Fixed bug #12382.
  INSERT statement effectively changed thd->set_query_id to 0,
  while SELECT statement changed it to 0. As a result
  the insert_fields function that expanded '*' was called
  with different values of thd->set_query_id for the query
  SELECT * FROM view depending on whether it was run after
  an INSERT or after a SELECT statement. This was corrected
  by restoring the old value of thd->set_query_id when
  returning from the function setup_fields where possible
  reset could occur.
  If the value of thd->set_query_id == 0 then the fields
  substituted instead of '*' were not registered as used
  for bitmaps used_keys. This caused selection of an invalid
  execution plan for the query SELECT * from <view>.
view.result, view.test:
  Added a test case for bug #12382.


mysql-test/t/view.test:
  Added a test case for bug #12382.
mysql-test/r/view.result:
  Added a test case for bug #12382.
sql/sql_base.cc:
  Fixed bug #12382.
  INSERT statement effectively changed thd->set_query_id to 0,
  while SELECT statement changed it to 0. As a result
  the insert_fields function that expanded '*' was called
  with different values of thd->set_query_id for the query
  SELECT * FROM view depending on whether it was run after
  an INSERT or after a SELECT statement. This was corrected
  by restoring the old value of thd->set_query_id when
  returning from the function setup_fields where possible
  reset could occur.
  If the value of thd->set_query_id == 0 then the fields
  substituted instead of '*' were not registered as used
  for bitmaps used_keys. This caused selection of an invalid
  execution plan for the query SELECT * from <view>.