From: Nick Mathewson <nickm@torproject.org>
Date: Thu, 17 Nov 2011 22:42:45 +0000 (-0500)
Subject: add comment to new consider_reading code
X-Git-Tag: release-2.0.16-stable~2
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=ebf82199dfc91c437b304649829fb0a259726d5c;p=libevent

add comment to new consider_reading code
---

diff --git a/bufferevent_openssl.c b/bufferevent_openssl.c
index 53a4d686..d703279c 100644
--- a/bufferevent_openssl.c
+++ b/bufferevent_openssl.c
@@ -772,6 +772,14 @@ consider_reading(struct bufferevent_openssl *bev_ssl)
 		if (do_read(bev_ssl, n_to_read) <= 0)
 			break;
 
+		/* Read all pending data.  This won't hit the network
+		 * again, and will (most importantly) put us in a state
+		 * where we don't need to read anything else until the
+		 * socket is readable again.  It'll potentially make us
+		 * overrun our read high-watermark (somewhat
+		 * regrettable).  The damage to the rate-limit has
+		 * already been done, since OpenSSL went and read a
+		 * whole SSL record anyway. */
 		n_to_read = SSL_pending(bev_ssl->ssl);
 	}