[Memcacheddotnet-devel] Re: memcache patch

[Maxim M. <mm...@um...>]

After a bit more digging this seems to be happening because the socket is
not removed from the availPool. The reason it's not removed is because the
objects hash code changes as a result of the TrueClose method.  The
GetHashCode method is overriden in SockIO and changes based on the private
Socket variable. Once the socket is closed, gethashcode returns 0.

Was there a reason for overriding the default gethashcode method? It appear=
s
to be working fine without it.

Max

On 1/17/06, Maxim Mass <mm...@um...> wrote:
>
> There appears to be another bug in that maintenance thread. It's still no=
t
> closing connections but this time due to a different exception.  This lin=
e:
>
> DateTime expire   =3D (DateTime) sockets[ socket ];
>
> throws a null ref exception since that socket doesn't exist in the hash.
> socket comes from a foreach loop that iterates over all the keys... So
> looking at this for about 10 mins now seems like there are a couple issue=
s..
> For some reason, it's iterating over keys (sockets) that have been alread=
y
> removed. Or it's not removing and closing correctly.
>
> I'm going to look at this further but let me know if you come up with a
> fix for it
>
> Max
>
> On 1/17/06, Tim Gebhardt <ti...@ge...> wrote:
> >
> >  The patch looks good.  The timing stuff is a lot easier to read now.  =
I
> > redeployed the SourceForge.net download.
> >
> >
> >
> >
> >
> > -Tim
> >
> >
> >  ------------------------------
> >
> > *From:* max...@gm... [mailto: max...@gm...] *On Behalf Of
> > *Maxim Mass
> > *Sent:* Monday, January 16, 2006 9:04 PM
> > *To:* ti...@ge...
> > *Subject:* memcache patch
> >
> >
> >
> > I fixed a few things in the SockIOPool.. Take a look and sanity check
> > this please..
> >
> >
>