why UPDATE requires share lock(S) for table joined (3 replies)

for example
update t inner join lktest on lktest.c3=t.x set t.y=t.y where lktest.c3=2;

in this case , mysql always places share lock on the rows(c3=2) of lktest table no matter what transaction level it is


I am wondering why the S is needed , why mysql doesn't take advantage of MVCC to avoid lock on lktest ?

many thanks .