Revert 3 commits pushed directly to main

[beastoin]

Summary

• Reverts 3 commits that were incorrectly pushed directly to main instead of through a PR
• Reverted commits:
  • af8606f67 — Update test harness to match safe cast fix
  • a68365d7a — Fix safe cast for fair use stage and add refresh support
  • dd690961b — Bump app version to 784

Why

These changes should have gone through a PR. The fixes (safe cast, refresh support, evidence cleanup, version bump) will be re-submitted as a proper PR after this revert lands.

🤖 Generated with Claude Code

by AI for @beastoin

[greptile-apps]

Greptile Summary

This PR reverts three commits (dd690961b, a68365d7a, af8606f67) that were pushed directly to main without going through a PR, restoring the branch to the state before those changes. The reverted work (safe cast fix for fair use stage, refresh support for _loadFairUseStatus, version bump to 784) will be re-submitted as a proper PR.

Key points:

• The revert is mostly correct and matches the intent described in the PR description.
• The as String? ?? 'none' unsafe cast pattern is knowingly reintroduced as a temporary regression — ?? 'none' only handles null, not non-null non-String values, which would throw a TypeError. The follow-up PR is expected to address this.
• One small impurity: a new comment (// Using Future.wait to run both fetches concurrently) is added at usage_page.dart:584 that did not exist before the reverted commits, making this a slightly non-clean revert.
• _loadFairUseStatus() is removed from the pull-to-refresh handler as part of the revert, meaning fair use status will not update on pull-to-refresh until the follow-up PR lands.

Confidence Score: 4/5

• Safe to merge as a cleanup revert; the known unsafe cast regression is intentional and tracked for a follow-up PR.
• The PR achieves its stated goal of reverting three direct-to-main commits. The unsafe cast regression is acknowledged and temporary. The only unintentional change is a single explanatory comment that does not affect runtime behavior. The follow-up PR to re-land the fixes properly should be prioritized to avoid leaving the unsafe cast in production.
• app/lib/pages/settings/usage_page.dart — contains the unsafe cast and the extra comment introduced by the revert.

Important Files Changed

Filename	Overview
app/lib/pages/settings/usage_page.dart	Reverts safe is String cast guard to as String? ?? 'none' (known regression, follow-up planned) and removes _loadFairUseStatus() from pull-to-refresh; also adds an extra comment not present in the original code.
app/pubspec.yaml	Clean version bump revert from 784 back to 783 as expected.
app/test/widgets/usage_page_fair_use_banner_test.dart	Test harness reverted to match the production cast pattern; note there is no test covering a non-null, non-String stage value, which is the scenario that would expose the unsafe cast bug reintroduced by this revert.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Pull-to-Refresh triggered] --> B["Future.wait(\n  fetchUsageStats,\n  fetchSubscription\n)"]
    B --> C[Stats & Subscription updated]
    C --> D{Fair use status\nstale?}
    D -->|Yes - not refreshed| E[Banner shows old stage]
    D -->|No| F[Banner up to date]

    subgraph "After revert (this PR)"
        B
    end

    subgraph "Reverted commit a68365d7a"
        B2["Future.wait(\n  fetchUsageStats,\n  fetchSubscription,\n  _loadFairUseStatus\n)"] --> C2[All three updated]
    end

    G[_buildFairUseBanner called] --> H["_fairUseStatus!\['stage'\]\nas String? ?? 'none'"]
    H -->|null| I[stage = 'none' → hidden]
    H -->|non-null String| J[stage = value → shown]
    H -->|non-null non-String| K["⚠️ TypeError thrown"]

Loading

_{Last reviewed commit: "Revert "Bump app ver..."}

[greptile-apps]

Extra comment not present in original code

The comment // Using Future.wait to run both fetches concurrently was not in the codebase before the three commits being reverted. This means the revert is not a clean reversal — it introduces a net-new change. A pure revert should restore the exact prior state. Consider removing this comment to keep the diff minimal and the revert verifiably clean.

Suggested change

	// Using Future.wait to run both fetches concurrently
	await Future.wait([provider.fetchUsageStats(period: period), provider.fetchSubscription()]);
	await Future.wait([provider.fetchUsageStats(period: period), provider.fetchSubscription()]);

[greptile-apps]

Unsafe cast reintroduced by this revert

_fairUseStatus!['stage'] as String? ?? 'none' will only safely handle a null value — the ?? 'none' fallback is never reached for non-null, non-String values (e.g., an int or bool). Dart's as operator throws a TypeError at runtime when the actual type doesn't match. The reverted commit a68365d7a replaced this with the defensive rawStage is String ? rawStage : 'none' pattern specifically to guard against this.

This is noted in the PR description as a known regression that will be addressed in a follow-up PR, but it's worth flagging so reviewers are aware this is an active risk until that follow-up lands.