Steinar H. Gunderson modified c-ares to now also do to DNS retries even when

TCP is used since there are several edge cases where it still makes sense.

diff --git a/ares/CHANGES b/ares/CHANGES
index 160489c64c9b4e36c962c0946c9aed2d01e27df1..4f28e3b4caf3e8bf80abe2b996715c8500a73d3c 100644 (file)
--- a/ares/CHANGES
+++ b/ares/CHANGES
@@ -2,6 +2,9 @@
 
 * September 22 2007 (Daniel Stenberg)
 
+- Steinar H. Gunderson modified c-ares to now also do to DNS retries even when
+  TCP is used since there are several edge cases where it still makes sense.
+
 - Brad House provided a fix for ares_save_options():
 
   Apparently I overlooked something with the ares_save_options() where it

diff --git a/ares/ares_process.c b/ares/ares_process.c
index e8b9ef6fa79b2a3e80e6cb1d35bc1c3b40754c51..66007f0fb2e65c3b2b64c35e3cdf18254232b0f3 100644 (file)
--- a/ares/ares_process.c
+++ b/ares/ares_process.c
@@ -504,9 +504,12 @@ static struct query *next_server(ares_channel channel, struct query *query, time
         }
       query->server = 0;
 
-      /* Only one try if we're using TCP. */
-      if (query->using_tcp)
-        break;
+      /* You might think that with TCP we only need one try. However,
+       * even when using TCP, servers can time-out our connection just
+       * as we're sending a request, or close our connection because
+       * they die, or never send us a reply because they get wedged or
+       * tickle a bug that drops our request.
+       */
     }
   return end_query(channel, query, query->error_status, NULL, 0);
 }