fix: the real-fit strengths test read a gitignored snapshot #39

Merged
kristofferopsahl merged 1 commit from fix/strengths-test-needs-no-local-snapshot into feat/captaincy-board 2026-09-18 23:44:03 +02:00
Collaborator

Targets feat/captaincy-board (#31), not main.

test_the_real_fit_reads_the_results_it_is_given loaded its clubs from data/pit/20260913T180117Z/bootstrap.json. data/pit/ is gitignored, so the test passed on the machine that took that snapshot and raised FileNotFoundError on every clean checkout. It failed in 9 of #31's browser runs, and in 4 of them it was the only failure.

The clubs now come from the committed 2025/26 archive the test already loads. The fit is still unstubbed.

Checked on a clean checkout with no data/pit/:

  • tests/web/test_strengths_follow_results.py: 15 passed.
  • Still bites: with strengths_deps ignoring results_path, the test fails ("40 match(es)" read from the real file instead of the 1-match file it was given).

Not touched: test_header_narrow[400]. It fails on #31's head every time, locally too ("the page is 410px wide on a 400px screen"). It passed on earlier #31 runs, so a recent commit on the branch made the page 10px wider than a 400px phone. That's a real regression for #31 to fix, not noise.

The CSS gate flake that also shows up on #31 is fixed in #38 (against main). #31 gets it when it merges main.

🤖 Generated with Claude Code

Targets `feat/captaincy-board` (#31), not `main`. `test_the_real_fit_reads_the_results_it_is_given` loaded its clubs from `data/pit/20260913T180117Z/bootstrap.json`. `data/pit/` is gitignored, so the test passed on the machine that took that snapshot and raised `FileNotFoundError` on every clean checkout. It failed in 9 of #31's browser runs, and in 4 of them it was the only failure. The clubs now come from the committed 2025/26 archive the test already loads. The fit is still unstubbed. **Checked on a clean checkout with no `data/pit/`:** - `tests/web/test_strengths_follow_results.py`: 15 passed. - Still bites: with `strengths_deps` ignoring `results_path`, the test fails ("40 match(es)" read from the real file instead of the 1-match file it was given). **Not touched: `test_header_narrow[400]`.** It fails on #31's head every time, locally too ("the page is 410px wide on a 400px screen"). It passed on earlier #31 runs, so a recent commit on the branch made the page 10px wider than a 400px phone. That's a real regression for #31 to fix, not noise. The CSS gate flake that also shows up on #31 is fixed in #38 (against `main`). #31 gets it when it merges `main`. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
fix: the real-fit strengths test read a gitignored snapshot
Some checks failed
CI / syntax (pull_request) Successful in 9s
CI / tests (pull_request) Successful in 1m12s
CI / wheel (pull_request) Successful in 29s
CI / browser (pull_request) Failing after 18m14s
34bffd8d1b
test_the_real_fit_reads_the_results_it_is_given loaded its clubs from
data/pit/20260913T180117Z/bootstrap.json. data/pit/ is gitignored, so the
test passed on the machine that took that snapshot and raised
FileNotFoundError on every clean checkout: it failed in nine of PR #31's
browser runs, and was the only failure in four of them.

The clubs now come from the committed 2025/26 archive the test already
loads. The fit is still unstubbed, and still proved to bite: with
strengths_deps ignoring results_path the test fails, reading the real
40-match file instead of the 1-match one it was given.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
kristofferopsahl merged commit e824f6cac6 into feat/captaincy-board 2026-09-18 23:44:03 +02:00
kristofferopsahl deleted branch fix/strengths-test-needs-no-local-snapshot 2026-09-18 23:44:03 +02:00
Sign in to join this conversation.
No description provided.