the lock is in the bss. maybe the wrong page gets accessed after fork.

Charles Forsyth <charles.forsyth@gmail.com> wrote:

I've got one of those that was fine last time I tried it. I'll try it in the morning.
I wonder whether the change of lock to use semacquire instead of tas doesn't work well on the (that) ARM.
It seems a strange coincidence that it always fails there.