Drop "svg" from the names since flat/precision settings apply
equally to PDF output. The new names follow the hb_vector_draw_*
/ hb_vector_paint_* convention:
hb_vector_svg_set_flat -> hb_vector_draw_set_flat
hb_vector_svg_set_precision -> hb_vector_draw_set_precision
hb_vector_svg_paint_set_flat -> hb_vector_paint_set_flat
hb_vector_svg_paint_set_precision -> hb_vector_paint_set_precision
Assisted-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move the sweep gradient color stop tiling (pad/repeat/reflect
across 0-2π) into a shared template hb_sweep_gradient_tiles()
in hb-vector-svg-utils.hh. Both SVG and PDF backends now call
it with their own patch emitter callback.
Also adds hb_color_lerp() to the shared header (fixing an R/B
channel swap bug in the PDF version's interpolation).
Net reduction of ~170 lines of duplicated tiling code.
Assisted-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add PNG image support to the PDF paint backend. Parses PNG chunks
to extract IHDR metadata and IDAT compressed data, then embeds
directly as PDF XObject Image with /FlateDecode and /Predictor 15
(PNG predictor), avoiding any re-encoding of pixel data.
Supports:
- Grayscale, RGB, Gray+Alpha, RGBA (8-bit)
- Indexed color (palette PNG) via /Indexed colorspace with PLTE
- Transparency for indexed PNG via SMask: decompresses IDAT,
un-filters PNG rows, maps palette indices to tRNS alpha values,
and emits a separate DeviceGray soft mask image
Tested with NotoColorEmoji.ttf (CBDT, indexed PNG with partial
alpha) and sbix.ttf (SBIX, RGBA PNG).
Assisted-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Use the new push_group_for API to emit PDF ExtGState with /BM
(blend mode) when the compositing mode has a PDF 1.4 equivalent.
Maps 15 blend modes: Multiply, Screen, Overlay, Darken, Lighten,
ColorDodge, ColorBurn, HardLight, SoftLight, Difference, Exclusion,
Hue, Saturation, Color, Luminosity.
Porter-Duff modes (Clear, Src, Dst, etc.) have no direct PDF /BM
equivalent and fall back to the default (Normal/SrcOver).
Assisted-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add a new push_group variant that receives the compositing mode
upfront. This lets backends like PDF emit blend-mode state before
painting the group content, avoiding the need to buffer and
retroactively wrap.
The default (nil) implementation calls hb_paint_push_group(), so
existing backends are unaffected.
Callers updated:
- COLR PaintComposite now calls push_group_for() for both groups
- Rust/fontations push_layer passes the mode through
- hb-raster SVG context accepts mode parameter (default SRC_OVER)
All 233 tests pass.
Assisted-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The previous implementation only covered start_angle to end_angle
with a fixed number of equal sectors, which broke repeat/reflect
extend modes and produced too few patches for oscillating gradients.
Rewrite to use the same tiling logic as the SVG sweep gradient:
map color stops to angles, handle pad/repeat/reflect extend modes,
and tile across the full 0-2π range. Each color stop interval
becomes one Coons patch (further split at 90° max for arc quality).
Tested with sweep_270_440_reflect_narrow glyph from
test_glyphs-glyf_colr_1_variable.ttf.
Assisted-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The buffer helper functions (append_str, append_len, append_c,
append_num, append_unsigned, append_hex_byte, append_base64,
recover_recycled, blob_from) are format-agnostic and used by
both SVG and PDF backends. Rename from hb_svg_* to hb_buf_*
to reflect this.
SVG-specific helpers (append_color, append_instance_transform,
transform_point, scale_precision, etc.) keep their hb_svg_*
names.
Assisted-by: Claude <noreply@anthropic.com>
Replace the pie-slice sector fallback with proper Type 6
Coons patch mesh shading. Each sector of the sweep is a
patch with:
- Outer arc approximated by cubic Bézier (kappa method)
- Inner arc at small radius to avoid degenerate geometry
- Radial straight edges connecting inner and outer
- Corner colors from the color line at sector boundaries
Sectors are limited to 90° for accurate Bézier arc
approximation, with a minimum of 4 patches per sweep.
