From: oetiker Date: Fri, 7 Nov 2008 13:51:24 +0000 (+0000) Subject: Much simpler handling of timestamp errors. Return an error to the user X-Git-Url: https://git.verplant.org/?a=commitdiff_plain;h=f942d3c825ad1073bb6d68edd9653b0c999bbe49;p=rrdtool.git Much simpler handling of timestamp errors. Return an error to the user when any of the time stamp values are invalid. This is similar to RRDTool's normal behavior. Removed the complex logic previously used to return error codes to the user. This solves a bug where non-advancing timestamps could have produced incorrect error output during "BATCH" mode. The bug was cause by using the sock->wbuf pointer for the error output. -- kevin brintnall git-svn-id: svn://svn.oetiker.ch/rrdtool/trunk/program@1648 a5681a0c-68f1-0310-ab6d-d61299d08faa --- diff --git a/src/rrd_daemon.c b/src/rrd_daemon.c index c6211a6..52e9f12 100644 --- a/src/rrd_daemon.c +++ b/src/rrd_daemon.c @@ -1384,7 +1384,6 @@ static int handle_request_update (listen_socket_t *sock, /* {{{ */ { char *file, file_tmp[PATH_MAX]; int values_num = 0; - int bad_timestamps = 0; int status; char orig_buf[CMD_MAX]; @@ -1488,18 +1487,17 @@ static int handle_request_update (listen_socket_t *sock, /* {{{ */ stamp = strtol(value, &eostamp, 10); if (eostamp == value || eostamp == NULL || *eostamp != ':') { - ++bad_timestamps; - add_response_info(sock, "Cannot find timestamp in '%s'!\n", value); - continue; + pthread_mutex_unlock(&cache_lock); + return send_response(sock, RESP_ERR, + "Cannot find timestamp in '%s'!\n", value); } else if (stamp <= ci->last_update_stamp) { - ++bad_timestamps; - add_response_info(sock, - "illegal attempt to update using time %ld when" - " last update time is %ld (minimum one second step)\n", - stamp, ci->last_update_stamp); - continue; + pthread_mutex_unlock(&cache_lock); + return send_response(sock, RESP_ERR, + "illegal attempt to update using time %ld when last" + " update time is %ld (minimum one second step)\n", + stamp, ci->last_update_stamp); } else ci->last_update_stamp = stamp; @@ -1534,21 +1532,7 @@ static int handle_request_update (listen_socket_t *sock, /* {{{ */ pthread_mutex_unlock (&cache_lock); if (values_num < 1) - { - /* journal replay mode */ - if (sock == NULL) return RESP_ERR; - - /* if we had only one update attempt, then return the full - error message... try to get the most information out - of the limited error space allowed by the protocol - */ - if (bad_timestamps == 1) - return send_response(sock, RESP_ERR, "%s", sock->wbuf); - else - return send_response(sock, RESP_ERR, - "No values updated (%d bad timestamps).\n", - bad_timestamps); - } + return send_response(sock, RESP_ERR, "No values updated.\n"); else return send_response(sock, RESP_OK, "errors, enqueued %i value(s).\n", values_num);