ObjectSpace.dump: Include string coderange

I suspect that some shared pages are invalidated because
some static string don't have their coderange set eagerly.

So the first time they are scanned, the entire memory page is
invalidated.

Being able to see the coderange in `ObjectSpace` would help debug
this.

And in addition `dump` currently call `is_broken_string()`  and `is_ascii_string()`
which both end up scanning the string and assigning coderange. I think it's
undesirable as `dump` should be read only.

2 files changed, 52 insertions, 5 deletions

diff --git a/ext/objspace/objspace_dump.c b/ext/objspace/objspace_dump.c
index b570acbd95..18e052f210 100644
--- a/ext/objspace/objspace_dump.c
+++ b/ext/objspace/objspace_dump.c
@@ -313,6 +313,16 @@ reachable_object_i(VALUE ref, void *data)
     dc->cur_obj_references++;
 }
 
+bool
+dump_string_ascii_only(const char *str, long size) {
+    for (long i = 0; i < size; i++) {
+        if (str[i] & 0x80) {
+            return false;
+        }
+    }
+    return true;
+}
+
 static void
 dump_append_string_content(struct dump_config *dc, VALUE obj)
 {
@@ -323,9 +333,17 @@ dump_append_string_content(struct dump_config *dc, VALUE obj)
         dump_append_sizet(dc, rb_str_capacity(obj));
     }
 
-    if (is_ascii_string(obj)) {
-        dump_append(dc, ", \"value\":");
-        dump_append_string_value(dc, obj);
+    if (RSTRING_LEN(obj) && rb_enc_asciicompat(rb_enc_from_index(ENCODING_GET(obj)))) {
+        int cr = ENC_CODERANGE(obj);
+        if (cr == RUBY_ENC_CODERANGE_UNKNOWN) {
+            if (dump_string_ascii_only(RSTRING_PTR(obj), RSTRING_LEN(obj))) {
+                cr = RUBY_ENC_CODERANGE_7BIT;
+            }
+        }
+        if (cr == RUBY_ENC_CODERANGE_7BIT) {
+            dump_append(dc, ", \"value\":");
+            dump_append_string_value(dc, obj);
+        }
     }
 }
 
@@ -389,8 +407,6 @@ dump_object(VALUE obj, struct dump_config *dc)
       case T_STRING:
         if (STR_EMBED_P(obj))
             dump_append(dc, ", \"embedded\":true");
-        if (is_broken_string(obj))
-            dump_append(dc, ", \"broken\":true");
         if (FL_TEST(obj, RSTRING_FSTR))
             dump_append(dc, ", \"fstring\":true");
         if (STR_SHARED_P(obj))
@@ -403,6 +419,27 @@ dump_object(VALUE obj, struct dump_config *dc)
             dump_append(dc, rb_enc_name(rb_enc_from_index(ENCODING_GET(obj))));
             dump_append(dc, "\"");
         }
+
+        dump_append(dc, ", \"coderange\":\"");
+        switch (RB_ENC_CODERANGE(obj)) {
+          case RUBY_ENC_CODERANGE_UNKNOWN:
+            dump_append(dc, "unknown");
+            break;
+          case RUBY_ENC_CODERANGE_7BIT:
+            dump_append(dc, "7bit");
+            break;
+          case RUBY_ENC_CODERANGE_VALID:
+            dump_append(dc, "valid");
+            break;
+          case RUBY_ENC_CODERANGE_BROKEN:
+            dump_append(dc, "broken");
+            break;
+        }
+        dump_append(dc, "\"");
+
+        if (RB_ENC_CODERANGE(obj) == RUBY_ENC_CODERANGE_BROKEN)
+            dump_append(dc, ", \"broken\":true");
+
         break;
 
       case T_HASH:
diff --git a/test/objspace/test_objspace.rb b/test/objspace/test_objspace.rb
index ed9c998597..1392447e4f 100644
--- a/test/objspace/test_objspace.rb
+++ b/test/objspace/test_objspace.rb
@@ -486,6 +486,16 @@ class TestObjSpace < Test::Unit::TestCase
     end
   end
 
+  def test_dump_string_coderange
+    assert_includes ObjectSpace.dump("TEST STRING"), '"coderange":"7bit"'
+    unknown = "TEST STRING".dup.force_encoding(Encoding::BINARY)
+    2.times do # ensure that dumping the string doesn't mutate it
+      assert_includes ObjectSpace.dump(unknown), '"coderange":"unknown"'
+    end
+    assert_includes ObjectSpace.dump("Fée"), '"coderange":"valid"'
+    assert_includes ObjectSpace.dump("\xFF"), '"coderange":"broken"'
+  end
+
   def test_dump_escapes_method_name
     method_name = "foo\"bar"
     klass = Class.new do