refactor(paint): extract transforms/stacking/clip_path helpers from paint/mod.rs

[AndrewAltimit]

Summary

Splits three cohesive helper clusters out of the 2957-line crates/oasis-browser/src/paint/mod.rs into dedicated sibling modules:

• paint/transforms.rs -- transforms_have_3d, resolve_transform_origin, resolve_perspective_origin, compute_transform_matrix, compute_transform_offsets (+ unit tests for each).
• paint/stacking.rs -- creates_stacking_context, creates_compositing_layer, is_positioned.
• paint/clip_path.rs -- resolve_clip_path_rect, intersect_rects. The 3 clip-path tests that lived in mod.rs move with it, and 2 new intersect_rects tests are added (the helper had none).

paint/mod.rs keeps these names available via pub(crate) use re-exports, so paint/record.rs (the only sibling caller, via super::) needs no edits at the call site. paint_box's preserve3d_child_z helper stays in mod.rs because it threads PaintContext and is called only from the recursive walker in the same file.

Result: paint/mod.rs shrinks 2957 -> 2584 lines (-373). The remaining bulk is the paint_box recursive walker plus its integration tests.

Test plan

• cargo build --workspace
• cargo test -p oasis-browser --lib -- 1954 passed, 0 failed
• cargo clippy --workspace -- -D warnings -- clean
• cargo fmt --all -- --check -- clean
• Screenshot regression unchanged (CI will verify)

Generated with Claude Code

[github-actions]

Claude AI Security & Correctness Review

Issues (if any)

(none)

Previous Issues (for incremental reviews)

(none)

Suggestions (if any)

• crates/oasis-browser/src/paint/transforms.rs:119 - compute_transform_matrix is pub(crate) but is not included in mod.rs's pub(crate) use transforms::{...} re-export list. This is functionally fine (callers can reach it via super::transforms::compute_transform_matrix), but it's inconsistently published compared to its four siblings. If any future caller in record.rs [UNVERIFIED - File does not exist and is not in PR] uses the flat super:: path, it will be a compile error. Either add it to the re-exports for consistency, or make it pub(super) to signal it's internal to the transforms module.

Notes

• 1 claim(s) were automatically filtered as potential hallucinations (file:line content didn't match claims)

• This is a mechanical extraction refactor. The logic in all three new modules (clip_path.rs, stacking.rs, transforms.rs) is byte-for-byte identical to what was in mod.rs; no behavioral change is possible.

• The visibility promotions (fn → pub(crate) fn for resolve_clip_path_rect and intersect_rects) are correct — they are now called across module boundaries via the re-export.

• The m.e as i32 / m.f as i32 casts in compute_transform_offsets (NaN → 0, saturating on overflow) are pre-existing behavior, not introduced here.

---

Generated by Claude AI (sonnet). Supplementary to human reviews.

[github-actions]

Openrouter AI General Review

Issues (if any)

• [BUG] crates/oasis-browser/src/paint/mod.rs:31 - compute_transform_matrix is missing from the pub(crate) use transforms::{...} re-exports.
  • The PR states record.rs [UNVERIFIED - File does not exist and is not in PR] calls helpers via super::. Without this re-export, any call to super::compute_transform_matrix will fail to compile.
  • Add compute_transform_matrix to the re-export list in mod.rs.

Previous Issues (for incremental reviews)

(none)

Suggestions (if any)

• crates/oasis-browser/src/paint/transforms.rs:119 - If compute_transform_matrix is only consumed by compute_transform_offsets and record.rs [UNVERIFIED - File does not exist and is not in PR], consider marking it pub(super) instead of pub(crate) to accurately reflect its module boundary scope.

Notes

• 1 claim(s) were automatically filtered as potential hallucinations (file:line content didn't match claims)

• The extraction is clean and maintains byte-for-byte logic parity with the original mod.rs.

• All new modules correctly import their own dependencies, allowing mod.rs to safely drop unused imports (PerspectiveOrigin, TransformOrigin, ClipPath, etc.).

• Test coverage for intersect_rects and transform helpers was appropriately added during the move.

---

Generated by Openrouter AI (qwen/qwen3.6-plus). Supplementary to human reviews.

