fix(test): make tailwind integration test rerun-safe

[masenf]

Two compounding bugs caused the (redis, 3.14, split 2) integration job to occasionally fail at test_tailwind_app[tailwind_v4]:

1. The test wrote its sentinel stylesheet to Path(file).resolve().parent.parent / "assets". On a rerun triggered by pytest-rerunfailures, importlib.reload reuses the original module spec, so file keeps pointing at the previous tmp_path while the compiler reads stylesheets from Path.cwd() / "assets" in the new tmp_path. The mismatch raised FileNotFoundError at setup for every rerun. Write to Path.cwd() directly so the location matches the compiler regardless of reload-induced file staleness.

2. The v4 color assertion read value_of_css_property without polling. In dev mode the @tailwindcss/postcss step can finish painting after the element appears, so under runner load the first attempt intermittently saw the default color and failed; that failure is what kicked off the buggy reruns above. Poll until the expected color is applied so first-attempt timing variance no longer fails the test.

[greptile-apps]

Greptile Summary

This PR fixes two bugs that caused the tailwind integration test to fail on reruns triggered by pytest-rerunfailures. The first bug used a stale __file__ path after importlib.reload, writing the sentinel stylesheet to the wrong directory; it is fixed by switching to Path.cwd(). The second bug asserted the v4 color synchronously without polling, causing intermittent failures when the Tailwind PostCSS step finished after the element appeared.

• Asset path fix: assets = Path.cwd() / \"assets\" ensures the stylesheet is always written to the directory the compiler reads from, regardless of module reload state.
• Polling fix: The color assertion for both v3 and v4 now uses poll_for_or_raise_timeout, waiting until the expected utility-class color is applied in the browser.

Confidence Score: 5/5

The change is safe to merge; it corrects two independent test-reliability issues without touching production code.

Both fixes are narrowly scoped to the integration test file. The path change from Path(__file__) to Path.cwd() directly matches how the harness sets up its working directory, and the polling loop mirrors the existing pattern used elsewhere in the test suite. No logic that could introduce a regression was touched.

No files require special attention.

Important Files Changed

Filename	Overview
tests/integration/test_tailwind.py	Fixes two rerun-safety bugs: replaces Path(__file__) with Path.cwd() for stylesheet placement, and polls for CSS color instead of asserting it immediately in tailwind v4.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[pytest runs test_tailwind_app] --> B[AppHarness.create with tmp_path]
    B --> C["TailwindApp() called via importlib"]
    C --> D["Write stylesheet to Path.cwd() / assets/\n(compiler's working directory)"]
    D --> E[App compiles with Tailwind]
    E --> F[Browser renders page]
    F --> G[poll_for_or_raise_timeout: find #p-content]
    G --> H{tailwind_version?}
    H -- 0 disabled --> I[assert color NOT in v3 colors]
    H -- 3 or 4 --> J["poll_for_or_raise_timeout:\nwait until color in expected_colors"]
    J --> K[Color matches - test passes]
    I --> K
    K --> L[Assert external stylesheet color]

Loading

_{Reviews (1): Last reviewed commit: "fix(test): make tailwind integration tes..." | Re-trigger Greptile}

[codspeed-hq]

Merging this PR will not alter performance

✅ 24 untouched benchmarks
⏩ 2 skipped benchmarks¹

---

_{Comparing claude/debug-integration-test-68Vvy (5001c06) with main (1eed933)}

Footnotes

1. 2 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