Commit Graph
17349 Commits
Author SHA1 Message Date
Behdad Esfahbod 5b1aa7c9db [repacker] Bound clone buffers by each ligature split range
Many short ranges allocated the original LigatureSet offset array for
every clone, retaining quadratic buffer space even after shrink_buffer
moved the tails. Bound capacity by the number of offset slots in the
range, since each emitted set consumes at least one slot.

Advance through earlier ligature counts once across ordered ranges and
stop scanning at each range's end. Add a compact graph requiring one
subtable per coverage entry and verify counts, glyphs and connectivity.

An 8000-range run reduces peak RSS from 163 MB to 32 MB. Source-link
movement and coverage filtering still take quadratic time across many
ranges; document that remaining work rather than claiming a full fix.

Testing: full suite (303 passed, 1 skipped), ASan/UBSan repacker tests,
HB_NO_BEYOND_64K repacker tests, and 520 injected allocation failures
in a temporary harness. The combined fix series also passed 228650
ASan fuzzer runs with failing malloc.

Assisted-by: OpenAI Codex
2026-11-12 18:24:28 -06:00
Behdad Esfahbod 95b711ccb8 [repacker] Cache LigatureSet ids throughout subtable splitting
Refresh the id map once after unsharing and keep it in the split
context. Reuse it in every clone_range and in shrink, where repeated
index_for_offset scans previously made a large retained run quadratic.
Pass ligature counts by reference instead of copying them per range.

Exercise the 4000-set shared run both before and after a large ligature
to cover retained and moved runs, checking distinct ids in each case.

Testing: meson test -C build (303 passed, 1 skipped).

Assisted-by: OpenAI Codex
2026-11-12 18:24:12 -06:00
Behdad EsfahbodandClaude Opus 4.8 b17af8a343 [repacker] Un-share LigatureSets before splitting, in bounded time and memory (#6659)
* [repacker] Un-share LigatureSets before splitting, in bounded time and memory

When the subset serializer deduplicates a LigatureSet that is reachable
from several coverage entries of one LigatureSubst, pop_pack() links all
of those entries to a single LigatureSet vertex. The LigatureSubst
splitter numbers ligatures per coverage entry and then edits the
LigatureSet vertices in place (move_children, compact_liga_set, shrink),
which assumes one vertex per entry, so the shared entries silently lose
their ligatures. #6088 removed the duplicate_shared_liga_sets pre-pass
that used to guard this, on the assumption that as_mutable_table()'s
copy-before-mutate made it redundant; but that only un-shares a child
shared with *other* parents, not one reached by several links from the
same parent, so the bug came back.

Restore the pre-pass, but make it linear in both time and memory, since
a naive restore is O(n^2) on a subtable whose n coverage entries all
share one set:

- Memory: LigatureSet::subset()/Ligature::subset() each add one virtual
  link to Coverage ("ensure Coverage is packed after this") per subset
  call, and pop_pack()'s merge_virtual_links appends them to the deduped
  vertex without collapsing, so the shared set accumulates n redundant
  virtual links. duplicate(copy_table) copies that whole vector into
  every per-entry clone, i.e. O(n^2). unshare_duplicate_children()
  collapses each set's duplicate virtual links to one (dedup_virtual_links,
  fixing the dropped links' incoming-edge counts) before cloning, so each
  clone copies O(1), and does the whole un-share in a single pass over the
  parent's links instead of an O(n) remap-and-rescan per entry.

- Time: the split's analysis passes (compute_split_points,
  total_number_ligas, liga_counts) and clone_range() each did an O(n)
  index_for_offset() per entry. Precompute the ligatureSet index -> object
  id map once per split and pass it to the analysis passes, and in
  clone_range() move each run of fully-contained sets with a single
  move_children() (and look moved sets up through the new, small subtable)
  instead of a move_child()/as_mutable_table() per entry.

The un-share runs after the analysis passes: the deep copies share the
same Ligatures and keep the same ligature.len, so the counts and split
points are unchanged.