The mesh data is encoded as a binary stream with 16-bit
coordinates, 8-bit color components, and 8-bit flags per
the PDF Type 6 specification.
Assisted-by: Claude <noreply@anthropic.com>
Decompose sweep gradients into 360 colored pie-slice sectors.
Each sector evaluates the color line at its midpoint and emits
a filled triangular path from the center to two points on a
large-radius arc.
Includes a color line evaluator that handles pad, repeat, and
reflect extend modes with linear interpolation between stops.
Assisted-by: Claude <noreply@anthropic.com>
Rewrite hb-vector-pdf-paint.cc with a proper PDF object
system that supports dynamic resource generation:
Alpha:
Solid colors with alpha < 1 now emit ExtGState objects
with /ca (fill opacity) and reference them via /GSn gs
in the content stream.
Gradients:
Linear gradients → PDF Type 2 (axial) shading with
Type 2/3 interpolation functions for color stops.
Radial gradients → PDF Type 3 (radial) shading.
Sweep gradients → solid fallback (no PDF equivalent).
Architecture:
A hb_pdf_resources_t collector tracks extra PDF objects
(ExtGState, Shading, Function) during painting. At render
time, these are emitted as numbered objects and referenced
from the Page's Resources dictionary.
Mark internal symbols HB_INTERNAL to fix check-symbols test.
Assisted-by: Claude <noreply@anthropic.com>
Remove the -xy/-yy negation from the PDF paint glyph
positioning transform. Font and PDF coordinates are both
Y-up, so no flip is needed. Same fix as was applied to
the draw path earlier.
Assisted-by: Claude <noreply@anthropic.com>
Add hb-vector-pdf-paint.cc with a complete set of PDF paint
callbacks for the HarfBuzz paint API:
push/pop_transform → q + cm / Q
push/pop_clip_glyph → q + path + W n / Q
push/pop_clip_rect → q + re + W n / Q
color → rg + re + f
push/pop_group → q / Q
gradients → solid fallback (first stop color)
The PDF paint backend reuses the draw funcs for glyph clip
path emission and shares the hb_vector_paint_t struct with
the SVG backend via hb-vector-paint.hh.
hb_vector_paint_glyph dispatches to SVG or PDF paint funcs
based on format. hb_vector_paint_render delegates to the
PDF render function for PDF output.
Tested with COLRv0 (RocherColorGX) — color layers render
correctly with proper clipping.
Assisted-by: Claude <noreply@anthropic.com>
Move the paint struct and color glyph cache key to
hb-vector-paint.hh so the upcoming PDF paint backend can
share the type. Convert the SVG-specific transform helper
methods to free functions.
No functional change.
Assisted-by: Claude <noreply@anthropic.com>
The pen coordinates from vector-output.hh are in font space
(Y-up), not SVG space. The Y-flip in compute_extents only
affects the extents, not the pen positions passed to
hb_vector_draw_glyph. Use pen_y directly.
Assisted-by: Claude <noreply@anthropic.com>
Font and PDF coordinates are both Y-up, so no Y-flip is
needed in the path data. The pen position from the SVG-space
layout is the only thing that needs negating.
Remove the Y negation from append_xy and the CTM from the
content stream. Set the glyph transform to identity with
negated pen_y.
Assisted-by: Claude <noreply@anthropic.com>
The previous CTM-based Y-flip reversed path winding, causing
inverted fills. Instead, negate Y directly in append_xy for
PDF format. This preserves winding direction and produces
correct non-zero fill results.
Assisted-by: Claude <noreply@anthropic.com>
Prepend a 1 0 0 -1 0 (ey+eh) cm to the content stream so
that the SVG-space (Y-down) path coordinates render correctly
in PDF (Y-up). Set the MediaBox to [ex 0 ex+ew eh].
