This is a test build. Stable stays on beta_65. Flash it from wadamesh.com with the Beta toggle on, or turn on "Get test builds (beta)" on the device.
If your T-Deck types the wrong letters, there is now a switch
Settings > Keyboard > "Older keyboard protocol" skips the automatic detection and uses the protocol every T-Deck keyboard understands.
It exists because the detection cannot always be certain, and the failure is self-trapping: when it guesses wrong your keyboard types the wrong letters, so any remedy that requires typing is not a remedy. This switch is operated by touch and applies immediately, with no reboot.
Turning it on costs modifier latching, which a keyboard that needs the switch cannot do anyway. Automatic detection is unchanged and remains the default, so if beta_73 fixed your keyboard nothing here changes for you.
This is not a substitute for finding the cause on the devices where detection still gets it wrong (#351). It stops people being stuck while that continues, which is a different and more urgent thing.
The web settings page could quietly change your radio
Reported by @bradhawkins85 (#345). The remote page displayed a bandwidth of 62.5 as 62, and pressing Apply sent 62 back to the device. So opening the page and applying anything reconfigured the radio onto a bandwidth the mesh was not using.
Two independent faults, both needed: the device serialised bandwidth with no decimal places, and the form's bandwidth input had no step attribute so it would have rejected a decimal anyway. The apply direction was never wrong, which is exactly why it looked cosmetic while actually changing the radio.
Keyboard
Shift latches like symbol and alt already did (#344, @jrote1): tap once for the next letter, tap twice to lock until you tap again. It was a plain "is shift held" test while the other two went through the latching class, which is why they behaved differently. The two shift keys share one latch, so either arms it and either releases it.
The Attaky Mesh Deck's front D-pad and SELECT work, by @oumike (#346). The board has nine buttons and only POWER did anything. They hang off an expander rather than direct GPIO, so the driver drains an expander queue in its own pump.
Reading a route
Each hop now shows its path prefix beside the repeater name (#348, @jrote1): 3A Repeater Name rather than the name alone. The prefix is what the node is keyed by on the wire, so it stays true when a repeater is named misleadingly or two share a name, which is precisely when you are reading a route.
Translations
"Known regions" and the Bluetooth pairing hint are translatable and translated, from @pisti87 (#347).
For testers
Still open and not fixed here: a map that renders nothing since beta_71 (#339), a screen that will not stay dark after a manual lock (#333), the web flasher hanging at "wrapping up" on the T-Display P4 (#342), and an M9 reporting no SD card when formatting (#343).
A reminder, because it caught someone: two separate settings control whether an arriving message lights a dark screen. "Flash on new message" is in Settings > Keyboard and is the one that wakes the screen. "At a glance" is in Settings > Display and draws the preview text. Turning off the second alone will not stop the screen lighting up.