Every call site that touched `browser.*` or `chrome.*` now goes through
`src/shared/browserApi.js`, the only file in the tree that names either.
It exposes lazily-resolved namespace handles for events and synchronous
methods, and promise-returning wrappers for everything that is
callback-shaped on Chrome. Callers await; `runtime.lastError` is gone,
folded into the rejection the wrapper produces on the Chrome path.
The Firefox suite gains the four dApp round trips the issue's definition
of done asks for — `eth_requestAccounts`, `personal_sign`,
`eth_sendTransaction`, and a closed approval window rejecting with
EIP-1193 4001 — driven through the real content script, background page
and approval windows. `--network none` was thought to rule that out
because it leaves no `http://` origin to inject into; loopback survives
it, so the page and a JSON-RPC node are served from 127.0.0.1 inside the
container and the run still reaches nothing but itself.
That harness refutes the premise it was built to verify. On Firefox
153.0.3, `browser.*` honours a trailing Chrome-style callback and does
populate `runtime.lastError`, both measured directly, and all four flows
pass against the unconverted code. So this is a uniformity and coverage
change, not a repair of a broken target; the PR records the measurement
in full.
One real defect is fixed on the way past: the window id written back
into a pending approval after `windows.create()` was unguarded, so an
approval settled during the open — an address switch will do it —
dereferenced a deleted entry.
Drives the real popup in a real Firefox with dist/firefox/ installed as an
unpacked MV2 temporary add-on via geckodriver. make test-e2e-firefox, outside
make check like the Chrome suite. Zero npm dependencies: plain fetch and
child_process against geckodriver's HTTP API. Base image, Firefox tarball and
geckodriver are each pinned by digest and verified at build time.
Error capture reads the privileged console service through Marionette's chrome
context, not WebDriver BiDi. BiDi delivers nothing at all for extension pages,
so a BiDi-based harness would observe zero events and report success -- the
vacuous-check shape this repo has shipped twice. Both the driver and the README
say so where someone would be tempted to simplify.
Demonstrated to discriminate: a background page that throws at the top of the
file, a missing import, and an async throw where every UI assertion still
passes each fail the run.
Three limits are measured and documented rather than papered over: capture is
poll-based so an error is attributed to a step, not a moment; the console ring
buffer holds 250 messages and evicts the oldest, measured against a clean-run
peak of 4; and the drained window ends roughly 1.5s after the last step, with
observed jitter rather than a hard boundary. Content-script capture is marked
unverified because --network none leaves no page to inject into, and that same
choice inverts coverage of network-dependent code.
