feat(core): restack bounded BiDi pointer click

[seonghobae]

Partial implementation of #28. This Draft owns deterministic primary-button click serialization and remains inert until a trusted adapter binds current session/context/origin/document/policy authority.

Current non-force stack adoption

Current exact head is 881c7f09ee9161ce8664dd75226938ecf60b85e5, an ordinary tree-identical two-parent topology adoption over predecessor ced4a851ca66c08d895a098725c7f0ad3ecf0c38 and repaired parent #255 6865faa8185c5fe6a8b6aaba9535d774575b0061. No source, test, docs, workflow, dependency, gate or authority changed in this adoption. The predecessor pointer-click tree and its prior CI/MV3 successes remain evidence for that exact predecessor only.

The repaired parent skips the unsafe #253 caller-supplied operational-completion intermediate while preserving #253/#254 valid deltas and #255 connection-provenance repair. This child therefore keeps the safe successor chain without force-push or destructive rebase.

Fresh exact-head CI 34003259914 completed success and exact-head Manifest V3 Compatibility 34003259888 completed success on 881c7f09ee9161ce8664dd75226938ecf60b85e5. These results validate this exact adopted tree only; they do not establish browser action/post-condition evidence, protected-main integration, central review/security acceptance, or release readiness.

The existing pointer-click command remains inert transport data: successful serialization is not browser action, process/profile cleanup, navigation, or observed post-condition evidence. Keep Draft; parent-first protected integration and all required review/security/ruleset gates remain independent. No workflow/ruleset/secret mutation, self-approval, bypass, force update, merge, tag, release or publication was performed.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[seonghobae]

Exact-head repair review: predecessor 172a18a... produced a real repository RED, not a queue artifact. Rust contracts were GREEN while exact production coverage failed on the sole impossible success + missing command id branch. This head removes that representable-invalid state by retaining validated routing structurally (CommandSuccess { command_id }, nullable-id CommandError, Event) and exposes the routing only crate-privately for correlation. Public envelope behavior and browser/policy authority are unchanged. Fresh CI/MV3 on this exact head is still required; no predecessor result is transferred.

[seonghobae]

Parent #255 has advanced by normal fast-forward from 626026c5037ee23e5b435bd864e7ad4280da5643 to exact 88e68f37d655ae62dae6a38a50c12b30e2801160 after hosted exact coverage exposed the common-envelope success/missing-id representation defect. This #256 head is therefore stale-parent evidence even though it already contains a compatible structural routing repair plus its unique pointer-click delta. Do not transfer #256 checks or reconstruct descendants yet. Keep Draft; after #255 88e68f37... reaches terminal exact-head verification, non-destructively restack #256 onto that verified parent and preserve only its unique PointerClick command slice and still-valid evidence, without reintroducing duplicate parent repair.

[seonghobae]

Parent/base reconstruction completed for exact head f4f5bb71b54f3df3887d79b52f361e5c8e2d30e7.

I merged the current #255 head (d60e98679f09cc067d46fae1170b464d53c0951f) without rewriting history. The bounded pointer-click slice remains inert: it serializes a fixed primary-button input.performActions sequence only from already admitted identifiers and does not bypass the still-required current session/context/origin/document, policy, approval, or transport binding. I added the missing buyer-visible changelog entry.

Exact-tree local evidence: documentation contracts 22 passed / 177 subtests; rustfmt clean; strict all-target/all-feature Clippy clean; full workspace tests passed; rustdoc with warnings denied passed; pinned-nightly LLVM production functions, lines, regions, and branches are 100% covered. Hosted exact-head checks remain authoritative.

[seonghobae]

Current-head lineage review: live PR metadata is base=d60e98679f09cc067d46fae1170b464d53c0951f, head=f4f5bb71b54f3df3887d79b52f361e5c8e2d30e7, with 8 ahead / 0 behind and five unique semantic paths. The earlier d24b8dc... CI/MV3 GREEN is predecessor evidence after this non-destructive parent refresh and must not be promoted to current-head GREEN. Current CI 33898817836 and MV3 33898817825 remain non-terminal. #247 has also moved again to the no such client window test-only generation, so parent-first reconstruction remains required after that canonical ancestor reaches exact GREEN.

[seonghobae]

Exact-head parent refresh: bd1f5ac60a76d2edb35e63095d406e53bc43931f now adopts #255 f8edec38cf8ab7fde22b8d1de9305728c1a2f25b without conflict. Rustfmt, all-feature/all-target check, strict Clippy, rustdoc, 141 Python contracts, and CI-equivalent pinned-nightly 100% production coverage all passed. Hosted checks remain authoritative; the PR stays Draft.

[seonghobae]

Non-force restacked on PR #255 exact head a13de5f9321e72c1867974eb7a43230f031e58df; PR #256 is now exact head 9f2e6f29be46371762e3031a97c1cac04720694f. Verification: rustfmt, all-feature/all-target check, strict Clippy, rustdoc, 141 Python contracts, and CI-equivalent pinned-nightly production function/line/region/branch coverage at 100%. Fresh hosted exact-head checks and ordered parent integration remain required.

[seonghobae]

Reviewed the five-file parent-relative change at 9f2e6f2 against a13de5f. No additional defect found in this bounded pointer-click serialization change. It reuses the existing JSON string encoder and protocol-text validation, enforces the js-uint command-id and context byte bounds, and emits one fixed mouse move/down/up sequence. The element origin/sharedId and performActions envelope agree with the current W3C Working Draft (3 September 2026), §§7.9.2.1 and 7.9.3.1: https://www.w3.org/TR/2026/WD-webdriver-bidi-20260903/#command-input-performActions .

Rust 1.97.1 formatting, all four focused pointer-click tests, strict core Clippy with all targets/features, and warning-free core rustdoc passed locally. This inert serializer grants no policy authority, performs no browser input and proves no click post-condition. Keep Draft: parent #255's connection-evidence finding remains unresolved, and current hosted checks, complete coverage, ordered parent integration and policy gates remain independent requirements. This COMMENT is not an eligible non-author approval.

[seonghobae]

Current-parent repair finding on exact 9f2e6f29be46371762e3031a97c1cac04720694f: live #255/base branch has advanced to 3e7057443d7c9532ff526acb5eefe8cd4778c767. Fresh comparison is 10 child commits ahead / 61 parent commits behind, with merge base a13de5f9321e72c1867974eb7a43230f031e58df. The valid child delta remains five paths. Two paths overlap with the parent generation: CHANGELOG.md and crates/originweave-network/src/webdriver_bidi_command_correlation.rs. The latter must retain #255's connection-generation provenance machinery and add only the child-owned PointerClick command kind; the pointer-click CHANGELOG record must also survive. This is a repair/restack finding, not authorization to close or discard the child. Current 9f2e... CI/MV3 success is predecessor-tree evidence and cannot transfer to the future current-parent adoption. Keep Draft; ordinary non-force adoption of exact #255 is required before #257 follows this branch.