remove special treatment of

encoding.c

@@ -754,10 +754,10 @@ rb_enc_compatible(VALUE str1, VALUE str2)
     enc1 = rb_enc_from_index(idx1);
     enc2 = rb_enc_from_index(idx2);
 
-    if (TYPE(str2) == T_STRING && RSTRING_LEN(str2) == 0)
-	return (idx1 == ENCINDEX_US_ASCII && rb_enc_asciicompat(enc2)) ? enc2 : enc1;
-    if (TYPE(str1) == T_STRING && RSTRING_LEN(str1) == 0)
-	return (idx2 == ENCINDEX_US_ASCII && rb_enc_asciicompat(enc1)) ? enc1 : enc2;
+    if (BUILTIN_TYPE(str2) == T_STRING && RSTRING_LEN(str2) == 0)
+	return enc1;
+    if (BUILTIN_TYPE(str1) == T_STRING && RSTRING_LEN(str1) == 0)
+	return (rb_enc_asciicompat(enc1) && rb_enc_str_asciionly_p(str2)) ? enc1 : enc2;
     if (!rb_enc_asciicompat(enc1) || !rb_enc_asciicompat(enc2)) {
 	return 0;
     }

test/ruby/test_m17n.rb

@@ -811,15 +811,15 @@ class TestM17N < Test::Unit::TestCase
   end
 
   def test_sprintf_p
-    enc = "".inspect.encoding
-    asc = Encoding::US_ASCII
     Encoding.list.each do |e|
       format = "%p".force_encoding(e)
       ['', 'a', "\xC2\xA1", "\x00"].each do |s|
         s.force_encoding(e)
-        assert_strenc(s.inspect, e.ascii_compatible? && enc == asc ? e : enc, format % s)
+        enc = (''.force_encoding(e) + s.inspect).encoding
+        assert_strenc(s.inspect, enc, format % s)
       end
       s = "\xC2\xA1".force_encoding(e)
+      enc = ('' + s.inspect).encoding
       assert_strenc('%10s' % s.inspect, enc, "%10p" % s)
     end
   end
@@ -1086,7 +1086,6 @@ class TestM17N < Test::Unit::TestCase
     assert_equal(false, s.ascii_only?, "[ruby-core:14566] reported by Sam Ruby")
 
     s = "abc".force_encoding(Encoding::ASCII_8BIT)
-    t = s.gsub(/b/, "\xa1\xa1".force_encoding("euc-jp"))
     assert_equal(Encoding::ASCII_8BIT, s.encoding)
 
     assert_raise(Encoding::CompatibilityError) {
@@ -1400,7 +1399,7 @@ class TestM17N < Test::Unit::TestCase
     Encoding.list.each do |enc|
       next if enc.dummy?
       strs = strings.map {|s| s.encode(enc)} rescue next
-      yield *strs
+      yield(*strs)
     end
   end
 end

test/ruby/test_m17n_comb.rb

@@ -286,7 +286,7 @@ class TestM17NComb < Test::Unit::TestCase
       assert_strenc(a(s), s.encoding, "%s".force_encoding(s.encoding) % s)
       if !s.empty? # xxx
         t = enccall(a("%s"), :%, s)
-        assert_strenc(a(s), s.encoding, t)
+        assert_strenc(a(s), (a('')+s).encoding, t)
       end
     }
   end
@@ -633,13 +633,9 @@ class TestM17NComb < Test::Unit::TestCase
   def test_str_casecmp
     combination(STRINGS, STRINGS) {|s1, s2|
       #puts "#{encdump(s1)}.casecmp(#{encdump(s2)})"
-      begin
-        r = s1.casecmp(s2)
-      rescue ArgumentError
-        assert(!s1.valid_encoding? || !s2.valid_encoding?)
-        next
-      end
-      #assert_equal(s1.upcase <=> s2.upcase, r)
+      next unless s1.valid_encoding? && s2.valid_encoding? && Encoding.compatible?(s1, s2)
+      r = s1.casecmp(s2)
+      assert_equal(s1.upcase <=> s2.upcase, r)
     }
   end
 
@@ -1031,7 +1027,7 @@ class TestM17NComb < Test::Unit::TestCase
         t1.insert(nth, s2)
         slen = s2.length
         assert_equal(t1[nth-slen+1,slen], s2, "t=#{encdump s1}; t.insert(#{nth},#{encdump s2}); t")
-      rescue Encoding::CompatibilityError, IndexError => e
+      rescue Encoding::CompatibilityError, IndexError
       end
     }
   end