[geonnave]

The host half of storing LH2 stations by a 16-bit mask instead of a dense count, so a robot can hold the stations on, say, channels 3 and 9 without anything in slots 0 to 7. Today send_lh2_calibration refuses anything but stations 0 to N-1, and device info reports a count.

Wire, as the new bootloader and net core expect it (byte-for-byte with PyDotBot's packer):

• 0xA3 calibration message, still 84 bytes: u32 station_mask (bit i = station i, bits 16-31 zero) at 0, u32 station_index at 4, the float32 matrix at 8, this station's own u32 valid_mm[4] at 44, site name at 60, calibration id at 76. One message per station in index order; the mask, site and id are identical across a push.
• Device info 4, 179 bytes: lh2_homography_count (u8) becomes lh2_station_mask (u16) at offset 152, so lh2_flags moves to 154.

Host behaviour.

• send_lh2_calibration checks a push is one push: a mask naming 1 to 16 stations below 16, messages carrying exactly the mask's stations once each in index order, a non-empty rectangle in each, the same mask, site and id in every message.
• helpers.read_lh2_calibration_payload reads calibration schema 4 (each station's own valid_mm is required), builds the mask from the station indices, and refuses an index outside 0 to 15 or a duplicate. Schema 3 files are refused.
• DeviceInfo.lh2_summary names stations by index and channel, e.g. stations 2, 8 (channels 3, 9; valid, from flash); the fleet table's cell lists the indices. Robots below device info 4 read as firmware to reflash.

Flash together. This only talks to a bootloader and net core built for the station mask (config page magic SWRP); those come by cable, both images on every robot in one visit, and nothing should be pushed over the air to a robot that has not been reflashed. Install this host first: an old host would send the old count format, which the new net core cannot commit. PyDotBot's matching change (schema 4 files, mask messages, the baked page) is in DotBots/PyDotBot#322 and the PRs stacked on it.

Validation. Host only: 240 passed (pytest swarmit/tests), including the shared wire fixture pinned in both repos (stations 2 and 8 with different rectangles, so keying by position or sending the union in every message fails it), an empty rectangle refused, stations 2 and 8 going out under mask 0x0104, the refusals above, the device-info layout at offsets 152-155, and an emulated node's mask surviving the round trip. The calibration files PyDotBot writes produce identical bytes through both packers. Dashboard frontend builds; pre-commit clean.

File	+	-
swarmit/testbed/controller.py	+43	-35
swarmit/testbed/helpers.py	+43	-35
swarmit/testbed/protocol.py	+21	-23
swarmit/tests/lh2_wire_fixture.py	+32	-22
swarmit/tests/test_controller.py	+53	-31
swarmit/tests/test_helpers.py	+47	-10
swarmit/tests/test_protocol.py	+19	-15
5 small files: swarmit/cli/main.py, dashboard/frontend/src/App.tsx, tests/test_client.py, tests/test_webserver.py, tests/utils.py	+15	-12
Total, 12 files	+273	-183

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.04%. Comparing base (db07551) to head (8019e7a).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #175      +/-   ##
==========================================
- Coverage   91.30%   91.04%   -0.27%     
==========================================
  Files          27       27              
  Lines        5082     5124      +42     
==========================================
+ Hits         4640     4665      +25     
- Misses        442      459      +17     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.