From: Stefan Fritsch Date: Sun, 16 Jun 2013 21:34:41 +0000 (+0000) Subject: comment X-Git-Tag: 2.4.5~109 X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=3247468a925c4537e324fa6447a9e234c264a481;p=apache comment git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/branches/2.4.x@1493591 13f79535-47bb-0310-9956-ffa450edef68 --- diff --git a/STATUS b/STATUS index 57b7044a71..1dcd237252 100644 --- a/STATUS +++ b/STATUS @@ -128,6 +128,25 @@ PATCHES PROPOSED TO BACKPORT FROM TRUNK: https://svn.apache.org/r1463046 2.4.x patch: trunk patches work +1: jim + sf comments (did not do a full review, though): + - The description of the proposal is misleading. The above commits add + the eventopt MPM. Do we really need another MPM in 2.4? Shouldn't we + merge the optimizations into event, instead? + - The skiplist API lacks documentation. + - ap_skiplist_alloc()/ap_skiplist_free() should at least have a free list + for sizeof(ap_skiplistnode) sized items. Otherwise it will leak lots of + memory. + - There does not seem to be any need for structs ap_skiplist and + ap_skiplistnode to be public. I would be in favor of making them + private to skiplist.c to allow changing implementation details later + without breaking compatibility. + - register_timed_callback is intended to be used from different threads. + The g_timer_skiplist_mtx mutex ensures that access to the skiplist + itself is serialized. However, the skiplist is created using pchild + and will allocate from that pool, but other use of pchild outside of + the skiplist is not protected by g_timer_skiplist_mtx. You must either + use a sub-pool or document why using pchild is save (I don't think it + is). * core, mod_ssl: Lift the restriction that prevents mod_ssl taking