github ALLFATHER-BV/wadamesh beta_61
wadamesh beta_61

latest releases: beta_106, beta_105, beta_104...
pre-release3 months ago

Test build. Stable stays beta_50.

If you are on beta_60, update. It shipped with three regressions from the MeshCore 1.17 rebase, all fixed here. Flashing beta_61 over the top is enough — nothing needs erasing, and no data was lost in any of them.

Flash it from https://flasher.wadamesh.com (flip the Beta toggle), or on-device via Settings → Firmware with "Get test builds (beta)" enabled.


The beta_60 regressions

Boot loop on an empty contact list (#249). A device with no contacts saved — a fresh install, an erase-and-flash, or a card that had not mounted yet — crashed during startup and rebooted forever. 1.17 reserved eight slots at the head of the contact table and claimed them before the table was allocated, so a new startup routine walked a null pointer. Every board on our bench had contacts on it, which is exactly why this got past testing; the release process now checks that case. Reported by @Pierre747 and @rustinmyeye, whose boot log is what cracked it.

The newest contacts were unreachable (#252). Same reserved slots, different consequence: the contact list, the action sheets and the phone-app sync were still counting from the old start of the table, so the eight most recently added contacts could not be shown or acted on. A contact you had just added was invisible, and a room server's Log in silently did nothing because it resolved to an empty reserved slot. Nothing was lost — those contacts were in the table the whole time. Reported by myshoeisonfire.

The T-Display P4 could send but never receive (#253). No channel messages, no delivery confirmations, no adverts, RX counters flat at zero — while the Monitor app still showed signal, because it talks to the radio directly. The P4 is the one board whose radio interrupt is not wired to the processor, so it has to ask the radio over SPI whether a packet arrived. That polling code was never in the repository — it lived only in a generated directory that gets rebuilt — so the 1.17 rebase erased it. It is rewritten and now version-controlled, with a build-time check so it cannot vanish again. Reported by @D-Melhede.

Also fixed

  • A contact that changes its node name no longer leaves two chat threads behind. The thread is matched by key, but its title was never updated, while incoming messages are filed by sender name — so the first message after a rename started a second thread. Reported by myshoeisonfire.
  • Release builds now derive the firmware version, build date and core version from the release itself, and a release aborts if any board's image does not carry its own tag. beta_60 reported v1.16.0-touch in About because that value was maintained by hand.
  • All 13 language files gained the 48 UI strings none of them had, so those stop being permanently English and translators can see the gaps. New tool scripts/build/audit-lang.py reports drift in both directions.

Known, not fixed here

The intermittent T-Lora Pager crash (#233, #251) is still open. Both dumps decode to a freed LVGL callback — the same underlying fault seen from two angles — and @PixPMusic's #234 restructures exactly that teardown. It is not a 1.17 regression and predates beta_60.

Don't miss a new wadamesh release

NewReleases is sending notifications on new releases.