A microphone the machine has and the app does not
1.3.0 taught the app to notice a capture stream that opens and is then never
called, and to rebuild when it happens. That fix could leave you worse off than
the fault it was for.
If the rebuild raced the device being released, no capture stream got built at
all — and nothing could recover from there. With no stream there is no callback
to watch, and the device check compares device names, which had not changed.
The app sat in listen-only until it was restarted, which is the same
"restart it and it works" shape as the original bug, only self-inflicted.
Now, when there is no capture stream but the machine does have a microphone, it
tries to open one every fifteen seconds. That happens on its own, without
touching playback — which matters, because on a machine whose microphone is
blocked in Windows privacy settings the retry will never succeed, and cutting
the audio every few seconds to keep trying would be worse than the problem.
Failing costs a line in the log and nothing else.
Two downloads fewer
Five files here instead of seven.
No more MSI. The firewall rule is added by an NSIS installer hook and the
MSI had no equivalent, so it produced an install that looked identical to the
working one and then left you with Windows' own firewall prompt — and
cancelling that writes a silent block rule that outranks everything after it.
That is the exact failure this application was written to remove, and shipping
a download that walked people into it was a mistake. Use
Zetta.Com_1.4.0_x64-setup.exe.
No more AppImage. It was 79 MB against the 4 MB of the .deb beside it,
because an AppImage has to carry WebKitGTK and GStreamer for a machine that has
neither, where the .deb and .rpm depend on what the distribution already
ships. Nothing could shrink that — carrying the engine is the format's whole
bargain. Debian, Ubuntu, Fedora and RHEL are covered by the packages; anything
else builds from source in two commands.
Updating
1.4.0, 1.3.x and 1.2.0 all work together. The wire format is unchanged, so
machines can be updated one at a time, in any order, with nothing going quiet in
between. Machines on 1.0.0 still cannot hear anybody and need updating.
If you are coming from 1.2.0 or earlier, the 1.3.0
notes are the ones
that matter — voice no longer needs a logout after installing, there is a log at
%APPDATA%\com.joy.zetta-com\zetta-com.log, and the roster says which build
each machine is on.
macOS is still unsigned: move the app to Applications and run
xattr -cr "/Applications/Zetta Com.app", then enable
System Settings → Privacy & Security → Local Network.