Thorsten,
I guess we should let it use __sync_val_compare_and_swap instead then?That one’s interesting: it makes an idle-spin instead of busy-spin:
....
Is there another thread supposed to join in and break the deadlock? Lock initialized wrong?Is there something wrong with threads (whatever façette of it)?
Just a stray thought ... Cheers, Michael(hoping this will reach you via one of the lists, as you appear to refuse mail from me otherwise)