Testing: the K86 correctness reproducer now loses 0 of 8400 ligatures
(was lossy). The 1P5QWD69 reproducer (32759 entries sharing one set, 4
split subtables) drops from ~24 s / an OOM at ~12 GB to ~0.65 s CPU and
~137 MB peak RSS; instruction count vs entry count is linear (doubling n
multiplies work by ~1.8x, not ~4x), and peak RSS is O(n). Adds
test_resolve_with_liga_split_shared_liga_set to test-repacker. The full
meson test suite (including the subset and subset+repack suites) passes.

Discovered by Anthropic's OSS Scanner, as vulnerabilities
ANT-2026-K86LK9JR (dropped ligatures) and ANT-2026-1P5QWD69 (quadratic
DoS).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [repacker] Cache LigatureSet ids while cloning split ranges

Moving a run of LigatureSets still looked each set up by scanning the
new subtable's links. This left clone_range quadratic in the run length.

Resolve the original offsets once before moving sets, and reuse their
object ids. Extract the existing copy-before-mutate check so cached ids
retain the same handling of children shared with other parents.

Add a 4000-set moved-run test checking alternating shared ligatures.
Instrumented lookup comparisons for runs of 1000/2000/4000 sets drop
from 501505/2003005/8006005 to 5/5/5.

Testing: full build and meson test -C build (281/281 passed).

Assisted-by: OpenAI Codex

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-12 14:18:34 -06:00
Behdad EsfahbodandMattias Eriksson e563058688 [font] Add hb_font_create_without_funcs_or_fail() (#6651)
Let clients install their own font funcs without initializing and
immediately discarding the default backend. Return NULL if the font
object or its initial variation coordinates cannot be allocated.
Keep hb_font_create()'s initialization order and behavior.

Use the new constructor in hb-ft, Fontconfig, explicit tool backend
selection, and the font benchmarks. This avoids loading a second
FT_Face when FreeType is the default while preserving the caller's
face. Document the custom-client use case and add API and stream-read
regressions.

Testing: full Meson suite (302 passed, one skipped), generated docs,
reduced-feature builds, allocator failures, and tool/benchmark checks.

Supersedes: https://github.com/harfbuzz/harfbuzz/pull/6648

Assisted-by: OpenAI Codex

Co-authored-by: Mattias Eriksson <me@axentia.se>
2026-11-12 13:30:19 -06:00
Behdad Esfahbod 438c3a8b11 [repacker] Cache LigatureSet ids while cloning split ranges
Moving a run of LigatureSets still looked each set up by scanning the
new subtable's links. This left clone_range quadratic in the run length.

Resolve the original offsets once before moving sets, and reuse their
object ids. Extract the existing copy-before-mutate check so cached ids
retain the same handling of children shared with other parents.

Add a 4000-set moved-run test checking alternating shared ligatures.
Instrumented lookup comparisons for runs of 1000/2000/4000 sets drop
from 501505/2003005/8006005 to 5/5/5.

Testing: full build and meson test -C build (281/281 passed).

Assisted-by: OpenAI Codex
2026-11-11 22:58:56 -06:00
Behdad Esfahbod 136b04439b Merge pull request #6656 from harfbuzz/fix/raster-bgra-endianness-6655
[raster] Preserve BGRA byte order on big-endian systems
2026-11-11 21:30:00 -06:00
Behdad EsfahbodandClaude Opus 4.8 73a5bc580a [repacker] Un-share LigatureSets before splitting, in bounded time and memory
When the subset serializer deduplicates a LigatureSet that is reachable
from several coverage entries of one LigatureSubst, pop_pack() links all
of those entries to a single LigatureSet vertex. The LigatureSubst
splitter numbers ligatures per coverage entry and then edits the
LigatureSet vertices in place (move_children, compact_liga_set, shrink),
which assumes one vertex per entry, so the shared entries silently lose
their ligatures. #6088 removed the duplicate_shared_liga_sets pre-pass
that used to guard this, on the assumption that as_mutable_table()'s
copy-before-mutate made it redundant; but that only un-shares a child
shared with *other* parents, not one reached by several links from the
same parent, so the bug came back.