[github-actions]

Claude AI Incremental Architecture & Quality Review

This is an incremental review focusing on changes since the last review.

Issues (if any)

(none)

Previous Issues (for incremental reviews)

• [STILL UNRESOLVED] crates/oasis-browser/src/paint/mod.rs:38 - compute_transform_matrix is absent from the pub(crate) use transforms::{...} re-export block while all four sibling functions are included.
  • Confirmed not a compile error: the only call site is internal to transforms.rs:102 (compute_transform_offsets calls it within the same file). The mod.rs:480 reference is a doc comment, not a call.
  • Severity is lower than the original [BUG] report — this is an API surface inconsistency, not a broken build. The function is pub(crate) but is only reachable externally via super::transforms::compute_transform_matrix, not the flat super:: path that every other extracted helper uses.

Suggestions (if any)

• crates/oasis-browser/src/paint/transforms.rs:77 - Since compute_transform_matrix is only called from within transforms.rs itself, change its visibility to pub(super) to match its actual access pattern. This closes the inconsistency without requiring a re-export addition, and signals to future readers that it is a module-private helper rather than a crate-wide utility.

Notes

• The three-way split (clip_path.rs, stacking.rs, transforms.rs) is a clean mechanical extraction with no behavioral changes. The pub(crate) promotions on resolve_clip_path_rect and intersect_rects are correct since they now cross a module boundary via re-export.
• intersect_rects gained two tests that were missing from the original — good coverage improvement.
• The m.e as i32 / m.f as i32 truncating casts in compute_transform_offsets:102 are pre-existing behavior, not introduced here.

---

Generated by Claude AI (sonnet). Supplementary to human reviews.

[AndrewAltimit]

Review Response Agent (Iteration 1)

Status: Changes committed, pushing...

Commit: 8bb5808

Fixed Issues

• crates/oasis-browser/src/paint/transforms.rs:77 — Changed compute_transform_matrix visibility from pub(crate) to pub(super) to resolve the API surface inconsistency raised by all three reviewers. Verified the function is only called within transforms.rs itself (line 102); the mod.rs:480 reference is a doc comment, not a call site. pub(super) keeps it reachable from the parent paint module if ever needed while signaling it is not a crate-wide utility, which is the cleaner of the two reviewer-suggested fixes (the alternative being to add it to the re-export list).

Ignored Issues

• The original [BUG] claim that omission of compute_transform_matrix from re-exports would cause a compile error was a false positive — the third reviewer (Claude Quality, incremental) downgraded it correctly to a consistency suggestion, since no external caller exists. The fix above resolves the consistency concern at the visibility level instead of the re-export level.

Deferred to Human

• (none)

Notes

• Verified cargo build -p oasis-browser and cargo clippy -p oasis-browser -- -D warnings both pass after the change.
• Two reviewers also flagged [UNVERIFIED - File does not exist] warnings about record.rs. That file does exist at crates/oasis-browser/src/paint/record.rs (referenced from paint/mod.rs:26); the reviewers' filter incorrectly flagged the path. No action needed.

---

Automated summary of agent fixes.

[github-actions]

Claude AI Incremental Architecture & Quality Review

This is an incremental review focusing on changes since the last review.

Issues (if any)

(none)

Previous Issues (for incremental reviews)

• [RESOLVED] crates/oasis-browser/src/paint/mod.rs:31 - compute_transform_matrix missing from pub(crate) use transforms::{...} re-exports.
  • Fixed by changing visibility to pub(super) in transforms.rs:77, which is correct since the only call site is compute_transform_offsets within the same file. The re-export list in mod.rs is now consistent with actual access patterns.

Suggestions (if any)

• crates/oasis-browser/src/paint/transforms.rs:120 - The doc link [`compute_transform_matrix`] in compute_transform_offsets's rustdoc is valid today (same module), but if compute_transform_matrix is ever inlined or removed, rustdoc will emit a broken-link warning. No action required — just worth knowing the link is module-local.

