Measurement source: 7d6748ed25dfdd1d0359ea72e1ca09fe19d211d5, version 0.8.0. Source counts, native sizes, configured dependencies, and coverage below use this source. Qualified PR evidence is identified separately where reused.
Generated with tools/cloc.sh (cloc 2.02), from the clean measurement source. LOC covers C++ and C/C++ Header files under donner/; JavaScript, Python, WGSL, and Objective-C++ are outside this count. The comment/code ratio is calculated from raw cloc CSV: 89,256 / 490,185 = 18.21%, avoiding the shell summary’s intermediate truncation.
Generated with python3 tools/feature_loc.py from the same source. The cloc product bucket excludes _tests.cc and tests/ directories; feature product buckets additionally exclude fuzzers and benchmarks. Their totals use different classifications and are not interchangeable partitions.
File-level attribution by first matching feature bucket. Product LOC excludes tests/, fuzzer, and benchmark files; support LOC includes them. Shared files stay in the broad bucket that owns the file.
| Feature | Product LOC | Support LOC | Total LOC | Files | Notes |
|---|---|---|---|---|---|
| Editor | 67,625 | 83,404 | 151,029 | 357 | ImGui/GLFW editor, async rendering, repro, and headless tooling |
| Composited rendering | 6,227 | 9,870 | 16,097 | 33 | Layer promotion, dual-path verification, and compositor hints |
| Text | 11,252 | 14,583 | 25,835 | 87 | SVG text DOM, layout, shaping, font loading, and font parsers |
| Filters | 7,297 | 3,857 | 11,154 | 74 | SVG filter DOM, filter graph components, and CPU/GPU filter engines |
| Rendering | 32,492 | 43,273 | 75,765 | 187 | Renderer facade, renderer driver, backends, and image comparison helpers |
| SVG document model | 34,510 | 34,841 | 69,351 | 444 | SVG DOM wrappers, parser, properties, ECS components, resources, and tools |
| CSS | 6,037 | 4,977 | 11,014 | 75 | CSS values, parsing, selector matching, and declarations |
| Base/shared | 17,711 | 23,759 | 41,470 | 169 | Core utilities, geometry, XML, parsers, encoding, and element helpers |
| Benchmarks | 0 | 1,703 | 1,703 | 11 | Benchmark harnesses under donner/benchmarks |
| Other | 43,932 | 42,835 | 86,767 | 545 | Files not covered by a more specific feature bucket |
| Total | 227,083 | 263,102 | 490,185 | 1,982 |
| Feature | Product LOC | Support LOC | Total LOC | Files | Notes |
|---|---|---|---|---|---|
| Editor app and shell | 16,400 | 3,499 | 19,899 | 24 | Main entrypoints, top-level app wiring, menus, sidebars, dialogs, and command queue |
| Editor host UI | 4,534 | 3,537 | 8,071 | 19 | GLFW/ImGui host integration, clipboard, profiling wrapper, and window glue |
| Document sync and undo | 2,503 | 4,727 | 7,230 | 15 | SVG document state, source sync, attribute writeback, incremental reparse, and undo |
| Viewport and tools | 4,935 | 8,378 | 13,313 | 34 | Viewport math, pointer gestures, selection, overlays, and interactive tools |
| Async rendering | 6,065 | 7,217 | 13,282 | 15 | Editor render coordination, async worker path, texture cache, and render perf repros |
| Text editor | 10,326 | 10,747 | 21,073 | 23 | Standalone source text editor, text buffer, text patching, and related tests |
| Repro files | 3,420 | 1,992 | 5,412 | 10 | User-visible repro recording and .donner-repro encoding/decoding |
| Editor misc and integration tests | 19,442 | 43,307 | 62,749 | 217 | Editor files not covered by a more specific editor sub-bucket |
| Total | 67,625 | 83,404 | 151,029 | 357 |
| Feature | Product LOC | Support LOC | Total LOC | Files | Notes |
|---|---|---|---|---|---|
| Renderer API and contracts | 1,411 | 6,727 | 8,138 | 9 | Public renderer facade, backend selection glue, interface contracts, and stroke params |
| Traversal and render context | 5,890 | 2,679 | 8,569 | 12 | RendererDriver traversal, RenderingContext instantiation, utility code, and shared views |
| TinySkia backend | 3,267 | 166 | 3,433 | 4 | TinySkia software renderer backend |
| Geode backend | 16,430 | 17,158 | 33,588 | 76 | Geode GPU backend (Metal/Vulkan/WebGPU), GPU device, encoders, pipelines, shaders, and tests |
| Image and debug utilities | 634 | 1,670 | 2,304 | 11 | Image IO, terminal image preview, and bitmap comparison helpers |
| Wasm bridge | 259 | 0 | 259 | 3 | Renderer wasm bridge entrypoints |
| Renderer test harness | 0 | 14,873 | 14,873 | 51 | resvg suite harness, test backend abstractions, and renderer error-path tests |
| Rendering misc | 4,601 | 0 | 4,601 | 21 | Rendering files not covered by a more specific renderer sub-bucket |
| Total | 32,492 | 43,273 | 75,765 | 187 |
Current-source Linux aarch64 measurements use Clang 21.1.6 (LLVM commit a832a5222e489298337fbb5876f8dcaf072c5cca) and the existing linux-binary-size profile: opt, -Os, debug information for attribution, function/data sections and linker section collection. The stripped files exclude debug information.
| Target | Non-LTO stripped file bytes | ThinLTO stripped file bytes |
|---|---|---|
| svg_parser_tool | 5,007,264 | 4,895,480 |
| donner-svg | 7,802,528 | 7,416,552 |
These measure the parser and CLI, not the editor application. Parser Bloaty attribution totals 5,007,264 file bytes and 2,612,476 VM bytes; the chart and treemap display file bytes. Chart/treemap labels use 1024-based units (KiB/MiB), even where their generated labels say KB/MB. Exact byte counts in the table are authoritative.
The chart combines omitted groups with the existing “Everything else” bucket; its displayed groups sum to all file bytes. The treemap preserves bytes at intermediate symbol prefixes as explicit (self) children. Every parent equals its children, and all leaves sum to the same file-byte total. The generated assets include lossless data corrections and container CSS layout normalization; the repository generators and renderer JavaScript are unchanged. The generated container has a viewport-relative height, a 320px minimum height, and relative positioning for initial layout.
Interactive parser file-byte treemap. HTML validation covers the static tree structure, exact byte totals, and unchanged renderer JavaScript; only generated container CSS and data changed.
ThinLTO reduces stripped file bytes by 2.2% for the parser and 4.9% for the CLI in this matched pair. This is an existing build profile, not a change to product build flags.
Reproduce the native target/profile configuration with:
Attribution used the stripped parser plus its unstripped debug file and the existing tools/binary_size_config.bloaty grouping. Debug-file bytes are not included in the stripped-size table.
The qualified Editor WASM run 39800234119, attempt 1 built synthetic PR commit 195ebf77133902b68bdc8aac658977ff7430f641. Its Git tree 0420666d0d310cfd1510f8e9ba145be6451d18e5 is identical to the measurement source's tree. The embedded identity and provenance remain the synthetic commit; this is not represented as a new build of the merged commit, and a different embedded commit can change compressed bytes.
| Artifact | Bytes | SHA-256 |
|---|---|---|
| Raw editor Wasm | 10,637,562 | fd4f86f3a085b67c5338859e40719b3049c15d8cfa922b3a517b2df504378ef4 |
| Served gzip Wasm | 3,475,312 | f2d064d516b5ce907abd36da75b3583d84b6e5ddcda41e497b2e08b0a9c8abee |
Artifact 12437537934 was checked against all 44 site/deployment checksums. Bounded decompression matched the raw Wasm exactly, with valid EOF and no trailing or extra gzip members. The exact 0.8.0 plus synthetic-commit marker occurred once.
Coverage run 39813628374, attempt 1 measured the exact main commit above with //donner/... (proof schema 2, complete-main). Build job 119946238138 and upload job 119998109117 succeeded. The workflow used Ubuntu 24.04 x86_64 and tools/ci/coverage_lane.sh → tools/coverage.sh, with the instrumented latest_llvm configuration. This coverage venue is separate from the Linux aarch64 native-size measurements.
| Coverage metric | Numerator / denominator | Percent |
|---|---|---|
| Fully covered lines; partials count as misses | 125,616 / 147,340 | 85.2559% |
| Raw LCOV lines with a positive hit | 134,835 / 147,340 | 91.5128% |
| Branches hit | 59,241 / 72,334 | 81.8992% |
The line denominator includes 9,219 partial and 12,505 missed lines. The fully-covered-line metric exceeds 85%; genhtml's 91.5% line-hit headline is a different metric.
The retained LCOV contains 1,187 source records and has SHA-256 6a9a2c08ee3576e931f4724549a25ba42237248c2fa3dbdc61a0991e3d967daf, matching the workflow proof. Proof artifact 12447173035 and filtered-report artifact 12447642188 are bound to this run and source.
For source checkout and PR review, download the coverage HTML ZIP; it contains coverage/index.html. The archive is 10,272,662 bytes, SHA-256 b88ac6cfaf74c22122d537108605136dd15a13797f7f6f8d7696b47cbd6f5563. It contains 2,574 regular HTML/style/image members, with no raw logs, command metadata, BEP, or LCOV. HTML was generated from the retained LCOV with genhtml 2.0-1, without rerunning tests. All source paths were already repository-relative.
The published coverage view reflects this snapshot only after documentation promotion; until then it may show an older report.
Qualified CI run 39800234110, attempt 1 used the same synthetic commit and identical Git tree described above. These are affected-target subsets, not a fresh full-repository test suite. Cached passes are not fresh execution; selected target counts are not inner test-case counts.
| Platform / invocation | Selected targets | Passed, including cache | Executed targets | Skipped targets |
|---|---|---|---|---|
| Linux main group | 115 | 107 | 8 | 8 |
| macOS main group | 111 | 108 | 7 | 3 |
Separate macOS invocations executed 1/1, 6/6, 1/1 and 1/1 selected targets. Browser 4/4 and Geode 6/6 groups passed from cache. These overlapping groups are not summed as unique tests. This is qualified PR evidence, not fresh execution at the merged commit or the later report-only candidate.
Separately, the exact-main instrumented Coverage invocation selected the complete product pattern //donner/.... Its proof records 575 PASSED label/configuration status records and 165 skipped labels. The proof does not encode a fresh/cache partition or inner test-case counts; those counts are not inferred or added to the overlapping PR groups.
Current-source query: kind(library, set(//donner/... //:*)) intersect attr(visibility, public, //...).
These are build targets with public visibility, not a recommendation to use internal implementation APIs.
All three dependency queries use the common latest_llvm configuration on Linux aarch64. “Default” means no additional text configuration, not the host-default compiler. Configured transitive build queries include toolchain/sysroot and helper-tool inputs; this is not a runtime-only bill of materials. Vendored components can appear in the generated NOTICE without being separate external repositories. Repository counts and license-package counts therefore differ.
Query: deps(//examples:svg_to_png); configuration: default. Attribution: //third_party/licenses:notice_default.
External build repositories: entt, stb, sysroot_linux_aarch64, tiny-skia-cpp, toolchains_llvm, zlib.
| NOTICE package | License identifiers | Upstream |
|---|---|---|
| EnTT | MIT | upstream |
| Public Sans | OFL-1.1 | upstream |
| stb | MIT | upstream |
| tiny-skia-cpp | BSD-3-Clause | upstream |
| libpng | libpng-2.0 | Not supplied by the manifest |
| zlib | Zlib | Not supplied by the manifest |
Query: deps(//examples:svg_to_png); configuration: text-full. Attribution: //third_party/licenses:notice_text_full.
External build repositories: brotli, entt, freetype, harfbuzz, libpng, stb, sysroot_linux_aarch64, tiny-skia-cpp, toolchains_llvm, woff2, zlib.
| NOTICE package | License identifiers | Upstream |
|---|---|---|
| EnTT | MIT | upstream |
| Public Sans | OFL-1.1 | upstream |
| stb | MIT | upstream |
| tiny-skia-cpp | BSD-3-Clause | upstream |
| libpng | libpng-2.0 | Not supplied by the manifest |
| zlib | Zlib | Not supplied by the manifest |
| brotli | MIT | upstream |
| FreeType | FTL | upstream |
| HarfBuzz | MIT | upstream |
| woff2 | MIT | upstream |
Query: deps(//donner/editor:editor); configuration: target transition. Attribution: //third_party/licenses:notice_release_editor.
External build repositories: brotli, entt, freetype, gfont_bebas_neue, gfont_bitter, gfont_inter, gfont_jetbrains_mono, gfont_lato, gfont_lora, gfont_montserrat, gfont_open_sans, gfont_oswald, gfont_pacifico, gfont_playfair_display, gfont_roboto_mono, glfw, harfbuzz, imgui, libpng, python_3_12_aarch64-unknown-linux-gnu, stb, sysroot_linux_aarch64, tiny-skia-cpp, toolchains_llvm, vulkan_headers, woff2, zlib.
| NOTICE package | License identifiers | Upstream |
|---|---|---|
| EnTT | MIT | upstream |
| stb | MIT | upstream |
| tiny-skia-cpp | BSD-3-Clause | upstream |
| brotli | MIT | upstream |
| FreeType | FTL | upstream |
| HarfBuzz | MIT | upstream |
| woff2 | MIT | upstream |
| Dear ImGui | MIT | upstream |
| GLFW | Zlib | upstream |
| Roboto | Apache-2.0 | upstream |
| Fira Code | OFL-1.1 | upstream |
| Public Sans | OFL-1.1 | upstream |
| Bootstrap Icons | MIT | upstream |
| Curated Google Fonts catalog | OFL-1.1 | upstream |
| glad | MIT, Apache-2.0 | upstream |
| Vulkan-Headers | Apache-2.0, MIT | upstream |
| libpng | libpng-2.0 | Not supplied by the manifest |
| zlib | Zlib | Not supplied by the manifest |
| Noto Sans | OFL-1.1 | upstream |
The Linux editor NOTICE sidecar reports zero uncovered component labels, but 11 unexempted linker-option entries: -lGL, -lX11, -lXcursor, -lXi, -lXinerama, -lXrandr, -ldl, -lpthread, -lrt. These are recorded limitations, not new exemptions or complete license-closure clearance for a Linux editor distribution.
API documentation. This link is informational; it does not claim a fresh documentation build at the measurement source.
1.15.0 