Restore the pre-pass, but make it linear in both time and memory, since
a naive restore is O(n^2) on a subtable whose n coverage entries all
share one set:

- Memory: LigatureSet::subset()/Ligature::subset() each add one virtual
  link to Coverage ("ensure Coverage is packed after this") per subset
  call, and pop_pack()'s merge_virtual_links appends them to the deduped
  vertex without collapsing, so the shared set accumulates n redundant
  virtual links. duplicate(copy_table) copies that whole vector into
  every per-entry clone, i.e. O(n^2). unshare_duplicate_children()
  collapses each set's duplicate virtual links to one (dedup_virtual_links,
  fixing the dropped links' incoming-edge counts) before cloning, so each
  clone copies O(1), and does the whole un-share in a single pass over the
  parent's links instead of an O(n) remap-and-rescan per entry.

- Time: the split's analysis passes (compute_split_points,
  total_number_ligas, liga_counts) and clone_range() each did an O(n)
  index_for_offset() per entry. Precompute the ligatureSet index -> object
  id map once per split and pass it to the analysis passes, and in
  clone_range() move each run of fully-contained sets with a single
  move_children() (and look moved sets up through the new, small subtable)
  instead of a move_child()/as_mutable_table() per entry.

The un-share runs after the analysis passes: the deep copies share the
same Ligatures and keep the same ligature.len, so the counts and split
points are unchanged.

Testing: the K86 correctness reproducer now loses 0 of 8400 ligatures
(was lossy). The 1P5QWD69 reproducer (32759 entries sharing one set, 4
split subtables) drops from ~24 s / an OOM at ~12 GB to ~0.65 s CPU and
~137 MB peak RSS; instruction count vs entry count is linear (doubling n
multiplies work by ~1.8x, not ~4x), and peak RSS is O(n). Adds
test_resolve_with_liga_split_shared_liga_set to test-repacker. The full
meson test suite (including the subset and subset+repack suites) passes.

Discovered by Anthropic's OSS Scanner, as vulnerabilities
ANT-2026-K86LK9JR (dropped ligatures) and ANT-2026-1P5QWD69 (quadratic
DoS).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 21:24:29 -06:00
Behdad EsfahbodandClaude Opus 4.8 f87e9d774c [gvar] Charge tuple region eval and point-number decoding to work budget (#6654)
apply_deltas_to_points() only charged the glyph work budget
(HB_BUDGET_GLYPH) for tuple variations that actually applied, and
only by the glyph's point count, after calculate_scalar() had
already walked gvar.axisCount axes for every tuple header.  A
~35 KB font with axisCount 8192 and 4095 non-applying tuples on a
composite-multiplied glyph could hold one hb_font_get_glyph_extents()
call for over a minute while using ~3% of the budget.  Two further
pieces of work were also uncharged: the shared point-number list is
decoded on every glyph load, and an applied tuple decodes/applies
indices.length deltas, which can exceed the point count charged.

Following the project's "charge before doing the work" idiom
(#6256, #6410), charge 1 + axisCount per tuple before its region is
evaluated (bounding calculate_scalar() regardless of the scalar
cache or outcome), shared_indices.length once per load after the
shared point numbers are decoded, and indices.length per applied
tuple before its deltas are decoded.  A font whose tuple work
exceeds the budget now fails the glyph load (outline/extents return
false) instead of finishing slowly; real fonts (a dozen axes, tens
of tuples) pay a few hundred units per load.

Adds test_hb_draw_gvar_budget to test/api/test-draw.c covering both
the unapplied-tuple region charge and the point-number charges.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-S2OJWJHX.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 20:38:47 -06:00
Behdad EsfahbodandClaude Opus 4.8 388d91d96e [instancer] Dedup VarStore delta rows by sorting, not by a row hash (#6653)
item_variations_t::as_item_varstore() and compile_varidx_map() removed
duplicate ItemVariationStore delta rows and mapped them to new
VariationIndex values with an hb_hashmap_t keyed by pointer-to-row,
which hashes and compares the row contents. hb_array_t::hash() is an
invertible FNV-1a over whole 32-bit words with a fixed seed, so a font
author can solve the last delta of each row to force thousands of
distinct rows to share one 30-bit hash. Keys with equal hashes probe
the same slots at every table size, so every insert and lookup walks
all earlier rows and the dedup goes quadratic: a 320 KB font cost about
30 s of CPU in hb-subset --variations where an ordinary font of the
same size took 0.02 s.

