Stop the driver sweep at the first driver that claims the image

[jasiek]

What

Loading a .img with no metadata trailer used to import all 191 CHIRP driver modules before detection could run. In the browser every module is its own CDN round trip, so that sweep is the slowest thing the app does. Detection only ever needs the modules up to the winning driver, so the sweep now imports in the same order and stops at the first driver that claims the image.

Measured over the 108 metadata-less images in chirp/tests/images/: all 108 match, min 1 module, p25 30, median 72.5, p75 116, max 188 (Yaesu_VX-8*). Roughly 60% of the fetches disappear at the median; the worst case is unchanged.

Why this cannot pick a different driver

directory.get_radio_by_image() returns the first match in DRV_TO_RADIO insertion order, and insertion order is a pure function of import order. Import the same list in the same order and stop at the first hit, and the winner is exactly the one the full sweep would have returned.

What would not be safe is reordering the list to try likely drivers first: 111 registered classes inherit the default match_model, which is the bare comparison len(filedata) == cls._memsize, and several classes share a memory size — among those the winner is decided by order alone. The parity test below is what pins this down.

Changes

web/python/runtime_bridge.py

• New detect_image_driver_incremental(image_b64, module_short_names, progress_cb): imports one module at a time, checks only newly-registered classes after each, returns the match plus imported / total / exhausted.
• _image_class_matches() mirrors the per-class body of get_radio_by_image — both the match_model branch (metadata-less images) and the vendor/model/variant alias branch (images whose trailer the catalog could not resolve) — including swallowing driver exceptions during detection. One deliberate divergence, from review: it returns after the match_model branch instead of falling into the alias comparison with meta_vendor/meta_model both None the way upstream does. No registered class declares a None VENDOR or MODEL, so the comparison upstream still runs can never match; the docstring and FINDINGS record that so a future CHIRP bump has something to trip over.
• Classes registered before the sweep starts are checked first: a driver the user's radio selection already imported sits at the front of DRV_TO_RADIO, and the full sweep would have considered it first too.
• Factored out _decode_image_b64 and a _temp_image_file context manager so detection and loading hand drivers the same .img filename shape (some drivers key off the extension).
• import_all_driver_modules is untouched — scripts/build-catalog.mjs and the existing tests still use it.

web/js/runtime-rpc.js

• ensureAllDriverModules() → importDriverModulesUntilImageMatches(imageBase64).
• The session memo changes from a cached promise to an allDriverModulesImported boolean set only when the module list is exhausted: "some drivers are imported" is not a reusable fact the way "all of them are" is.
• The progress strip keeps its per-module callback; its total is now explicitly an upper bound and the strip usually ends early.
• Loading still runs through load_image_base64 → get_radio_by_image, so one code path picks the driver and one reads the image.

web/js/image-metadata.mjs — rebased onto the ImageDetectionError backstop merged in #46. The incremental sweep is what loadImageWithDriverFallback() injects, so both paths stop early: the unresolved-image path and the retry after a wrong fast-path resolve. The injected function is renamed importAllDrivers → importDriversForDetection, because it no longer imports every driver.

FINDINGS.md — new incremental-detection-is-order-safe entry with the ordering argument, the measured distribution and the upstream divergence; image-driver-resolution and all-driver-import-mechanics updated to describe the incremental sweep rather than the all-or-nothing one.

Tests

Four new cases in scripts/test-metadataless-image-load.mjs, alongside the retry-gate test from #46:

• Parity + early exit — a reference harness runs the full sweep and loads two metadata-less images (Baofeng_UV-3R, 13th module; Wouxun_KG-UV8D, 94th); a fresh Pyodide runtime per image then runs incremental detection and asserts the same module, class and channel rows, with imported < total. Parity is derived from the full sweep rather than hardcoded, so this is the test that catches a future reordering "optimisation". Also asserts progress reports stop exactly where detection stopped.
• Pre-registered drivers win — ensure_radio_module('baofeng_uv3r') first, then detect against a module list that excludes it; matches at imported === 0.
• Unclaimable image — a 7-byte payload no driver matches; asserts exhausted: true and every module imported.
• Metadata alias branch — Kenwood_TS-480_CloneMode.img resolves to ts480.TS480_CRadio incrementally (at module 168 of 191), exercising the branch where match_model is never called.

npm test — 768/768 pass (177 + 32 + 553 + 6). py_compile on web/python/runtime_bridge.py clean.

New dependencies

None.

Not in this PR

_ensure_chirp_module_file still blocks on one fetch at a time, so the remaining ~72 median fetches are strictly serial. Prefetching sources into the Pyodide FS with real concurrency is order-preserving and would compose with this change.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR optimizes .img driver detection in the Pyodide runtime by importing CHIRP driver modules incrementally and stopping as soon as the first matching driver class is registered, preserving CHIRP’s import-order–based selection semantics while reducing the number of module fetches in the browser.

Changes:

• Added incremental driver detection in runtime_bridge.py that imports modules one-by-one and checks newly registered radios after each import.
• Updated runtime-rpc.js to use the new incremental detection path and to cache only the “all modules exhausted” state, not partial imports.
• Added parity/early-exit and edge-case tests for metadata-less and metadata-unmatched images, plus updated documentation in FINDINGS.md.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
web/python/runtime_bridge.py	Introduces incremental detection helpers and refactors image base64 decode + temp file handling for consistent driver interactions.
web/js/runtime-rpc.js	Switches image-load fallback from full driver sweep to incremental detection with early exit and updated session caching semantics.
scripts/test-metadataless-image-load.mjs	Adds tests to prove parity with the full sweep, early-exit behavior, pre-registered precedence, and exhausted-list behavior.
FINDINGS.md	Updates operational notes and adds an ordering-safety finding documenting why early exit preserves correctness.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.