Re: [Libevent-users] Re: evhttp: persistent connections over ssl with chunked transfer encoding bug?

2011-12-22 Thread Hochhaus, Andrew
On Thu, Dec 22, 2011 at 5:31 PM, Catalin Patulea wrote: > Seems like bev_ssl flushes data to SSL_write and notifies back more > eagerly than bev_sock, which is confusing evhttp_send_reply_end. Thanks Catalin. Based on your comments, I tried moving evhttp_write_buffer() above evbuffer_add() and i

Re: [Libevent-users] Re: evhttp: persistent connections over ssl with chunked transfer encoding bug?

2011-12-22 Thread Catalin Patulea
On Thu, Dec 22, 2011 at 5:12 PM, Catalin Patulea wrote: > I'll be digging into this to try to figure out what's going on. Seems like bev_ssl flushes data to SSL_write and notifies back more eagerly than bev_sock, which is confusing evhttp_send_reply_end. evhttp_send_reply_end tries to write the l

Re: [Libevent-users] Re: evhttp: persistent connections over ssl with chunked transfer encoding bug?

2011-12-22 Thread Catalin Patulea
On Thu, Dec 22, 2011 at 4:55 PM, Hochhaus, Andrew wrote: > You can also reproduce the behavior in a web browser by running the > server (main.cc), browsing to https://127.0.0.1:8889/ and then > refreshing a few times (until one of the TCP connections is reused). I can reproduce this too against ma

[Libevent-users] Re: evhttp: persistent connections over ssl with chunked transfer encoding bug?

2011-12-22 Thread Hochhaus, Andrew
On Thu, Dec 22, 2011 at 3:29 PM, Hochhaus, Andrew wrote: > http.py is an example HTTP client that creates a persistent HTTP > connection and sends two requests to each of the above handlers over a > shared TCP connection You can also reproduce the behavior in a web browser by running the server (

[Libevent-users] evhttp: persistent connections over ssl with chunked transfer encoding bug?

2011-12-22 Thread Hochhaus, Andrew
Hi, While working on a comet http server, I have run into a problem where the second request on a reused (persistent) TCP connection is not processed by evhttp. Has anyone experienced behavior like this? I created a minimal test case [1] which demonstrates that all of the following must be true f

[Libevent-users] Re:in-order callbacks

2011-12-22 Thread Haseeb Abdul Qadir
So could a possible fix be to call event_base_once() with callback==consider_read and timeout == 0 to make sure it gets processed in the next iteration of the loop? On Thursday, 22 December, 2011 at 8:57 PM, Mark Ellzey wrote: > On Thu, Dec 22, 2011 at 01:33:28PM +0500, Haseeb Abdul Qadir wrot

Re: [Libevent-users] PATCH: bufferevent_openssl consider_read doesn't drain underlying BE in all cases

2011-12-22 Thread Mark Ellzey
On Thu, Dec 22, 2011 at 01:33:28PM +0500, Haseeb Abdul Qadir wrote: > What you guys mean when you say 'build correct infrastructure'? Does it mean > the ability to schedule callbacks and guarantee that callbacks will called in > the order they've been queued? For example a timer callback with a t

Re: [Libevent-users] PATCH: bufferevent_openssl consider_read doesn't drain underlying BE in all cases

2011-12-22 Thread Haseeb Abdul Qadir
What you guys mean when you say 'build correct infrastructure'? Does it mean the ability to schedule callbacks and guarantee that callbacks will called in the order they've been queued? For example a timer callback with a timeout of 0 is usually enough to enough to defer processing to the next i