robustness tests

Robustness tests

Status: in progress.

Property tests that look for panics and wrong behavior in the pure logic of the browser shell. No GUI is needed.

How it works

• proptest makes random input. Each test names the requirement it checks in a Req: comment. The sources are the ADR, the IPC contract and the Browser shell page.
• src-tauri/tests/robust_nav.rs: the host predicate, the address bar, the engine rules, and the planner that the gatekeeper uses. One test checks that the L1, L3 and L4 layers agree on every host.
• src-tauri/tests/robust_gatekeeper.rs: the gatekeeper with real loopback sockets and a fake router, in its own test process so its sockets cannot disturb the gatekeeper unit tests. Random and malformed request heads, CONNECT lines, huge heads, partial reads, slow and half-closed clients, and eight concurrent slow images.
• src-tauri/tests/robust_core.rs: random command sequences on the state machine, back and forward, and a soak test of 1 000 tabs.
• src-tauri/tests/robust_stores.rs: random bookmark and history commands, corrupt files, the 10 000-entry history cap, import and export.

How to use / run locally

• cargo test --manifest-path src-tauri/Cargo.toml --test robust_nav --test robust_core --test robust_stores --test robust_gatekeeper
• A failing property prints the smallest input that fails.

Limits

• The shell glue (webviews, the main loop) is not covered. Check by hand on each OS: open a second content tab, toggle site JavaScript, and close and reopen a tab. The window must stay responsive.
• The gatekeeper tests use a fake router, not a real I2P router.

History

• 2026-10-03 — Add the robustness tests — #52