The runner image's `reuse` failed to import (NoEncodingModuleError): it
needs an encoding-detection backend, which the bare `reuse` install does
not provide and which libmagic/`file` is not present to satisfy. Install
`reuse[charset-normalizer]` so the reuse job can run in the image.
Rebuild the runner image and recreate the containers to pick this up.
Assisted-By: Claude Code (claude-opus-4)
Claude-Session: https://claude.ai/code/session_01NFpGtGTWUEYXLwZeZs2RAV
Signed-off-by: Stan Grams <sjg@haxx.space>
The host-executor runners run jobs inside one container that already has
the Rust toolchain and all build dependencies baked in, so the per-job
`apt-get install` and rustup steps were redundant. Worse, with two jobs
running concurrently in the same runner they collided on the dpkg lock
("Could not get lock /var/lib/dpkg/lock-frontend").
Remove the system-dependency, rustup and cache steps; jobs now run cargo
directly. The cargo registry persists in the long-lived runner container.
Assisted-By: Claude Code (claude-opus-4)
Claude-Session: https://claude.ai/code/session_01NFpGtGTWUEYXLwZeZs2RAV
Signed-off-by: Stan Grams <sjg@haxx.space>
The host-executor runners have no Docker daemon, so fsfe/reuse-action
(a Docker action) fails with "Cannot connect to the Docker daemon". Call
the reuse CLI directly; it is baked into the runner image.
Assisted-By: Claude Code (claude-opus-4)
Claude-Session: https://claude.ai/code/session_01NFpGtGTWUEYXLwZeZs2RAV
Signed-off-by: Stan Grams <sjg@haxx.space>
Add container/: a rootless Podman + systemd (Quadlet) setup to run
per-project Gitea Actions runners on one host instead of VMs. Uses the
host executor with a purpose-built image that bakes in the Rust toolchain
and all build dependencies (opus, alsa, soapysdr, clang), so CI runs skip
the per-run install and cold soapysdr-sys build.
Includes the runner Containerfile, first-boot registration entrypoint,
act_runner config template, two Quadlet units (trx-rs + a second project),
and a README covering build, registration, the required host-executor
workflow tweak, and tuning.
Assisted-By: Claude Code (claude-opus-4)
Claude-Session: https://claude.ai/code/session_01NFpGtGTWUEYXLwZeZs2RAV
Signed-off-by: Stan Grams <sjg@haxx.space>
clippy needless_range_loop (rust 1.97) flagged the loops in ldpc_check
and ldpc_decode that use a range only to index one array. Replace them
with iterator/enumerate forms. The belief-propagation loops that index
several arrays by the same variable are left as-is (not flagged).
Behaviour is unchanged; the transformations are index-for-index
equivalent. Verified the lib compiles and is clippy-clean; the crate's
LDPC tests run in the CI test job (they need a dev-dependency not
available in the local offline sandbox).
Assisted-By: Claude Code (claude-opus-4)
Claude-Session: https://claude.ai/code/session_01NFpGtGTWUEYXLwZeZs2RAV
Signed-off-by: Stan Grams <sjg@haxx.space>
get_uint returns Option<u32>, so `? as u32` is an unnecessary same-type
cast flagged by clippy under -D warnings (rust 1.97).
Assisted-By: Claude Code (claude-opus-4)
Claude-Session: https://claude.ai/code/session_01NFpGtGTWUEYXLwZeZs2RAV
Signed-off-by: Stan Grams <sjg@haxx.space>
CI runs a newer clippy (1.97) than was available locally, which flagged
three lints in trx-ftx not caught earlier:
- question_mark: replace the Some/None match in CallsignHashTable::lookup
with `self.entries[idx].as_ref()?`
- collapsible_match: fold the nested `if` in text.rs char/nchar into match
guards on the AlphanumSpaceSlash arm
Behaviour is unchanged; verified clean with nightly clippy (1.93).
Assisted-By: Claude Code (claude-opus-4)
Claude-Session: https://claude.ai/code/session_01NFpGtGTWUEYXLwZeZs2RAV
Signed-off-by: Stan Grams <sjg@haxx.space>
sjg
Stan Grams