An empty dashboard after a restart, and history that keeps itself
Fixed
- Signing in after a restart sometimes gave a dashboard of empty, stale cards. The router showed
in the corner with a green dot and the connection was genuinely fine, which is what made it so
confusing — everything that still worked was something that did not need the page. A refresh fixed
it, and that was the only cure. The browser sends two messages as it starts up and their order is
not guaranteed; if they arrived the less common way round, the page was left subscribed to nothing
for as long as that tab stayed open. Reported twice. Verified against live routers by forcing the
bad order: the affected cards received nothing before, and everything after. - A traffic graph that silently stopped was never restarted. RouterOS pushes traffic readings
and never acknowledges them, so a router that stops sending leaves an open connection, a healthy
status, and a chart that has simply stopped. Nothing errored, so nothing retried. MikroDash now
notices a stream that has gone quiet for ten seconds and reopens it, and says so on the card when
a stream keeps failing. - Test email failed against Microsoft 365 and Outlook with "504 5.7.4 Unrecognized authentication
type". MikroDash only ever offered one sign-in method, and those servers do not accept it. It now
uses whichever method the server actually advertises. Other providers are unaffected. - Bandwidth and traffic history was only recorded while somebody had the Dashboard open. Which
interfaces got written to history depended on which ones a browser happened to be watching, so a
report covered a series full of holes and presented the total as if it were complete. History is
now a property of the device: an interface is recorded continuously or not at all. - The install-wide "default interface" setting was ignored. It was read under the wrong name, so
changing it did nothing anywhere. It also had three different fallbacks in three places; a device
with no interface of its own recorded no history at all in the background.
Changed
- A router that rejects the sign-in is retried far less often. A wrong password used to put a
failed login into the router's own log every five seconds for as long as MikroDash ran; it now
backs off to at most five minutes. A router that is unreachable or rebooting is unaffected and
still retried quickly, and correcting the password takes effect immediately.
Internal
- Three new checks for the class of bug behind two of the fixes above: that the router-select
handler re-establishes every subscription it drops, that every entry point on the history recorder
is actually called by the shipped binary, and that URLs built into links are covered by the
endpoint audit rather than only ones passed tofetch.