Replace both content-keyed maps with a sort-based dedup that does not
depend on the row hash:

- as_item_varstore() collects the candidate rows in major/minor order,
  then sorts an index of them by row content (ties broken by position)
  and flags each row equal to its predecessor, keeping the first
  occurrence. Encodings are then rebuilt per major in the original
  order.
- compile_varidx_map() collects (row, new VarIdx) pairs while splitting
  encodings, sorts them by content, and matches each front-mapping row
  by binary search instead of probing the map.

Both are O(n log n) for any font and keep the output deterministic.
Rows reaching this code are already globally de-duplicated by content,
so the matches are unique. Output is byte-identical: the subset suite's
partial-instance tests pass unchanged, and instancing RobotoFlex,
MPLUS1 and AdobeVFPrototype (CFF2) is byte-identical to before. The
colliding reproducer now instances in well under a second.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-PT58IWSN.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 20:38:12 -06:00
Behdad EsfahbodandClaude Opus 4.8 0264035291 [subset] Seed the serializer object-dedup hash per serialization (#6657)
pop_pack() deduplicates packed objects through packed_map, an
open-addressing hashmap keyed by object content.  object_t::hash()
used fasthash32 with a fixed seed, which is invertible, so a font
could craft thousands of distinct objects -- e.g. 16-byte GPOS
PairSets -- that all share one 30-bit hash.  Every insert and lookup
then walks the whole collision chain and the dedup goes quadratic:
a 337 KB crafted font took 56 s of CPU in hb-subset where an
ordinary font the same size took 0.2 s.

Seed the hash with a value computed once per hb_serialize_context_t
from two ASLR-randomized addresses (the context's own and a stack
frame).  A font being subset cannot observe them, so it cannot
precompute colliding objects.  With ASLR disabled the seed folds to
a constant, i.e. no worse than the previous fixed seed.

The seed never changes output: packed_map is lookup-only and dedup
keeps the first-inserted object, independent of hash order.  The
per-object hash is cached on object_t so the map's del(), which
hashes the key itself, sees the same seeded value without touching
hb-map.hh.

Testing: full meson test passes (281/281).  Subset output is
byte-identical to before and identical across two different seeds,
verified on NotoNastaliqUrdu, Amiri, AwamiNastaliq, NotoSansDevanagari,
Roboto, NotoSansCJKsc, PingFang and NotoColorEmoji.  Instructions
retired are unchanged (<0.01%) on a same-build-dir A/B.  The crafted
reproducer drops from 56 s to 0.13 s.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-3T4G5NLB.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 20:37:06 -06:00
aa7a824cc4 [JSTF] Expose extender glyphs via hb_ot_layout_get_extender_glyphs (#6568)
* [JSTF] Expose extender glyphs via hb_ot_layout_get_extender_glyphs

The JSTF table is already parsed but never exposed to clients. Extender
glyphs are the most widely populated part of it and are useful for
cursive-script justification implementations that want to prefer a
font-provided extender over the plain U+0640 TATWEEL fallback.

Adds hb_ot_layout_get_extender_glyphs(face, script, language,
start_offset, &count, glyphs) following the standard iterator idiom
used by hb_ot_layout_get_attach_points and friends: NULL for size
probing, inout count for pagination.

Only the script tag is consulted: per the OT spec, extender glyphs live
directly on JstfScript and do not vary by language. JstfLangSys carries
per-language justification priorities, but exposing those is both
harder and less common among fonts than the (already rare) JSTF table
itself; that can be a separate accessor later.

Co-authored-by: Helmut Januschka <helmut@januschka.com>

* [JSTF] Minor simplification by refactoring exposed methods

---------

Co-authored-by: Helmut Januschka <helmut@januschka.com>
Co-authored-by: Behdad Esfahbod <behdad@behdad.org>
2026-11-11 18:58:24 -06:00
Behdad EsfahbodandClaude Opus 4.8 5ee557a991 [subset] Bound avar2 offset-compensation (binary search instead of quadratic) (#6642)
* [subset] Bound avar2 offset-compensation's quadratic breakpoint work

avar::_subset_avar2() builds the offset-compensation knots and runs
_avar2_estimate_offset_error() twice per narrowed axis; every
breakpoint and every estimator sample is looked up in the old
segment map, the instantiated mapping and the knots by linear
search, so the cost is quadratic in the axis's avar breakpoints and
nothing bounded it (42 s of CPU for a 131 KB font, re-run on each
serializer out-of-room retry).

