On Mar 7, 2006, at 11:05 PM, Ruediger Pluem wrote:
Shouldn't we make the size of FLUSH_WAIT configurable somehow
(possibly in a next step)?
I think that makes a lot of sense as well, but wanted to
hold off on that until we got some agreement that making
the whole flushing aspect configurable was a Good idea :)
@@ -366,6 +355,7 @@
"proxy: error processing body");
isok = 0;
}
+ /* XXX: what about flush here? See mod_jk */
We sent an eos bucket down the chain just a few lines up. Doesn't
this cause a flush
of the remaining data by the core output filter?
I mean that in mod_jk, depending on whether FlushPackets is
set or not, we do the *reverse* as the end, which is
confusing... Trying to track down why :)
===================================================================
--- modules/proxy/mod_proxy.h (revision 384045)
+++ modules/proxy/mod_proxy.h (working copy)
@@ -301,6 +301,11 @@
#if APR_HAS_THREADS
apr_thread_mutex_t *mutex; /* Thread lock for updating
address cache */
#endif
+ enum {
+ flush_off,
+ flush_on,
+ flush_auto
+ } flush_packets; /* how to deal with bad
headers */
Just a style question: Shouldn't we define a type for this outside
the struct and
declare flush_packets as this type?
Not really. Check out the proxy_server_conf struct for example.
Of course, that member should be tacked onto the end really...