script/lint ran `prettier --check .`, byte for byte what script/fmt-check
runs, so make check checked formatting twice and did no static analysis on
a cryptocurrency wallet. Two used-but-not-imported crashes shipped past it.
ESLint is pinned in package.json with @eslint/js recommended as the base and
a flat config in eslint.config.js. no-undef and no-unused-vars are restated
error-level so a future recommended-set change cannot downgrade them.
Globals are declared per tree rather than globally, because a too-wide set
hides the next unimported identifier: browser for the popup and content
scripts, service worker for src/background/ and src/shared/, browser for the
one documented POPUP ONLY module in src/shared/, jest for tests/, node for
build.js, and both for the e2e harnesses, which carry the callbacks they
ship into the page inline.
Two rules new to the recommended set are narrowed, and both would have cost
something to satisfy. no-useless-assignment is off for approval.js and
confirmTx.js only: it flags the `password = null` and `decryptedSecret =
null` wipes at 9 sites there, which are dead by construction — that is what
a best-effort wipe of decrypted key material is — and the rule's fix is to
delete the wipe. It stays on for the rest of the tree, so an ordinary dead
store elsewhere is still an error. preserve-caught-error is off tree-wide:
it would change what the wallet's error paths throw at 3 sites
(src/shared/balances.js 207 and 215, tests/e2e/firefox/run.js 131), and
adopting `{ cause }` is a decision of its own rather than a side effect of
turning a linter on, so new code is not held to it either pending that
decision.
Every remaining violation is fixed: 41 unused bindings and 53 undefined
identifiers. Unused catch bindings became `catch {`, which the repo already
used; the shared init(ctx) view signature keeps its parameter as _ctx in the
three views that do not read it. src/shared/uniswap.js keeps its unused
V2_SWAP_EXACT_OUT decoder behind a scoped disable, because deleting it would
widen the gap it represents rather than close it (#283). driver.js's waitFor
had a plain dead store in its `last` initializer, which the newly scoped
no-useless-assignment catches; the initializer is dropped.
Linting is containerized. script/lint builds the Dockerfile's new lint stage
so the ESLint deciding whether this repo is green is the pinned one and not
whatever the host has; AUTISTMASK_LINT_NATIVE, set only in that image, is
what makes make check inside the CI build lint in place instead of recursing
into docker, and a value set to anything else is now an error rather than a
silent fall-through to the docker path. The check stage takes a COPY --from=
lint dependency so a lint failure fails the whole build early rather than
racing it.
The lint stage roughly doubles the image build, which exposed script/test's
30s cap as marginal rather than a bound: on the first CI run to rebuild the
base stage cold it killed a healthy suite at 30.6s with nothing asserting
false. The cap is a guard against a hung suite, not a wall-clock budget, and
one a healthy suite can trip teaches "just run it again". It stays at 30s on
a host, where the suite runs in about 8s and REPO_POLICIES' figure holds,
and the Dockerfile raises it to 180s through AUTISTMASK_TEST_TIMEOUT for the
in-image run, which also pays a cold jest cache and shares the runner with
the rest of the build. script/test now names a timeout kill as one instead
of reporting it as a test failure, and skips the verbose rerun in that case,
which would only spend the same wall clock to be killed again.
No --fix anywhere in the lint path: make check remains non-mutating.
The README claim that a used-but-not-imported identifier is invisible to
make check, and the same claim in script/test-e2e, are no longer true and
are corrected.
ConfirmTx -- the screen that decides what gets signed -- had no automated
coverage of its own behaviour. The arithmetic underneath was well tested; the
wiring was not, so a mutant making the spend gate read the displayed fee
estimate instead of the reserve would have reintroduced the #154 overspend with
the suite still green.
Nine end-to-end tests now drive it for both the native and ERC-20 paths,
covering the pending, funded, over-balance and estimate-failed states, and
asserting that the gate reads the reserve rather than the estimate. Swapping the
two makes the suite fail. The view height is asserted constant across every
state transition rather than merely printed.
Reaching the screen needs a funded balance and a gas estimate, so the route
interception gains fixtures for both. Testing the estimate-failed state means
provoking the console error the code is supposed to emit, which the harness
otherwise fails a run on; an expectation mechanism consumes exactly one matching
record, is scoped to the declaring test, and fails that test if nothing matched,
so it cannot mask an unrelated error.
Runs the real popup in a pinned containerized Chrome and fails on any uncaught
page error or console.error. Also fixes the two defects it caught: the missing
showView import in addToken.js and the missing addressDotHtml import in
transactionDetail.js.
closes#150closes#151
