Getting epoll working...?

Brad Fitzpatrick brad@danga.com
Thu, 31 Jul 2003 15:34:44 -0700 (PDT)


Put a static build (recommended) of 1.1.8 with epoll support (if
available) at:

   http://www.danga.com/memcached/dist/binaries/linux-x86/

Don't use the -h flag to test the libevent method, like I accidentally
said earlier.  Just do:

$ EVENT_SHOW_METHOD=1 ./memcached-1.1.8

- Brad


On Thu, 31 Jul 2003, Alan Kasindorf wrote:

> Also, I fixed the other headers already. It's the mystery of epoll.h
> that's been bugging me.
>
> Getting a binary would probably be helpful for just seeing if I've even
> patched the kernel right at this point. Would much appreciate it!
>
> Note that this is a .... redhat 7.2 machine.... I might be able to get
> 8.0 on it if it ends up being required, but likely nothing newer.
>
> -Alan
>
> Brad Fitzpatrick wrote:
>
> >You don't need to modify libevent at all.  It'll auto-detect epoll if it
> >can find the header files and system call.  If you rebuilt your kernel,
> >you have the system call, but you don't have the headers.
> >
> >Look at BUILD in the memcached distribution.  You have to modify your
> >headers.  Then, when you run libevent's configure, you should see it
> >building with epoll support.
> >
> >Alternatively, you want our memcached static binary?
> >
> >Then, to test your kernel:
> >
> ># EVENT_SHOW_METHOD=1 ./memcached -h
> >
> >and it'll say:
> >
> >libevent using: epoll
> >
> >
> >
> >On Thu, 31 Jul 2003, Alan Kasindorf wrote:
> >
> >
> >
> >>Ok, so I'm trying to get epoll working happily and rapidly running out
> >>of time.
> >>
> >>What patch from *where* actually works with libevent? I patched a 2.4.21
> >>vanilla kernel with the latest 2.4.21 one, it's running fine, but the
> >>header files are not what libevent expects, and from there different
> >>things seem to be named wrong (looking for an epollops struct, etc).
> >>
> >>Did you have to patch libevent to get it to compile?
> >>
> >>I'm going to keep poking at it here/see if I can't use a redhat rpm, but
> >>any insight is greatly welcomed. We're really behind in getting this
> >>project done.
> >>
> >>Thanks,
> >>-Alan
> >>
> >>
> >>
> >>
> >>
>
>
>
>