|
From: Julian S. <js...@ac...> - 2010-05-20 08:38:08
|
> > > I have a test case which will not deadlock, but helgrind reports "lock > > > order "0x600DA0 before 0x600DE0" violated". > > > The code does indeed have lock order inversion, but due to reader locks > > > no deadlock is possible. > My question is whether an ideal deadlock detector should or should not > report such lock order violations. > (I don't know the answer!) /me no understand .. if you are sure that this example can never deadlock, then an ideal lock-order-error detector should not report any error, because there is no error. (Or did I miss something?) J |