refresh ziti_session.edge_routers #620
Conversation
| @@ -1339,7 +1340,8 @@ static void edge_routers_cb(ziti_edge_router_array ers, const ziti_error *err, v | |||
| // check if it is already in the list | |||
| if (model_map_remove(&curr_routers, tls) == NULL) { | |||
| ZTX_LOG(TRACE, "connecting to %s(%s)", er->name, tls); | |||
| ziti_channel_connect(ztx, er->name, tls, NULL, NULL); | |||
| ziti_channel_connect(ztx, er->name, tls); | |||
| ers_changed = true; | |||
There was a problem hiding this comment.
should this be done when the ER is successfully (re)connected?
There was a problem hiding this comment.
that would cause (maybe a lot) more frequent session refreshes that are probably unnecessary
library/model_support.c
Outdated
| for (int i = 0; i < meta->field_count; i++) { | ||
| if (strncmp(meta->fields[i].path, json + tok->start, tok->end - tok->start) == 0) { | ||
| if (strlen(meta->fields[i].path) == token_len && |
There was a problem hiding this comment.
This change seems behaviorally equivalent to me, but now the strncmp only happens if a match cannot be ruled out based on length. So I might think it's an optimization... but doesn't using strlen only guarantee that the entire string will be scanned once before possibly being scanned again in strncmp?
There was a problem hiding this comment.
the issue was with an JSON having both service and serviceId field.
token(service) matched against serviceId field
There was a problem hiding this comment.
I reworked matching to avoid extra string scan
make sure path length matches token being checked
[fixes #614]