From 329996b142455428a6b8cc66d01a0deb203f8d2e Mon Sep 17 00:00:00 2001 From: "Olle E. Johansson" Date: Sun, 3 Apr 2016 10:18:35 +0200 Subject: [PATCH] http_client Update TODO. --- modules/http_client/TODO.txt | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/modules/http_client/TODO.txt b/modules/http_client/TODO.txt index 81738c5dde4..c323c6dfa01 100644 --- a/modules/http_client/TODO.txt +++ b/modules/http_client/TODO.txt @@ -1,5 +1,7 @@ The http_client module - todo ====================== +- clean up + Clean up curl connections, pconn and conn objects at exit time. Just to be nice. - connection parameters read=line/all - Implemented in API, but not in parameter @@ -12,15 +14,14 @@ The http_client module - todo - save fingerprint of cert for trust-on-first-use procedure -- connection and last status for httpcon-object, needs to be object in process memory, so child_init - +- connection and last status for httpcon-object, needs to be object in process memory + - done. - test what happens if we allocate curlcon objects in shared memory and use locks Code implemeneted in git master branch for testing - if 5xx response or timeout set flag - function in dialplan - httpcon_exists(object) - pv $curlstatus(object) ?? - this is per process !!! - - can be used to fix the $curlredirect pseudovariable too - Figure out a way to communicate back the content-type of the response. Another pvar arg? - This is per process! @@ -28,6 +29,8 @@ The http_client module - todo - Reload of configuration file from RPC interface - Need to close connections (if we have open ones according to idea above) - Need to rebuild core httpcon objects and "mirrors" in pkg memory + - Should we just abandon old pkg objects and let them rot away + in space and time? - Can we add a lock to the connection list while switching if we load a new list, then switch active list and then free the old ones