Add HB_BUDGET_SUBSET_TABLE to hb-limits.hh (same hb_budget_spend()
idiom as the rendering budgets) and, right before the knot loop and
the estimator, charge (256 + breakpoints) * 8 * breakpoints from a
budget shared by all axes of the table.  Charging at the point where
the quadratic work starts covers the O(n^2) knot-collection loop and
its map_float() calls too.  On exhaustion the table fails through
c->serializer->check_success(), so the whole subset fails
(hb_subset_or_fail() returns NULL) rather than avar being silently
dropped from an avar2-instanced font.

The budget is 1<<30: a 5,000-breakpoint axis narrowed to -900:900 is
still admitted (0.6 s), while 16,385 and 32,769 breakpoints are
rejected in 0.01 s (the 42 s DoS).  A reviewer may tighten it; real
avar maps have a handful of breakpoints, far below the cap.

Adds test_avar2_restrict_many_breakpoints to test/api/test-subset-avar2.c.

This caps the cost; it does not remove the quadratic algorithm.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-CWJXX1BL.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [subset] Bound avar2 offset-compensation tuple memory to the table size

avar::_subset_avar2() adds one item_variations_t tuple per retained
avar v1 breakpoint and per biased axis, and add_tuple() stores each
as a dense column with one bool and one float per row of the VarData
it joins.  VarData.itemCount can be 65535 with zero-width rows (a
6-byte VarData with no regions passes sanitize), so breakpoints x
rows x 5 bytes is not bounded by the font: a 20 KB font drove
hb-subset to 2.8 GB, and add_tuple() silently dropped the failed
allocation so the result still exited 0.

Charge each tuple's item_count deltas to a budget before the
add_tuple() call, at all three avar2 sites (empty-region bias, knot
tents, free-axis bias), with the existing hb_budget_spend() idiom.
The budget is max(HB_AVAR2_MAX_DELTAS_MIN, HB_AVAR2_MAX_DELTAS_FACTOR
x table length) -- 128 deltas per byte, floor 65536, new macros in
hb-limits.hh next to the other table-derived limits.  Exhaustion
fails the subset through c->serializer->check_success(false) rather
than dropping avar.  The attack font is now rejected with a 19 MB
peak instead of 2.8 GB; valid fonts (a few deltas per byte) stay far
below the budget.

Adds test_avar2_empty_rows to test/api/test-subset-avar2.c: 513
breakpoints (which clear the quadratic estimate budget) with a
65535-row region-less VarData, so the delta budget is what rejects
it.

The charge is placed in the caller; the sibling improvement of
making add_tuple() return failure instead of swallowing the dropped
allocation lives in hb-ot-var-common.hh and is left for that file's
own change.  The as_item_varstore() rows x regions allocation and
duplicate_row() remain uncovered, as in the report.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-3HO8JDDV.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [subset] Binary-search avar2 offset compensation instead of budgeting it

Offset compensation looked every breakpoint and estimator sample up in
the old segment map, the instantiated mapping and the knots by linear
search, and inserted each breakpoint knot by shifting, so its cost was
quadratic in the axis's breakpoints.  The previous commit capped that
with HB_BUDGET_SUBSET_TABLE; remove the quadratic work instead:

- SegmentMaps::map_float() takes a sorted flag; for maps whose
  fromCoords ascend it finds the exact-match run and the segment by
  binary search, which lands where the linear scans do.
- The instantiated mapping is sorted by fromCoord, and the knots by z,
  so _avar2_map_new_mapping() and _avar2_eval_offset() binary-search.
- The estimator pulls the ascending knots back through the mapping
  with one forward scan over its toCoords.
- Breakpoint knots are pushed and then sorted, keeping the first of
  equal z as _avar2_add_knot() did.

An old segment map with fromCoords out of order, which the spec
forbids, would still need a linear scan per lookup; for such an axis
keep the {-1, 0, +1, z_old} anchors and skip the estimate.

Remove HB_BUDGET_SUBSET_TABLE, now unused.

test_avar2_restrict_many_breakpoints now expects the 32,769-breakpoint
axis to instance, and to match, rather than fail; the whole test runs
in 0.2 s.  65,000-breakpoint maps instance in under 0.25 s.

Outputs are byte-identical to before for every test-subset-avar2
instance and for 317 random avar2 maps (sorted, with duplicate
fromCoords, and with non-monotone toCoords) on random ranges; only
maps with unsorted fromCoords differ.

Tested: meson test -C build (277/277 OK).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* [subset] Reuse the one bsearch loop for avar2 instead of a new one

The avar2 offset-compensation searches went through a hand-rolled
_avar_partition_point loop -- a second binary search living next to
hb_bsearch_impl.  Route them through hb_bsearch_impl instead, with
comparators that never return 0 so it reports the insertion position
(a lower or upper bound, per the comparator) rather than a match.

No behavior change: the full test suite passes and the CWJXX1BL and
3HO8JDDV reproducers stay bounded (0.13 s CPU and 21 MB peak, the
table failing cleanly), same as before.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 18:39:32 -06:00
Behdad EsfahbodandClaude Opus 4.8 420bb6d0ba [subset/cff] Bound hint-drop/closure walks and share the budget across the whole subset (#6638)
* [subset] Bound the CFF hint-dropping and closure walks

drop_hints_in_str()/drop_hints_in_subr() and the closure walk
(collect_subr_refs_in_str/_in_subr) recursively walk the subroutine
calls recorded for each glyph with no depth limit and no work budget.
The records need not match what the interpreter executed: a string is
parsed on its first run only, global subr records are shared across
Font DICTs whose local subrs differ, and endchar does not end a walk.
So the recorded call graph can nest deeper than kMaxCallLimit or be
cyclic, and a binary-tree subr structure makes the hint-dropping walk
expand exponentially.  A 360-byte CID font crashed the subsetter with
a stack overflow under --no-hinting (or hb_subset_preprocess), and a
~1 KB font ran unboundedly.  PR #6428 rejected stack-supplied call
operands but added no depth, cycle or work limit to these walks, so
these fonts were still affected.

Hold both walks to the interpreter's call-depth limit (kMaxCallLimit),
which stops the cycle and the long chain before the stack runs out, and
charge the hint-dropping walk to the subsetting pass budget introduced
for the shared-subroutine work, so the exponential expansion is capped
too.  A walk that follows the interpreter's path stays within both, so
only fonts whose records diverge from their execution are affected;
over either limit the subset fails, as other CFF planning failures do.

Adds test_subset_cff1_subr_loop, which subsets the 360-byte cycle font
with --no-hinting and only requires that hb_subset_or_fail() returns.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-BCC7WHCA.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [subset] Stop CFF subr walks at the call-depth limit instead of failing

The hint-dropping and subr-closure walks enforce kMaxCallLimit
themselves, because the recorded call graph can nest deeper than the
interpreter allows or be cyclic.  Hitting the limit failed the whole
subset, and the hint-dropping walk signalled it by writing -1 into the
shared work budget, conflating "too deep" with "out of budget" and
clobbering the budget for later passes.

Make both walks best-effort at the limit: the closure walk keeps the
subr it reached but does not walk into it, and the hint-dropping walk
does not walk into the subr and keeps the call and its hints.  Only
running out of the work budget fails the subset now, so
drop_hints_param_t no longer needs set_error() and the budget is only
ever touched by hb_budget_spend().  Walks that follow the
interpreter's path never reach the limit, so real fonts are
unaffected; for fonts whose records diverge the output may keep a few
hints or call a subr whose callees were not retained, but no longer
fails.

test_subset_cff1_subr_loop now also asserts that the subset succeeds;
it failed before this change.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* [subset] Share one CFF charstring budget across a whole subset

Each CFF / CFF2 charstring pass of the subsetter (seac closure, CFF2
extents for instancing, CFF2 usage and instancing passes, subroutine
subsetting with hint dropping, and flattening) created its own budget
with CFF::get_subset_budget(), so one subset could spend that budget
several times over, and the two CFF2 instancing passes reset it
between them.

