[PM-40995] fix: Allow fill-assist to rescue autofill requests when heuristics find no fields

[aj-rosado]

🎟️ Tracking

https://bitwarden.atlassian.net/browse/PM-40995

📔 Objective

AutofillParserImpl.parseInternal returned AutofillRequest.Unfillable immediately whenever the built-in heuristic field-classifier found zero non-Unused fields on a page — before the app ever checked whether fill-assist (server-curated CSS-selector-based field targeting) was enabled or had matching rules for that host. This meant fill-assist could never rescue a page the heuristics failed to parse, defeating its purpose on pages designed to evade naive heuristics (e.g. generically-named, hint-less input fields). Confirmed via a live repro against a test page with no autofill suggestion offered at all.

This PR:

• Lets parseInternal fall back to an Unused-inclusive candidate view when heuristics find nothing, but only when fill-assist is actually enabled (feature flag + setting) — the disabled path is unchanged in both behavior and cost.
• Gives toEffectiveViews's Unused case (previously dead code, since a heuristic-classified focused view could never be Unused) real behavior: it checks against both login and card fill-assist categories, since heuristics gave no signal for which one applies.
• Adds a conservative guard so fill-assist is only trusted for the Unused case when it actually matches a field on the page — an empty match falls back to heuristics instead of being treated as authoritative, so an unrelated host rule (e.g. one covering a different pathname/category on the same host) can't silently claim a page it has no real match for. The pre-existing Card/Login branches keep their existing "authoritative even if empty" behavior since those already carry real heuristic confirmation.
• Fixes a pre-existing bug in the shared test setup()'s default toAutofillViewData mock stub, uncovered by the new tests: it used firstArg()/secondArg() for the autofillId/website params, but since it mocks an extension function, the receiver occupies argument position 0 at the JVM level, so the stub was assigning the ViewNode itself into the autofillId field. Fixed by shifting to secondArg()/thirdArg().
• Adds test coverage for the rescue path, the conservative empty-match guard, and confirmation the rescue path stays inert when fill-assist is disabled.

[github-actions]

🤖 Bitwarden Claude Code Review

Overall Assessment: APPROVE

Reviewed the fill-assist rescue path added to AutofillParserImpl.parseInternal, the new toEffectiveViews handling for the Unused case, the supporting refactors (selectCandidateAutofillViews, firstFocusedOrNull), and the accompanying test changes. This is a security-sensitive autofill change at a trust boundary, so the analysis focused on whether the new path could cause credentials to be offered on the wrong page.

The rescue path is well-guarded: it activates only when the feature flag and user setting are both enabled, host rules exist for the resolved host, and fill-assist actually matches a field on the page. The conservative empty-match guard for the Unused case correctly prevents an unrelated host rule from silently claiming a form it has no real match for, while preserving the existing "authoritative even if empty" behavior for the heuristic-confirmed Card/Login branches. The test-stub fix (firstArg/secondArg → secondArg/thirdArg for a mocked extension function) is correct, and the three new tests cover the rescue path, the empty-match guard, and the disabled-feature inert path.

Code Review Details

No findings. The change is well-reasoned, conservatively guarded, and adequately tested.

Note: Codecov reports 0% patch coverage, but this appears to be an artifact of the heavy extension-function mocking in these tests rather than a genuine coverage gap — the new tests do exercise parseInternal end to end.

[codecov]

Codecov Report

❌ Patch coverage is 96.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 85.42%. Comparing base (845e20b) to head (f62b943).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
...twarden/data/autofill/parser/AutofillParserImpl.kt	96.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #7214      +/-   ##
==========================================
- Coverage   86.19%   85.42%   -0.78%     
==========================================
  Files         888      962      +74     
  Lines       64776    66801    +2025     
  Branches     9691     9745      +54     
==========================================
+ Hits        55836    57065    +1229     
- Misses       5470     6260     +790     
- Partials     3470     3476       +6     

Flag	Coverage Δ
app-data	17.42% <96.00%> (-0.29%)	⬇️
app-ui-auth-tools	18.61% <0.00%> (-0.14%)	⬇️
app-ui-platform	16.66% <0.00%> (+0.14%)	⬆️
app-ui-vault	27.89% <0.00%> (+0.56%)	⬆️
authenticator	6.13% <0.00%> (+<0.01%)	⬆️
lib-core-network-bridge	4.12% <0.00%> (-0.01%)	⬇️
lib-data-ui	1.18% <0.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[david-livefront]

Since this is a multi-line statement, can we make sure it is wrapped in curly braces:

is AutofillView.Unused -> {
    rule.category in LOGIN_FILL_ASSIST_CATEGORIES ||
        rule.category in CARD_FILL_ASSIST_CATEGORIES
}