guesses for 1.2.6 release?

dormando dormando at
Fri May 30 01:21:12 UTC 2008

So far as I know there are no test cases.

We've found the refcount leak to be most likely based around spurious 
malloc failures (which aren't handled correctly and leak references).

the assoc bug we've not been able to reproduce, but a solid workaround 
exists so far.

Dustin Sallings wrote:
>   Is there a case that can demonstrate either of these bugs?
> -- Dustin Sallings (mobile)
> On May 29, 2008, at 16:37, dormando <dormando at> wrote:
>> ASAP - next week or two.
>> Not blocking on the binary protocol... mostly on me I guess. I'll kick 
>> out -rc1 this weekend and no later.
>> Miguel DeAvila wrote:
>>> AFAIK, there are two known bugs with 1.2.5,
>>>        - reference counts get screwed up under some conditions, which 
>>> leads to eviction failures and OOM errors
>>>        - infinite loop in assoc.c (again, only under certain 
>>> conditions), which leads to a pegged cpu and non-responsive server
>>> Any updates on these issues, or a guess about a 1.2.6 release?
>>> I'm hoping that these issues are not stuck behind the binary protocol 
>>> ...
>>> thanks,

More information about the memcached mailing list