Notes

• The transforms.rs extraction is clean: pub(super) on compute_transform_matrix correctly signals it is a module-private helper rather than a crate-wide utility, closing the API surface inconsistency from the previous review cycle.
• All nine tests in transforms.rs cover the distinct code paths (empty list, 3D variants, 2D-only, origin defaults, offset passthrough, translate shift). Coverage is sufficient.
• No behavioral changes in any of the three extracted modules — byte-for-byte logic parity with the original mod.rs is maintained.

---

Generated by Claude AI (sonnet). Supplementary to human reviews.

[github-actions]

Openrouter AI Incremental General Review

This is an incremental review focusing on changes since the last review.

Issues (if any)

(none)

Previous Issues (for incremental reviews)

• [RESOLVED] crates/oasis-browser/src/paint/mod.rs:31 - compute_transform_matrix missing from pub(crate) use transforms::{...} re-exports.
  • Fixed by changing visibility to pub(super) in transforms.rs:77, correctly scoping it to the module since it's only called internally by compute_transform_offsets.

Suggestions (if any)

(none)

Notes

• Mechanical extraction is clean: logic in clip_path.rs, stacking.rs, and transforms.rs is byte-for-byte identical to the original mod.rs implementation.
• Visibility adjustment for compute_transform_matrix (pub(super)) accurately reflects its internal-only usage within transforms.rs, resolving the API surface inconsistency flagged in prior reviews.
• Import cleanup in mod.rs correctly drops TransformFunction, ClipPath, ClipLength, PerspectiveOrigin, and TransformOrigin as they are now encapsulated in their respective sibling modules.
• Test coverage for intersect_rects and transform helpers was appropriately added/moved, covering edge cases like collapsed insets, disjoint rectangles, and 3D transform detection.
• intersect_rects and resolve_clip_path_rect correctly clamp negative dimensions to 0.0, preventing invalid layout rects from propagating.

---

Generated by Openrouter AI (qwen/qwen3.6-plus). Supplementary to human reviews.

[github-actions]

Claude AI Incremental Security & Correctness Review

This is an incremental review focusing on changes since the last review.

Issues (if any)

(none)

Previous Issues (for incremental reviews)

• [RESOLVED] crates/oasis-browser/src/paint/mod.rs:31 — compute_transform_matrix missing from pub(crate) use transforms::{...} re-exports.
  • Fixed in commit 8bb5808 by changing visibility to pub(super) in transforms.rs:77. The only call site (compute_transform_offsets, same file at line 102) is within the module, making pub(super) the correct scope. All subsequent reviewers confirmed resolution.

Suggestions (if any)

(none)

Notes

• All three extracted modules (clip_path.rs, stacking.rs, transforms.rs) are byte-for-byte logic parity with the original mod.rs. No behavioral change is possible from a mechanical extraction.
• compute_transform_matrix's pub(super) visibility is correct and consistent: it is called only within transforms.rs itself, and pub(super) keeps the parent paint module reachable if ever needed while correctly signaling it is not a crate-wide utility.
• The m.e as i32 / m.f as i32 truncating casts in compute_transform_offsets are pre-existing behavior (NaN → 0, saturating on overflow since Rust 1.45); no new risk introduced.
• resolve_clip_path_rect and intersect_rects correctly promoted to pub(crate) since they now cross a module boundary via re-export.

---

Generated by Claude AI (sonnet). Supplementary to human reviews.

[AndrewAltimit]

Review Response Agent (Iteration 2)

Status: No changes needed

Fixed Issues

• (none)

Ignored Issues

• The original compute_transform_matrix re-export inconsistency was already fixed in prior iteration (commit 8bb5808): visibility changed to pub(super) at crates/oasis-browser/src/paint/transforms.rs:77. All three incremental reviews on that commit explicitly mark it [RESOLVED].
• Claude Quality's note about a rustdoc link to compute_transform_matrix in compute_transform_offsets's doc — explicitly labeled "No action required" by the reviewer; the link is currently valid.

Deferred to Human

• (none)

Notes

• All three latest reviews (Claude Quality, Claude Security, OpenRouter General) on commit 8bb5808 report no issues and no actionable suggestions. The PR is in a clean review state.
• Verified transforms.rs:77 shows pub(super) fn compute_transform_matrix(...) as expected from the prior iteration's fix.

---

The agent reviewed feedback but determined no code changes were required.