[fbettic]

Summary

• accept the physically observed 12PSSSS5 serial and UOID structures alongside 12PSSSS3 and 12PSSSS4
• accept the physically observed 2026-04-17 HWCONFIG calendar-shaped value
• retain fail-closed rejection of unobserved identity families and date values
• record the fifth camera's non-sensitive fingerprints and physical recovery result
• update the read-only identification output and focused regression tests

Relation to existing reports

Related to #21 and provides a narrower, evidence-backed alternative to #23.

PR #23 correctly identifies that the digit after 12PSSSS is not limited to 3 and 4, but widens the pattern to [3-9]. This change accepts only the directly observed set [345]. It also adds the 2026-04-17 HWCONFIG date required by the same physical unit, plus regression coverage and the repository's compatibility documentation.

Physical evidence

The fifth camera is an EF-S7-V1.0.30B unit with manufacturing code 2526, an Ingenic T23 processor, and an 8 MiB ZB25VQ64D-family SPI NOR reporting JEDEC ID 5e4017. Before recovery it did not enumerate as a USB device.

Three consecutive 8 MiB reads were acquired at 1 MHz without moving the SOIC-8 clip. They are byte-for-byte identical with SHA-256:

f41e33b9927cfd8a5b11180872a2b940d645ad33ca3d13d5271476e930eb352e

The dump has:

• the existing normalized invariant SHA-256 4dee29ee9f996779a8af0f4e4a66ebad3c6c1ca353cf8b4287b4f1f9a8b12823
• exact matches for boot, kernel, root, system, partition layout, stock patch window, and known HWCONFIG bytes
• the observed 261-byte type-12 HWCONFIG record with extension 00 00 00 08 40
• structurally valid 12PSSSS5 serial and UOID fields
• HWCONFIG date 2026-04-17
• a live, CRC-valid serial.cfg
• 659 CRC-valid JFFS2 nodes, 657 obsolete nodes, and 98.457% non-FF config usage

The three-read recovery build passed strict post-build validation and produced SHA-256:

45acd32f451428513f14ca1238815e52a0508636060ef6973fa3ebafe0228a41

After programming, a complete 8 MiB readback was byte-for-byte identical to that image. The camera then booted, enumerated normally, and produced a working video feed.

Safety rationale

• no generic range or --force path is introduced
• identities 2 and 6 and an unobserved valid date remain explicitly rejected in tests
• all invariant firmware hashes and writable-region restrictions remain unchanged
• recovery continues to preserve the complete HWCONFIG partition and the camera's own serial payload
• no raw dump, recovery image, serial.cfg, complete serial, or complete UOID is committed

Validation

• Python 3.10: python -m unittest discover -s tests -q — 165 tests passed
• Python 3.12: python -m unittest discover -s tests -v — 165 tests passed
• package installation, CLI help, packaging smoke test, compileall, JSON parsing, and git diff --check passed
• a fresh build from all three physical reads reproduced the recorded recovery SHA-256 and matched the permanent full readback byte-for-byte

No package version is changed here; release/version selection is left to the maintainer.

[phryneas]

Looks good! Sorry I reacted so late, after we moved the repo, notifications were off!