refactor: extract oasis-app main.rs monolith into 7 modules

[AndrewAltimit]

Summary

• Extract the ~1,468-line main.rs (with a ~970-line main() function) into 7 focused modules, reducing main.rs to ~262 lines
• New modules: app_state.rs (Mode enum + AppState struct), input.rs (per-mode input dispatch), commands.rs (terminal command output + remote polling), render.rs (SDI scene graph update), launch.rs (deduplicated app launch logic), terminal_sdi.rs (terminal/wallpaper/media SDI helpers), vfs_setup.rs (demo VFS population + tests)
• No behavioral changes -- pure structural refactoring with all existing tests passing

Test plan

• cargo test --workspace -- all 33 test suites pass
• cargo clippy --workspace -- -D warnings -- clean
• cargo fmt --all -- --check -- clean
• cargo build --release -p oasis-app -- builds successfully

Generated with Claude Code

[github-actions]

Gemini AI Code Review

Issues (if any)

• [BUG] crates/oasis-app/src/input.rs:37,42 - Terminal output won't update after OSK usage
  • In handle_osk_input, status messages are pushed to state.output_lines without calling commands::trim_output.
  • Since setup_terminal_objects only displays the first 12 lines of the Vec, and trim_output is responsible for shifting newest lines to the front by removing index 0, failing to trim will cause the terminal to permanently display the first 12 lines ever recorded.
  • Fix: Call commands::trim_output(&mut state.output_lines) after pushing OSK status messages.

Previous Issues (for incremental reviews)

(none)

Suggestions (if any)

• crates/oasis-app/src/input.rs:222,244,334,441 - Deduplicate app launch and transition logic
  • The sequence of launching an app window (or terminal), applying the launch result to state.mode, and triggering a fade-in transition is repeated identically in four different places.
  • Suggest extracting this into a helper like AppState::launch_app(&mut self, app: &AppEntry).

Notes

• 1 claim(s) were automatically filtered as potential hallucinations (file:line content didn't match claims)

• Excellent refactoring of a complex monolith. Splitting into focused modules (input, render, commands, etc.) significantly improves maintainability.

• Borrow splitting is handled cleanly by maintaining vfs and sdi as separate locals in main.rs while grouping the rest of the application state into the AppState struct.

• Verification: 10/10 unit tests in vfs_setup.rs passed successfully.

---

Generated by Gemini AI (gemini-3-flash-preview). Supplementary to human reviews.

[AndrewAltimit]

Review Response Agent (Iteration 1)

Status: No changes needed

Fixed Issues

• [BUG] crates/oasis-app/src/input.rs:37,42 - Terminal output not trimmed after OSK usage (Gemini): Added commands::trim_output(&mut state.output_lines) after both OSK confirmation and cancellation pushes. Without this, if output_lines grew beyond MAX_OUTPUT_LINES (12), the terminal would display the oldest 12 lines instead of the newest, since setup_terminal_objects reads output_lines[0..12] and trim_output removes from the front.

Ignored Issues

• Suggestion: Deduplicate app launch and transition logic (Gemini): This is a style/refactoring suggestion, not a bug. The repeated launch+transition pattern at lines 222/244/334/441 could be extracted into a helper, but the PR already introduced launch::launch_app_window and launch::apply_launch to consolidate the core logic. The remaining repetition is the 3-line transition creation which is tightly coupled to caller context. Not worth addressing in this PR.

Deferred to Human

• (none)

Notes

• The OSK trim bug was pre-existing in the original monolithic main.rs (not a regression from this refactor), but was correctly identified by Gemini and is a legitimate fix.
• All 1,828 tests pass, clippy is clean with -D warnings.
• Codex review produced no output (CLI error/empty response).

---

The agent reviewed feedback but determined no code changes were required.

[github-actions]

Gemini AI Incremental Review

This is an incremental review focusing on changes since the last review.

Issues

(none)

Previous Issues

• [RESOLVED] crates/oasis-app/src/input.rs:37,42 - Terminal output won't update after OSK usage

  • commands::trim_output is now correctly called after both confirmation and cancellation of the OSK.

• [STILL UNRESOLVED] crates/oasis-app/src/commands.rs:20,105,135 - Terminal output not trimmed after command execution

  • Several functions (process_command_output, apply_skin_swap) push to state.output_lines but do not call trim_output.
  • While the caller in input.rs:404 calls trim_output, other potential callers or internal logic paths (like apply_skin_swap pushing a success message) might bypass this, leading to the terminal buffer potentially exceeding its display limit and showing stale data.
  • Fix: Call trim_output(&mut state.output_lines) within process_command_output and apply_skin_swap after pushing new lines.

Suggestions

• crates/oasis-app/src/input.rs:218,270,431 - Deduplicate app launch and transition logic
  • The sequence of cloning an app entry, calling launch::launch_app_window, launch::apply_launch, and setting up a make_transition is repeated identically in three places.
  • Suggest extracting this into a helper method on AppState or a standalone function in launch.rs.

Notes

• The refactoring successfully broke down the 1.4k line monolith into manageable modules.
• The AppState struct effectively manages borrow splitting for the main application loop.
• The OSK trimming fix addresses a critical UI staleness bug.

---

Generated by Gemini AI (gemini-3-flash-preview). Supplementary to human reviews.

[AndrewAltimit]

Review Response Agent (Iteration 2)

Status: No changes needed

Fixed Issues

• (none)

Ignored Issues

• [STILL UNRESOLVED] crates/oasis-app/src/commands.rs:20,105,135 - Terminal output not trimmed after command execution (Gemini): Not a real bug. process_command_output and apply_skin_swap are only called from one place (input.rs:398-401), and that call site already calls commands::trim_output(&mut state.output_lines) on line 404, after both functions complete. Gemini's concern about "other potential callers or internal logic paths" bypassing trim is speculative — no other callers exist. Adding redundant trim_output calls inside these functions would be unnecessary.
• Suggestion: Deduplicate app launch and transition logic (Gemini): Same style suggestion from iteration 1, already reviewed and intentionally declined by admin. The PR already consolidated core logic into launch::launch_app_window and launch::apply_launch. The remaining 3-line transition pattern is tightly coupled to caller context.
• Codex review: No output produced (CLI error), no issues to address.

Deferred to Human

• (none)

Notes

• All issues from the incremental Gemini review are either already resolved (OSK trim — fixed in iteration 1) or not actual bugs (trim after command execution).
• The codebase correctly applies the "trim once at the call site" pattern rather than trimming redundantly inside helper functions.

---

The agent reviewed feedback but determined no code changes were required.