Comment #4 on issue 127 by [email protected]: incr/decr operations are not
thread safe.
http://code.google.com/p/memcached/issues/detail?id=127
Err, I figured it out, but it's still slower and not entirely correct so
we'll have
to figure it more later.
do_add_delta has the global cache_lock, so it should've never had a race
condition?
But the item that's passed into it is fetched from outside of the big lock.
So an
incremented item could end up being junked by another thread. Added a
little bit to
trond's patch to re-fetch the item once inside do_add_delta and the test
came back
100% every time. It's still 7% slower, and I'm not sure if trond had that
fixed
elsewhere in the engine branch. I couldn't find it on a cursory look.
To fix it more, do_add_delta needs to be passed in a key,keylen tuple
instead of an
item and return a NOT_FOUND if the internal fetch doesn't find the item.
But again
we'll revisit it later :/
--
You received this message because you are listed in the owner
or CC fields of this issue, or because you starred this issue.
You may adjust your issue notification preferences at:
http://code.google.com/hosting/settings
--
To unsubscribe, reply using "remove me" as the subject.
