I've been running with this change -
evhttp_detect_close_cb(int fd, short what, void *arg)
{
struct evhttp_connection *evcon = arg;
- evhttp_connection_reset(evcon);
+ evhttp_connection_fail(evcon, EVCON_HTTP_EOF);
}
I think that you said that a similar change caused a segmentation fault for you
but I don't see how. If you remember can you send an email? Your suggested
change will result in calling evhttp_connection_fail eventually as well - from
the evhttp_write callback - since it will be the only one registered and it
will fail since the connection got closed. But the close detection close does
look like an unnecessary overhead and I'll try your suggestion as well.
Thx
--pv
> -----Original Message-----
> From: Niels Provos [mailto:pro...@gmail.com]
> Sent: Friday, December 19, 2008 2:32 PM
> To: Peter Vajgel
> Cc: libevent-users@monkey.org
> Subject: Re: [Libevent-users] http requests not freed by evhttp on
> client timeout
>
> I have attached a patch to fix this by not running close detection on
> servers. Severs already do this implicitly as they are waiting for
> more data on persistent connections. The svn branch also contains a
> test that makes sure that no connections are dangling on the server.
>
> Please, let me know if this works for you.
>
> Thanks,
> Niels.
_______________________________________________
Libevent-users mailing list
Libevent-users@monkey.org
http://monkeymail.org/mailman/listinfo/libevent-users
