From: Pavel Vasiliev (pavel_at_[hidden])
Date: 2003-02-11 13:04:00
Alexander Terekhov wrote:
> Pavel Vasiliev wrote:
> [...]
>> Thread A, in release_strong:
>> atomic_decrement(&strong_count) == 0,
>> enter strong_refs_lost(),
>> lock
> ACK.
>> Thread B, in acquire_strong_from_weak: ....
> NAK.
> Thread B, in release_weak:
> atomic_decrement(&weak_count) == 0,
>> Thread A, in strong_refs_lost:
>> see strong_count == 0,
> acquire_weak() --> weak_count == 1,
> ...
> enter release_weak(),
> atomic_decrement(&weak_count) == 0,
> Thread B, enter weak_refs_lost():
> ...
> destruct_self()
> Thread A, enter weak_refs_lost():
> WHACK!
Ok, indeed "WHACK!" :-)
But my idea with initial weak_count == 0 in any case was not good.
Please see my previous reply that fixes it, now reading not
"aesthetic feelings" but "ordinary correctness". (Hope to have this
correctness now.)
Pavel