Skip to content

Assert what a test's own setup guarantees, instead of skipping on it - #393

Merged
adamjohnwright merged 3 commits into
mainfrom
fix/321-self-skips
Oct 10, 2026
Merged

adamjohnwright merged 3 commits into
mainfrom
fix/321-self-skips

Conversation

@adamjohnwright

Copy link
Copy Markdown
Contributor

Refs #321 (self-skips that hide the behaviour under test).

Some tests skipped when a condition didn't hold that their own fixed pathway and window guarantee. A skip like that reports green having checked nothing. Each one is now an assertion. In CI's replay none of these ever skipped, and expected-skips.json already fails a new skip there, so this matters most for runs against a live site.

Each was shown failing where the old test skipped:

Spec Condition Old vs new, with the condition broken
back-button, content-page-urls the pathway offers a Molecule tab With the tab renamed, the old tests skip and the new ones fail with "this pathway offers a Molecule tab"
hierarchy-scroll (×2) the tree scrolls; rows sit above the fold In a 6000 px tall window, the old tests skip and the new ones fail with "the hierarchy scrolls"
interactor-threshold, says so, and offers the way to it the pathway opens zoomed out past 0.6 In a 4000×3000 window, the old test skips and the new one fails ("Received: 0.91")
interactor-threshold, raising it removes… the entity offers interactions same pattern
interactor-threshold, comes back with the Back button the tree click goes to another pathway It used to read the URL after a fixed 8 s; now it waits for the URL to change, then asserts

The fixed pauses after choosing a tab are now waits on the URL.

One skip stays, because a third party decides it: whether a second interaction resource holds anything for this pathway. The PSICQUIC servers answer in 6 to 17 s each, and the old test read their counts once, so it raced them. It now waits up to a minute for a count before it skips.

Tests: back-button and content-page-urls (6), hierarchy-scroll (2) and interactor-threshold (12) all pass.

🤖 Generated with Claude Code

adamjohnwright and others added 3 commits October 10, 2026 02:49
#321's self-skips: tests that skipped when a condition their fixed
pathway and window guarantee did not hold, which reports green having
checked nothing. Each is now an assertion, shown failing where the old
test skipped:

- back-button and content-page-urls: the Molecule tab. With the tab
  renamed, the old tests skip; the new ones fail by name. The fixed
  pauses after choosing a tab are now waits on the URL.
- hierarchy-scroll: that the tree scrolls, and that rows sit above the
  fold. With a 6000px-tall window the old tests skip; the new fail on
  'the hierarchy scrolls'.
- interactor-threshold: the entity offers interactions; the pathway
  opens zoomed out past the badge threshold (in a 4000x3000 window the
  old test skips, the new fails, 'Received: 0.91'); and the tree click
  leads elsewhere, now waited for instead of read after 8s.

One stays a skip, because third-party servers decide it: whether a
second interaction resource holds anything here. It now waits up to a
minute for their counts instead of reading them once, which raced them.

Refs #321

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Review of #393: the comment said finding no second resource is a skip,
not a failure. True against a live site; in CI the counts are replayed
and an unlisted skip fails the shard.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@adamjohnwright
adamjohnwright merged commit 3d1f354 into main Oct 10, 2026
6 checks passed
@adamjohnwright
adamjohnwright deleted the fix/321-self-skips branch October 10, 2026 03:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant