[skia-sync] Merge upstream chrome/m150

[mattleibow]

Automated upstream merge of chrome/m150.

Companion SkiaSharp PR: mono/SkiaSharp#4146

mono/skia PR Summary: Merge upstream Skia m148 → m150

Overview

Merged upstream chrome/m150 into skia-sync/m150, starting from the fork's skiasharp branch (HEAD 1a155bae3a). This advances the Skia engine from Chrome milestone 148 to 150.

Merge Details

Merge commit: 0a248d1b82 — "Merge upstream chrome/m150 into skia-sync/m150"
Upstream tip merged: c92514572e (tip of upstream/chrome/m150)
Conflicts resolved: 19 files
Branch tip: bceb12ad2c (merge + post-merge portability fixes — see "Post-Merge Portability Fixes" below)

Conflict Resolutions

File	Resolution	Reason
DEPS	Kept ours	Fork pins specific versions of expat (2.8.1), freetype (2.14.3), harfbuzz (14.2.0), libjpeg-turbo, libpng, libwebp, zlib — upstream would downgrade
RELEASE_NOTES.md	Took upstream	No fork-specific release notes
include/core/SkFont.h	Manual merge	Kept fork's breakText() overloads AND added upstream's new SkStrikeRef makeStrikeRef()
src/base/SkSafeMath.h	Took upstream	No fork changes
src/core/SkPixmap.cpp	Took upstream	No fork changes
src/core/SkRuntimeEffect.cpp	Took upstream	More defensive inline comment; no behavioral change
src/gpu/ganesh/SurfaceContext.cpp	Took upstream	No fork changes
src/gpu/ganesh/ops/AtlasTextOp.cpp	Manual merge	Combined fork's glyphSrcPadding()>0 check with upstream's atlasManager->supportsBilerp() condition; took upstream's graceful stride-mismatch handling (log+return instead of hard assert)
src/gpu/graphite/ (5 files)	Took upstream	Fork has no graphite changes (SkiaSharp uses Ganesh)
src/pathops/SkOpCoincidence.{cpp,h}	Took upstream	No fork changes
src/sksl/analysis/SkSLGetLoopUnrollInfo.cpp	Took upstream	No fork changes
tests/RegionTest.cpp	Manual merge	Kept fork's // TODO(kjlubick) comment + added upstream's new Region_setRects test (the TODO comment was later dropped to match upstream)
tests/SkGlyphTest.cpp	Took upstream	No fork changes
tests/SkRuntimeEffectTest.cpp	Took upstream	New SkRuntimeShader_b507643404 test

Breaking Changes in m148→m150 (Assessment)

Change	Impact on Fork	Action Taken
SkRegion::setRects signature change to SkSpan<const SkIRect>	None — C API doesn't call setRects	None needed
SkShader::isOpaque() pure virtual	None — fork doesn't subclass this	None needed
SkStrikeRef new class + SkFont::makeStrikeRef()	Minor — preserved in SkFont.h manual merge	Added to SkFont.h
SkLog.h gains #include <cstdarg>	Build environment issue (see below)	Fixed via build wrapper
Graphite API changes	Irrelevant — SkiaSharp uses Ganesh	None needed

Post-Merge Portability Fixes

The merge compiled on the primary toolchains, but SkiaSharp's full CI (SkiaSharp (Public)) builds the native libraries on older / niche toolchains (WASM emscripten 3.1.34 / libc++ 15, Tizen clang 10 + gcc 9.2, and the D3D backend on Windows) that upstream Skia CI does not. The following fork commits sit on top of the merge to make m150 build everywhere:

Commit	Area	Fix
047210c1bf	tests	Drop fork-only TODO(kjlubick) comment in RegionTest.cpp to match upstream
8912293c37	WASM / Tizen	std::views::reverse → reverse-iterator loop in SkPDFTag.cpp (libc++ 15 / clang 10 lack enabled ranges)
d0b422e9c9 (+ merge 391fb705ff, PR #255)	Tizen	Backport PDF backend's C++20 <=>/== to explicit C++17 operators (SkPDFTypes.h, SkPDFTag.cpp) — Tizen libstdc++ predates <compare>. Resolves #4155
d2fe11b410	Windows	Raise min_win_sdk_version floor to 10.0.22621.0 in gn/BUILDCONFIG.gn so the MSVC toolchain selects the Win11 SDK that defines IDWriteFontSet4 (m150 SkFontMgr_win_dw.cpp). Floor, not a pin
144ebf96d5	Tizen	optional::emplace on aggregate → aggregate-init (libstdc++ 9.2)
a8ea924948	Tizen	opts("ml3")/opts("ml4") use clang-10 -march=haswell / skylake-avx512 (m150's x86-64-v3/v4 need clang ≥ 12); keeps m150's new -mprefer-vector-width=512
159b1ec9e5	Windows D3D	GrD3DBackendSurface.cpp: GrBackendFormatData::equal unconditional, GrBackendTextureData::equal fully guarded under GPU_TEST_UTILS — completes upstream 20c304440c, which left the D3D backend abstract in non-test builds (only SkiaSharp compiles D3D)
bceb12ad2c	All platforms	Revert a SK_SUPPORT_PDF guard in src/c/sk_document.cpp — the C-API shim compiles into :core, which never receives :pdf's SK_SUPPORT_PDF public define, so the guard was always false → PDF factories returned null everywhere (caught by SKDocumentTest)

These are toolchain/portability fixes only — no change to Skia rendering behavior on any platform. Azure DevOps SkiaSharp (Public) build 158224 is green across all stages (Native Windows/macOS/Linux/WASM, Native, Package NuGets, API Diff, Build Managed, Tests, Samples).

Items Needing Human Attention

1. DEPS pinning: Fork continues to use custom-pinned native library versions. These were specifically upgraded as security/bug fixes. No action needed but worth tracking during releases.

2. AtlasTextOp.cpp merge: The upstream gracefully returns on stride mismatch (previously a hard assert). The combined supportsBilerp() || glyphSrcPadding() > 0 condition should be verified with a GPU test environment.

3. SkStrikeRef: New class added upstream. Not exposed in C API yet. Could be future API addition if users request access to glyph strike internals.

Created by skia-upstream-sync.