Allow sharing middlesubstring if the terminator is present

Sharing middle substrings have been behind a compilation flags for
a very long time, and it's unclear if we'll ever make it the default.

However, we can still share middle substrings without breaking the
zero-terminated contract if the source string happens to have the
necessary NUL characters at that specific offset.

This is the case for some file formats such as FlatBuffers, BSON keys
and a few others.

Author: byroot

string.c

@@ -198,15 +198,30 @@ VALUE rb_cSymbol;
 
 #define STR_ENC_GET(str) get_encoding(str)
 
+static inline bool
+zero_filled(const char *s, int n)
+{
+    for (; n > 0; --n) {
+        if (*s++) return false;
+    }
+    return true;
+}
+
 #if !defined SHARABLE_MIDDLE_SUBSTRING
 # define SHARABLE_MIDDLE_SUBSTRING 0
 #endif
-#if !SHARABLE_MIDDLE_SUBSTRING
-#define SHARABLE_SUBSTRING_P(beg, len, end) ((beg) + (len) == (end))
+
+static inline bool
+SHARABLE_SUBSTRING_P(VALUE str, long beg, long len)
+{
+#if SHARABLE_MIDDLE_SUBSTRING
+    return true;
 #else
-#define SHARABLE_SUBSTRING_P(beg, len, end) 1
+    long end = beg + len;
+    long source_len = RSTRING_LEN(str);
+    return end == source_len || zero_filled(RSTRING_PTR(str) + end, TERM_LEN(str));
 #endif
-
+}
 
 static inline long
 str_embed_capa(VALUE str)
@@ -2810,15 +2825,6 @@ rb_string_value_ptr(volatile VALUE *ptr)
     return RSTRING_PTR(str);
 }
 
-static int
-zero_filled(const char *s, int n)
-{
-    for (; n > 0; --n) {
-        if (*s++) return 0;
-    }
-    return 1;
-}
-
 static const char *
 str_null_char(const char *s, long len, const int minlen, rb_encoding *enc)
 {
@@ -3138,7 +3144,7 @@ str_subseq(VALUE str, long beg, long len)
     RUBY_ASSERT(beg+len <= RSTRING_LEN(str));
 
     const int termlen = TERM_LEN(str);
-    if (!SHARABLE_SUBSTRING_P(beg, len, RSTRING_LEN(str))) {
+    if (!SHARABLE_SUBSTRING_P(str, beg, len)) {
         str2 = rb_enc_str_new(RSTRING_PTR(str) + beg, len, rb_str_enc_get(str));
         if (ENC_CODERANGE(str) == ENC_CODERANGE_7BIT) {
             ENC_CODERANGE_SET(str2, ENC_CODERANGE_7BIT);

test/ruby/test_string.rb

@@ -3461,6 +3461,27 @@ def test_byteslice
     assert_equal(false, ("\u3042"*10).byteslice(0, 20).valid_encoding?, bug7954)
   end
 
+  def test_shared_middle_string_terminator
+    ten = "0123456789"
+    hundred = ten * 10
+    str = "#{hundred}\0#{hundred}".freeze
+
+    require 'objspace'
+
+    substr = str.byteslice(0, hundred.bytesize)
+    assert_equal hundred, substr
+    assert_includes ObjectSpace.dump(substr), ' "shared":true,'
+
+    # Larger terminator
+    substr.force_encoding(Encoding::UTF_16BE)
+    assert_equal hundred.dup.force_encoding(Encoding::UTF_16BE), substr
+    refute_includes ObjectSpace.dump(substr), ' "shared":true,'
+
+    substr = str.byteslice(0, hundred.bytesize + 1)
+    assert_equal hundred + "\0", substr
+    refute_includes ObjectSpace.dump(substr), ' "shared":true,'
+  end
+
   def test_unknown_string_option
     str = nil
     assert_nothing_raised(SyntaxError) do