[MaxFreedomPollard]

web-ext run prints nothing when Firefox refuses to reload the extension. If manifest.json is broken while it runs, the watcher fires, Firefox rejects the reload and keeps the old version loaded, and the terminal shows no error, whether the reload came from a file change or from pressing R. The "Last extension reload" time just stops updating. Firefox for Android behaves the same way, including when rebuilding the package during a reload fails.

Both Firefox runners catch their reload errors and resolve with them, as their docstrings describe: reloadExtensionBySourceDir returns [{ sourceDir, reloadError, runnerName }] (src/extension-runners/firefox-desktop.js:105-115, src/extension-runners/firefox-android.js:178-189), and reloadAllExtensions wraps those in a MultiExtensionsReloadError. MultiExtensionRunner then discarded that value, because its fulfilment handlers returned a fresh { runnerName } or { runnerName, sourceDir } (src/extension-runners/index.js:80 and 113). handleReloadResults therefore only saw errors from runners that reject, which in practice means Chromium alone. A broken manifest is already reported for Chromium, and tests/functional/test.cli.run-target-chromium.js:151-163 checks it.

MultiExtensionRunner now keeps the result each runner resolves with ({ ...result, runnerName }), so these errors reach handleReloadResults and get logged:

Error occurred while reloading "/home/dev/my-extension" on "Firefox Desktop" - Remote Firefox: addonRequest() error: Error: Extension is invalid

The runners are not changed. The rejection path and the one-result-per-runner shape of the returned array also stay the same. The Chromium runner resolves with [{ runnerName }] and logs its own Extensions.loadUnpacked failures (chromium.js:406), so nothing is logged twice.

The same function was missing a space when no source directory is given, which is the R key path: the message read Error occurred while reloadingon "Firefox Desktop". That is fixed too.

Two unit tests are added next to the existing reload tests in tests/unit/test-extension-runners/test.extension-runners.js. They mirror "logs errors to the console while reloading all extensions" and "collects reload errors in results when reloading an extension", but the failing fake runner resolves with its error the way the Firefox runners do instead of rejecting. The first test also checks the "reloading on" spacing.

Testing: I could not run the mocha suite, eslint or prettier, because the npm dependencies could not be installed in my environment. Instead I ran the real test file against the real src modules under Node 22, with small local stand-ins for the third-party packages (mocha, chai, sinon, pino, es6-error and a few imports these tests never call). All 22 tests in test.extension-runners.js pass with this change. On master the two new tests fail with expected 0 to equal 1 on the error count, and each test also fails when its part of the change is reverted. I also ran the real FirefoxDesktopExtensionRunner, RemoteFirefox and RDP client against a local fake RDP server that rejects reload the way Firefox does for an invalid manifest ({ error: "Error", message: "Extension is invalid" }, the packet Actor.writeError sends). On master neither path printed anything. With this change both print the error shown above.

🤖 Generated with Claude Code