Assisted-by: Claude <noreply@anthropic.com>
Add HB_VECTOR_FORMAT_PDF to hb_vector_format_t. The hb-vector
tool now produces self-contained PDF when invoked with -o out.pdf.
PDF path operators map directly from the draw callbacks:
move_to → m, line_to → l, cubic_to → c, close_path → h,
with quadratic-to-cubic promotion (PDF has no Q operator).
The render function builds a minimal PDF 1.4 with catalog,
page, and content stream objects plus a cross-reference table.
Implementation:
- Extract hb_vector_draw_t struct to hb-vector-draw.hh so
future backends can share it.
- Draw callbacks branch on format for SVG vs PDF syntax.
- hb_vector_draw_glyph and hb_vector_draw_render dispatch by
format.
- vector-output.hh selects format from file extension and
skips paint (color glyphs) for PDF.
Coordinate system needs refinement — the page layout inherits
the SVG Y-flip transform, placing glyphs at the top with
extra whitespace. Correct PDF coordinate handling is a
follow-up.
Assisted-by: Claude <noreply@anthropic.com>
Add draw and paint features to font.rs. Enable both by default in
Cargo.toml. This allow compilation without draw and paint functionality
enabled on the C++ side.
Assisted-by: Gemini CLI
Malformed reflected gradients can produce very large finite
gradient parameters. The raster backend normalized reflect
mode by casting floorf(t) to int, which can overflow and
drive LUT lookups past the fixed-size stack buffer.
Fold reflected coordinates with fmodf(), clamp LUT inputs
defensively, and add the minimized oss-fuzz raster
testcase to the corpus.
Fixes: https://oss-fuzz.com/testcase-detail/6597066439852032
Tested: meson test -C build
Co-Authored-By: OpenAI Codex <codex@openai.com>
hb_gpu_draw_encode() could allocate a fresh output buffer and then
return NULL before transferring ownership to a blob, leaking that
temporary buffer on late encode failures.
The recycled-blob growth path also tried hb_realloc() first and fell
back to hb_malloc() when that failed. Once the new buffer was committed
to the recycled blob, the old recycled buffer could be orphaned.
Free temporary buffers on the remaining late-failure exits and keep
the old recycled buffer alive until a replacement has been committed.
Tighten the GPU API recycle test to encode a larger second outline so
that reused-buffer growth stays covered.
Tested:
- meson test -C build
Assisted-by: OpenAI Codex <codex@openai.com>
edge_sweep_row() computed the clipped endpoint x coordinates with
int32_t subtractions before casting to int64_t. If an edge started at
INT32_MIN and the scanline clipped endpoint was near zero, ey - yL
overflowed before the slope multiply and UBSan flagged the raster
fuzzer path.
Keep the endpoint interpolation and follow-on dx math in 64-bit until
clamping back to int32_t. This preserves the existing rasterizer logic
while avoiding signed overflow for extreme edge coordinates.
Tested:
- meson test -C build
Assisted-by: OpenAI Codex <codex@openai.com>
libpng aborts via longjmp. Our PNG read/write helpers kept scratch
hb_vector_t objects as automatic locals after setjmp and then tried
to free them from the longjmp handler. Once longjmp fires, those
modified locals have indeterminate state, so the cleanup path can
miss live allocations and leak under the raster fuzzer.
Move the PNG scratch buffers and write accumulator into explicit
heap-backed state allocated before setjmp and cleaned up on every
exit path. This keeps both decode and encode cleanup deterministic
across libpng errors.
Tested:
- env LD_LIBRARY_PATH=/home/behdad/harfbuzz/build/src valgrind --leak-check=full --show-leak-kinds=all --error-exitcode=1 ./build/test/fuzzing/hb-raster-fuzzer test/fuzzing/fonts/clusterfuzz-testcase-minimized-hb-subset-fuzzer-5684014636859392
- meson test -C build raster-fuzzer-chunk-0 raster-fuzzer-chunk-1 raster-fuzzer-chunk-2 raster-fuzzer-chunk-3 raster-fuzzer-chunk-4 raster-fuzzer-chunk-5 raster-fuzzer-chunk-6 raster-fuzzer-chunk-7 --wrap='valgrind --leak-check=full --show-leak-kinds=all --error-exitcode=1' --print-errorlogs
- meson test -C build
Assisted-by: OpenAI Codex <codex@openai.com>
Transformed clip rectangles kept their intersection math in unsigned
loop bounds after clamping only the lower edge to zero. When the
transformed upper edge stayed negative, the signed-to-unsigned cast
wrapped and raster clip reuse read past the previous alpha mask.
The same path also reused emitter-generated clip buffers without
clearing bytes outside the rows it rewrote, which let Valgrind see
uninitialized alpha values on later clip intersections.
Keep the clip intersection in signed ints until after the empty-range
check, and zero newly sized emitter clip buffers before partial writes.
Testing:
- env LD_LIBRARY_PATH=/home/behdad/harfbuzz/build/src valgrind --leak-check=full --error-exitcode=1 ./build/test/fuzzing/hb-raster-fuzzer test/fuzzing/fonts/clusterfuzz-testcase-minimized-hb-raster-fuzzer-4872362293264384
- meson test -C build raster-fuzzer-chunk-0 --wrap='valgrind --leak-check=full --error-exitcode=1' --print-errorlogs
- meson test -C build
Assisted-by: OpenAI Codex <codex@openai.com>
The raster paint clip code indexed non-rect clip alpha rows directly,
assuming the clip metadata always matched a full stride*height backing
buffer. Fuzzing showed that reused clip objects can reach those reads
with stale scalar state, which turns the row access in
push_clip_rectangle() into a heap-buffer-overflow.
Reset cached clip geometry on acquire, add a validity check for
non-rect alpha masks, and refuse to build new clip intersections from
an invalid mask. The fallback is an empty clip, which keeps the paint
path safe without affecting well-formed clips.
Testing:
- meson test -C build --suite fuzzing --print-errorlogs
- meson test -C build
Co-authored-by: OpenAI Codex <codex@openai.com>
The raster SVG renderer asserted that each element restored the paint
transform, clip, and surface stacks to their entry depths. Fuzzing
found malformed documents that violate that assumption and abort the
process in svg_render_element().
Replace the assertions with explicit stack restoration, discard any
extra group surfaces, and mark the render as failed if the renderer
cannot get back to the saved baseline. The top-level SVG entry point
now checks that failure bit and stops rendering cleanly instead of
crashing.
Testing:
- meson test -C build --suite fuzzing --print-errorlogs
- meson test -C build
Co-authored-by: OpenAI Codex <codex@openai.com>
MemorySanitizer reported use of uninitialized data while building the
SVG document cache used by vector painting. The parser keeps small
aggregate structs on the stack, and their padding could stay poisoned
as they were copied through the open-element stack.
Zero-initialize the parser stack and the temporary span / element
records before filling their fields. That keeps later copies and
container insertions from carrying undefined padding into the MSan
instrumented path.
Testing:
- meson test -C build --suite fuzzing --print-errorlogs
- meson test -C build
Co-authored-by: OpenAI Codex <codex@openai.com>
The GPU fuzzer can build outline data large enough to stress the size
math inside hb_gpu_draw_encode(). Several totals and byte counts were
computed with unchecked unsigned arithmetic before the existing bounds
checks ran.
Use HarfBuzz's overflow helpers for band-index totals, encoded blob
lengths, byte sizes, and recycled-buffer growth. Also rewrite the
uint16-offset limit check so it does not rely on subtracting from an
already-overflowed total.
Testing:
- meson test -C build --suite fuzzing --print-errorlogs
- meson test -C build
Fixes: https://oss-fuzz.com/testcase-detail/6336503637934080
Co-authored-by: OpenAI Codex <codex@openai.com>
Failing-malloc fuzzing exposed two holes in the GPU draw path.
Accumulation ignored push failures, and encode reused scratch vectors
that could remain poisoned after a previous allocation failure.
Poison the draw object on accumulator OOM, bail out of the remaining
path-building callbacks, and have hb_gpu_draw_encode() return NULL once
the object has failed. Also check the encode scratch resizes and reset
poisoned vectors before reuse. Resetting the draw now fully resets the
curve vector as well, so callers can recover after an OOM by calling
hb_gpu_draw_reset().
Testing:
- meson test -C build --suite fuzzing --print-errorlogs
- meson test -C build
Co-authored-by: OpenAI Codex <codex@openai.com>
Document that raster, vector, and gpu still must not crash,
access out of bounds, or corrupt memory under failing-malloc,
but best-effort early failure is acceptable and complex rollback
logic is not required after OOM.
Co-authored-by: OpenAI Codex <codex@openai.com>
Add an explicit commit-guidance rule to keep each commit focused on a
single issue or concern, instead of combining unrelated fixes,
renames, refactors, or behavior changes.
Co-authored-by: OpenAI Codex <codex@openai.com>
Encode used adjacent endpoint equality to infer contour breaks, which
fails when two contours touch at the same point. Track contour starts
explicitly during accumulation instead, and use that flag when packing
shared endpoints.
Encode also rounded blob bounds to the nearest quarter unit. That can
shrink the encoded extents and clip coverage at glyph edges. Quantize
mins downward and maxes upward so the blob header always encloses the
actual geometry.
Add focused API tests for both regressions.
Testing:
- meson test -C build
Co-authored-by: OpenAI Codex <codex@openai.com>
Check navigator.gpu synchronously in main() before calling
the WebGPU C API. Previously the error was reported in an
async callback that raced with onRuntimeInitialized, which
hid the loading div before the error message could appear.
Also re-show the loading div and set window._webgpuError in
the adapter callback as a fallback.
Assisted-by: Claude <noreply@anthropic.com>
Docs are built in the dedicated docs CI workflow. Remove
gtk-doc-tools and disable docs in the Linux build.
Assisted-by: Claude <noreply@anthropic.com>
Compute pixelsPerEm once in hb_gpu_render and pass it down,
so MSAA samples share one fwidth call instead of each
computing their own. Required for WGSL (fwidth must be
called from uniform control flow) and a minor win for all
backends.
Assisted-by: Claude <noreply@anthropic.com>
Add hb_gpu_draw_set_scale() API and store x/y scale in the
glyph blob header (buf[1].b/a). hb_gpu_draw_glyph() sets it
automatically from the font.
hb_gpu_render() now reads the scale, computes ppem via fwidth,
and automatically blends 4 corner samples at ±1/3 pixel when
ppem < 16 to reduce moiré at small sizes. Define
HB_GPU_NO_MSAA to disable.
Add hb_gpu_ppem() public shader helper so clients can get
ppem without touching internal functions (e.g. for
hb_gpu_darken).
Remove hb_gpu_render_aaa(), u_upem uniform, and set_upem
plumbing from the demo — no longer needed.
GLSL only for now.
Assisted-by: Claude <noreply@anthropic.com>
Replace the two-program MSAA toggle with hb_gpu_render_aaa()
that takes ppem from the caller and auto-selects single-sample
(ppem >= 16) or 4-corner MSAA (ppem < 16). This keeps
hb_gpu_render() lean for the fast path while letting the
adaptive function handle small sizes.
The previous approach had the ppem calculation inside the
library shader using 1/fwidth(texcoord), which was wrong by
a factor of upem — MSAA fired at all sizes. Now the demo
passes correct ppem via a u_upem uniform.
Remove the 'm' key toggle, two-program compilation, and
set_msaa plumbing — no longer needed.
GLSL only for now.
Assisted-by: Claude <noreply@anthropic.com>