Bug #13601: Wrong int type for bit

  The wrong value was being reported as the field_length for BIT
  fields, resulting in confusion for at least Connector/J. The
  field_length is now always the number of bits in the field, as
  it should be.


mysql-test/r/type_bit.result:
  Add new results
mysql-test/r/type_bit_innodb.result:
  Add new results
mysql-test/t/type_bit.test:
  Add new regression test
mysql-test/t/type_bit_innodb.test:
  Add new regression test
sql/field.cc:
  Fix Field_bit->field_length to actually report the display width, and
  store the bytes stored in the rec in the new bytes_in_rec member.
sql/field.h:
  Fix Field_bit::field_length to store the correct value, adding
  Field_bit::bytes_in_rec to remember the number of bytes used for
  storing the value. Remove Field_bit_as_char::create_length, as it
  is now redundant.
sql/ha_ndbcluster.cc:
  Handle field_length of Field_bit actually being the display width (# of bits).
sql/key.cc:
  Fix inappropriate use of field->field_length for BIT field.

1 files changed, 2 insertions, 1 deletions

diff --git a/sql/key.cc b/sql/key.cc
index 9d86095f33e..75161e4f616 100644
--- a/sql/key.cc
+++ b/sql/key.cc
@@ -192,7 +192,8 @@ void key_restore(byte *to_record, byte *from_key, KEY *key_info,
       Field_bit *field= (Field_bit *) (key_part->field);
       if (field->bit_len)
       {
-        uchar bits= *(from_key + key_part->length - field->field_length -1);
+        uchar bits= *(from_key + key_part->length -
+                      field->pack_length_in_rec() - 1);
         set_rec_bits(bits, to_record + key_part->null_offset +
                      (key_part->null_bit == 128),
                      field->bit_ofs, field->bit_len);