Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

Partial-render residual: chrome mounts but content does not, so gaps.length==0 yields a false fixed

Open
  • Is something wrong?
#186 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
38/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Quiet
Tech stack
playwright, typescript
Domain
accessibility, testing-qa

Research direction

Start by reading markUnseenScreens, runProviders, and the gate that builds cleanlyScannedScreens; the issue describes the reproduced case but names no file or test. Reproduce the partial render where chrome mounts and content does not, then weigh a deterministic content or control signal for flagging it unseen and adding a gap. Done means the affected screen cannot produce a false fixed verdict, without making results depend on timing or DOM order.

Written by the indexing model from the issue text.

Description

area: engine bug

What this is

An accepted, documented residual of the #164 fix (no fixed on unscanned screens), filed so it
is visible rather than living only in a code comment. It is invariant-5 surface (a false green),
and it is a pre-existing weakness of gaps.length === 0 as a proxy for "this screen was measured."
#164 narrows the false-fixed population and does not widen it, so this is tracked, not a blocker.

The residual

cleanlyScannedScreens is built from scanned screens with gaps.length === 0. The gate treats
membership in that set as "the screen was genuinely measured this run," and only there does it mark
a disappeared floored identity fixed.

A screen whose chrome mounts but whose content region silently fails to mount defeats the proxy.
The keyboard walk lands on a real control, so isBodyOnly does not fire. No provider is denied or
throws, so runProviders records no gap. The affected content never renders, so its floored
identity is absent from the drafts. The screen therefore has zero gaps, enters
cleanlyScannedScreens, and the floored identity on the missing content is claimed fixed. That
is "absence of evidence read as evidence of absence," surviving at a narrower scope than before
#164.

Reproduced

Driven end to end through the real makeCheckRunner, makeConcreteStepRunner, markUnseenScreens,
and the gate during the #164 security review:

scan gaps: 0  stops: 1  drafts: 0
flagged unseen: []   gaps after unseen pass: 0
cleanlyScannedScreens: [ 'orders' ]
FALSE FIXED on partially-rendered screen: 1  [ 'orders/button-name' ]

Why it is not fixed in #164

The signal that would catch it, a genuine "content mounted" marker, is one the gate does not have.
markUnseenScreens detects only body-only blanks and duplicate renders; runProviders gaps only
on denied or thrown providers. A partial render with surviving chrome trips none of them. Closing
it needs a partial-render detector in markUnseenScreens, which is its own change, not a tweak to
the #164 guard.

Before #164 the gate marked fixed for any absent floored identity with no scan check at all, so a
completely unscanned floored screen got a false fixed. #164 excludes unscanned screens, gapped
screens, and body-only blanks. This partial-render case is the strict subset that remains.

Fix direction, if it is worth the cost

Add a partial-render detector to markUnseenScreens: a screen that reports a control count or a
content-region signal below a floor for its surface is flagged unseen and gets a gap, which removes
it from cleanlyScannedScreens. Whatever the detector, it must be deterministic, because a verdict
that depends on it must not depend on timing or DOM iteration order. This reintroduces some of the
per-application coupling the heuristics exist to avoid, so it is a tradeoff to weigh, not an obvious
win. This issue keeps it visible.

Origin

Found and reproduced in Chromium during the #164 security review. Accepted residual, not a blocker
for #164.

Dominant language
TypeScript
Stars
0
Forks
0
Avg merge
2h 48m
Merged PRs (30d)
8

Getting set up

  • No Dockerfile or Docker Compose file
  • Has a pull request template
  • No contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from usabl-dev/usabl

All issues in usabl-dev/usabl

Similar issues

More TypeScript issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.