fix(🐛): swap the skew shear axes in the native recorder too

[dennytosp]

Follow-up to #4015 — that PR is why main is currently red.

What went wrong

#4015 fixed skewX/skewY in processTransform3d, but the native recorder keeps its own copy of that conversion in cpp/api/recorder/Convertor.h, and that copy has the same two axes swapped. Only the TypeScript side was corrected, so the two backends now disagree.

The snapshots were regenerated for the corrected behaviour, so native is the one that fails. On main today:

job	result
build-test-ios-graphite (run)	Transforms.spec.tsx — both skew tests fail, 167673 px diff
test-android (on open PRs, which merge with main)	same two tests, 167047 / 167045 px diff

The fix

SkM44’s constructor reads its arguments in row-major order — the same layout as the Matrix4 tuple in TypeScript. So the two argument lists are now literally identical to the arrays in Matrix4.ts:

// skewX — shear factor in row 0, scaling y into x
SkM44 skewX(1, std::tan(angle), 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1);
// skewY — shear factor in row 0 of column 0, scaling x into y
SkM44 skewY(1, 0, 0, 0, std::tan(angle), 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1);

The perspective case a few lines below is a useful control: its C++ argument list already matches the TypeScript array element for element, and its snapshot passes on native — which is what confirms the constructor convention.

Testing

No new test: Transforms.spec.tsx already covers both axes and already fails on main, so test-android and build-test-ios-graphite going green here is the verification. I have no local Skia build, so I could not run the native path myself — CI is the check.

Sorry for the breakage.

[wcandillon]

thks a lot

[github-actions]

🎉 This PR is included in version 2.11.1 🎉

The release is available on:

• v2.11.1
• GitHub release

Your semantic-release bot 📦🚀