Record published recovery correction and final node restoration

This commit is contained in:
mikecarper committed 2026-10-02 04:14:38 -07:00
1 parent 7092488334
commit 7f0b689191
1 file changed
+42
@@ -243,3 +243,45 @@ confirmed the normal 2.4.10 bootloader, fixed application, saved radio tuple
`869.6179809,62.5,8,5`, and all seven unchanged settings digests. USB data
was restored. This checks the restored node's Bluetooth entry and exit;
it is not an Open transfer.
## Published recovery correction verification: PASS
On 2026-10-02, [bootloader recovery workflow run 36997468388](https://github.com/mikecarper/Adafruit_nRF52_Bootloader_OTAFIX/actions/runs/36997468388)
completed all 37 jobs: preparation, 35 board builds, and release
packaging/publication. The corrected production source is tag
`R_v0.11.0-OTAFIX2.4.11`, commit
`de88c5fcae72bb00992ac9db7fd144fec150835e`.
The [existing recovery 2.4.11 prerelease](https://github.com/mikecarper/Adafruit_nRF52_Bootloader_OTAFIX/releases/tag/R_0.11.0-OTAFIX2.4.11)
retains all 110 filenames and its original distribution tag object
`b3784854e72b1d3d68e6bef04c1890f251256f59`, which still points to
`d6b3edf7d8b58c15187ccae54ea212cfa806d01c`. The release remains non-latest;
normal latest remains OTAFIX 2.4.10. All downloaded digests, archive checksums,
provenance, 35 profile manifest CRCs, and HEX/DFU ZIP/UF2 reconstructions passed.
Six included profiles remain explicitly pending hardware qualification, and
`thinknode_m8` remains excluded. The published recovery bundle SHA-256 is
`803dffcfed20331205d1fbe581593865785a4e53988efd20d777e5fe19b05b3d`.
The physical Open and Nordic tests used RAK4631 `board` qualification CRC
`C2A4B64B`, raw SHA-256
`239c23fabcee8478475b3bc3e1c2fc94daa79f386c5fd7ccf980c689220eda7a`.
The published RAK4631 `board` production image has CRC `AA30573D`, raw SHA-256
`13188cc150682a23bdee1ed7d2340a4ff0e6701b6d18dc347b75803c9ae53ae2`.
Both have packed version `0x02040BFF`; their complete raw binaries differ.
The production/qualification comparison found exactly eight differing data
bytes out of 40,960: four in the `CURRENT.UF2` FAT create/update time fields
and four in the resulting whole-image manifest CRC. All remaining 40,952
bytes match, including executable instructions, vectors, CF2, version, and
policy guards. The production timestamp matches the clean source commit,
and the production DFU ZIP, published HEX, and workflow artifact agree.
This verifies the published implementation against the tested one while
keeping the physical test tied to its actual qualification artifact.
### Final test-node restoration
After both corrected 2.4.11 phone tests and publication verification, the
RAK4631 was restored to the published normal 2.4.10 bootloader and the fixed
`v1.17.1.8-ble-dfu-test` application. Final CLI verification confirmed those
versions, the saved radio tuple `869.6179809,62.5,8,5`, and all seven unchanged
settings digests. USB data was re-enabled. No erase firmware was used.