Code Review — findings & fixes

Reviewed lib/features/spaces/views/accord_home.dart (the only changed file). The fix is correct: PopScope(canPop: false) is the right tool here, the callback priority order (end-drawer → drawer → open drawer) matches the described behavior, and no Discord/Firebase code was re-introduced.

Findings ranked by severity

🔴 High — Missing test coverage (fixed in this review)

The new PopScope behavior had no tests. The three back-gesture cases and the core regression (back must not pop to sign-in) are now covered in a new file:

test/features/spaces/mobile_pop_scope_test.dart (commit 96d2aac)

Tests added:

1. Back when no drawer is open → opens the channel drawer
2. Back when the channel drawer is open → closes it
3. Back when the end drawer (members) is open → closes it
4. Back never pops the route back to the sign-in screen (regression guard)
5. Works correctly when there is no end drawer (no space selected)

The AccordHomeScreen widget has too many Riverpod/Hive dependencies to drive end-to-end in a unit test; the tests use a minimal StatefulWidget that replicates the same PopScope + Scaffold + GlobalKey structure so the drawer state machine can be exercised directly.

🟡 Medium — _MessagePane construction is duplicated (pre-existing, not introduced here)

The messageArea variable (used in the wide layout) builds _MessagePane with identical parameters to the inline construction in the narrow layout body. Any future change to those parameters must be applied in both places. This is a pre-existing issue outside the scope of this PR — noting it here for a follow-up refactor.

🟢 Low — Comment verbosity

The 7-line comment block before the PopScope is slightly longer than the project style recommends ("only add a comment when the WHY is non-obvious"). The WHY here (Android left-edge swipe == system back gesture) genuinely is non-obvious, so the comment is justified — but could be condensed to 3–4 lines. Not blocking.

Summary

One commit pushed to this branch: adds 5 widget tests covering all new behavior paths. No source-code changes were needed — the implementation is correct as written.

Generated by Claude Code