[cryptographicturk]

Summary

pyasic has restart_backend(), which restarts the mining program without rebooting the operating system. asic-rs only has restart(), a full device reboot, so callers moving from pyasic lose that action. This adds it.

• New RestartMining trait with restart_mining() and supports_restart_mining(), unsupported by default, added to HasMinerControl. It follows the shape of Restart / Pause / Resume.
• Python: await miner.restart_mining() and miner.supports_restart_mining, with regenerated stubs.
• C FFI: asic_rs_miner_restart_mining and a restart_mining capability flag.
• Go: Miner.RestartMining() and Supports.RestartMining.
• Docs: a "Restart Mining" column in the supported-devices matrix and a row in docs/api.md.

A rejected request is returned as an error carrying the firmware's own message, not as false.

Backends

Firmware	Request	Evidence
Braiins OS+ 25.07 / 26.04 backends	PUT /api/v1/actions/restart	Confirmed present in the device's own OpenAPI document on two newer Braiins OS+ releases, and in the vendored meta/braiins documents ("Restarts mining", distinct from actions/reboot)
ePIC UMC	POST /softreboot with {"param": null, "password": ...}	Confirmed present in the device's own OpenAPI document, and in the vendored meta/epic documents
Braiins OS+ 21.09 / 25.03 / 25.05 backends	GraphQL bosminer { restart }	Same mutation set_pools_config already sends; not confirmed on a device
VNish 1.2 / 1.3	POST /api/v1/mining/restart	From pyasic and the vendored meta/vnish documents; not confirmed on a device
Auradine	POST /restart with parameter gcminer	From pyasic; not confirmed on a device
LuxOS	RPC resetminer	From pyasic; not confirmed on a device

Every other backend reports unsupported, including AntMiner stock (pyasic returns False there too).

If you would prefer a smaller first PR, I am happy to drop the four unconfirmed backends and send them as a follow-up.

Testing

• cargo fmt --all -- --check
• cargo test --all --locked --exclude asic-rs-pydantic --exclude asic-rs-pydantic-macros --exclude pyasic-rs --exclude asic-rs-ffi: 418 passed (rebased onto current master, rustc 1.99.1)
• cargo test -p asic-rs-ffi --locked -- --test-threads=1: 11 passed
• Python tests after maturin develop --features python --extras test: 126 passed
• Go tests were not run locally.

This adds 19 mocked tests (request sent and error path per firmware, and the FFI error path). They are mocked tests only: no endpoint here has been exercised on real hardware. Whether each one keeps the operating system up, and how long mining is down afterwards, has not been measured.

Some of the new tests live in restart_mining_tests.rs files beside the Braiins and LuxOS backends rather than in the existing tests module, to stay clear of other open changes to those modules. I can fold them in if you prefer.

Unrelated observation

Not changed by this PR: on VNish, the existing restart() calls mining/restart rather than system/reboot, so it restarts mining instead of rebooting the device, and restart() and restart_mining() currently reach the same endpoint there.

🤖 Generated with Claude Code

[b-rowan]

Just drop a couple of redundant tests, looks good other than that.

[cryptographicturk]

@b-rowan Dropped the two AntMiner restart_mining_tests.rs files along with their mod declarations, and rebased onto master to resolve the conflict with #419. The Braiins backends merged cleanly, so #419's error propagation is kept alongside restart_mining. The only textual conflict was the support matrix in docs/supported-devices.md. Forgeaxe and AntMiner v2018 landed after this branch was cut, so they now have the default RestartMining impl too. fmt and tests pass locally (418 passed). Clippy already fails on master in asic-rs-core (35 existing errors), so I linted the rest with those lints allowed. Mocked tests only; not run against real miners.