Repository navigation
fix(web): validate the focused keyboard control - #903
Conversation
There was a problem hiding this comment.
All reported issues were addressed across 5 files
Reply with feedback, questions, or to request a fix.
Re-trigger cubic
|
Thanks, the bug is real: with focus inside a shadow root the check read the host, so a focused button let typing through. Two asks: drop the skills/e2e/references/agent.md addition, and cut the agent.mdx change and the changeset to one sentence, e.g. "The focused element is checked inside shadow roots and frames, so focus on a button there is refused." |
Signed-off-by: Rudy Celekli <rudy@gradiahq.com>
Signed-off-by: Rudy Celekli <rudy@gradiahq.com>
e874d81 to
8ea8169
Compare
|
[gpt-6.1-sol] responding on behalf of rudycelekli Removed the skills entry and shortened the documentation and changeset wording. All four built CLI cases pass for focused buttons and fields inside frames and shadow roots. Direct package typecheck and lint pass. The full monorepo check remains blocked at the same missing filtered mobile-benchmark dependency on untouched main; the current full Next benchmark was not rerun with this lean dependency environment. Hosted fork workflow authorization is reported separately; I am not marking these Ready for Human Review. |
|
Thanks! I merged main and added |
flowchart LR A[Focused element] --> B[Descend shadow focus] B --> C{Editable control or custom widget?} C -->|yes| D[Dispatch typing] C -->|no| E[Refuse before keys]Verified
Ran it locally: yes
The real built CLI proof uses the documented
StepExecutorseam to call native actions, with no model calls. Main and branch use locked Playwright 1.63.0 Chromium, Node 24.18.0 and pnpm 12.3.4. Main/branch videos are retained locally; installed GitHub CLI 2.83.1 lacks--attach.The full web suite passes: 56 files, 513 tests and one skipped. The complete testbed passes: 14 files, 71 tests and one skipped. The complete Next benchmark suite passes: 41 files and 91 tests, reusing the unchanged production app built in this worktree for the preceding ARIA branch. Web build, typecheck, lint and docs validation pass. An independent production/native-test review found no remaining critical or important findings. Semantic reader and oracle expectations are unchanged.
Six paired
scripts/bench-ab.ts --suite testbed --no-build --min-pairs 6 --max-pairs 6 --warmup 0 -- tests/forms.e2e.tsruns match every step, outcome and counter. Existing forms scenarios verify unaffected behavior; the new native fixtures prove the focus guard. No performance claim on the shared host.Local
pnpm checkpasses lint then stops at a mobile benchmarkreact-native-screensdead-code finding with Expo dependencies excluded from the filtered installation. The same finding reproduces on main with that installation; this does not establish a source defect. Full top-levelpnpm test,pnpm test:testbedandpnpm test:web-benchmarkwrappers were not run; the complete touched-package, testbed and Next suites above were run against built packages.CI status
The four workflows on this head require repository owner approval before execution:
Updated verification (October 6)
Removed the skills entry and shortened the documentation and changeset wording. All four built CLI cases pass for focused buttons and fields inside frames and shadow roots.
Direct package typecheck and lint pass. The full monorepo check remains blocked at the same missing filtered mobile-benchmark dependency on untouched main; the current full Next benchmark was not rerun with this lean dependency environment. Hosted fork workflow authorization is reported separately; I am not marking these Ready for Human Review.