Keep the budget on hb_subset_plan_t instead: cff_budget() sizes it on
first use from the table, as before, and every pass draws on the same
counter, so the total work of one subset is bounded once.  Retries of
a table with a larger serialize buffer draw on it too.  cff1 depend()
is not part of a subset plan and keeps its own budget per call.

Real fonts use about 1.5 operations per table byte per pass, against
a budget of 32 per byte (at least 2^24).  Tested: full meson test and
the subset suite; hb-subset on SourceSerif4Variable-Roman.otf (CFF2),
Roboto-Regular.otf (CFF), NotoSansCJKhk-VF.otf (30 MB CFF2) with
--gids='*' and --unicodes='*', with and without --no-hinting,
--desubroutinize and partial / full instancing, and on every .otf in
/Library/Fonts with --gids='*', all succeed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* [subset] Detect CFF charstring budget exhaustion at zero, not below

The shared CFF charstring budget is only ever spent through the
cs_interpreter_t interpret() cap, which clamps each glyph's operator
count to the remaining budget and charges back exactly what it
consumed.  By construction it never overshoots, so an exhausted
budget lands at exactly 0 rather than going negative.

The exhaustion checks therefore compare budget <= 0 (cff1 depend()'s
seac walk) and budget > 0 (the subset planner's seac-closure and CFF2
get_extents verdicts), not < 0 / >= 0, which that path can never
reach.  With < 0, cff1 depend() never failed, so
hb_subset_depend_from_face_or_fail() returned a result for an
adversarial font that exhausts the budget even though the equivalent
subset fails; that is the test-subset-cff1 shared-subrs case on CI
builds with -Dexperimental_api=true.  Document this at each check.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 18:33:22 -06:00
Behdad EsfahbodandClaude Opus 4.8 41a2880106 [CFF2] Charge blend region-scalar and default-value evaluation to the budget (#6637)
* [CFF2] Charge blend region-scalar evaluation to the work budget

cff2_cs_interp_env_t::process_blend() evaluates one VarRegionList scalar
per entry of the VarData's region indices (up to 65535, each walking up
to axisCount axes) the first time a charstring hits `blend`, but the
only budget charge for blend is the operand-proportional one in
cff2_cs_opset_t::process_blend(), which runs afterwards and is skipped
when the operand check fails.  So `1 blend` did regionIndexCount x
axisCount scalar evaluations for one budget unit, and a VARC fan-out
re-ran it tens of thousands of times within 2% of HB_BUDGET_GLYPH.

Charge region_count x max(1, axisCount) through the env's existing
spend_budget() right before resizing and filling the scalars vector,
matching how VARC charges values x coords for its own
ItemVariationStore evaluations; a failed charge errors the interpreter
exactly as a failed resize does.  spend_budget() is a no-op without a
budget, so the subsetter and instancer paths are unchanged, and
seen_blend gates the work to once per charstring run.

Adds test_hb_draw_cff2_blend_region_budget: a CFF2 charstring blending
one value over 256 region indices of a 16-axis region list must consume
at least 256 x 16 units and fail under a 2048-unit budget.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-WSM91D3G.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [CFF2] Charge the blend default values to the work budget

cff2_cs_opset_t::process_blend() walks all (k+1)*n values of a blend
(n default values plus k deltas each) but charged only the k*n deltas,
total - n.  With a VarData of regionIndexCount 0 (or no variation
store) k is 0, so `n blend` charges nothing, pops nothing, and the same
operands can be blended again and again for the price of two tokens.

Charge `total` = (k+1)*n, exactly what the loop touches.  spend_budget()
is a no-op without a caller budget, so only budgeted callers (draw,
extents, paint, shape, and the subsetter passes that now pass a budget)
are affected, and a well-formed charstring pays one extra unit per
blended value.

Adds test_hb_draw_cff2_blend_budget: a charstring that pushes 100
numbers and blends them 100 times over a region-less VarData must
consume at least 100 x 100 units and fail under a 5000-unit budget.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-0DAQUESS.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 17:51:25 -06:00
Behdad EsfahbodandClaude Opus 4.8 6aa278829e [subset] Charge gvar avar2 culling pass to the source table size (#6641)
In a well-formed gvar the per-glyph data ranges [offset[i],
offset[i+1]) are disjoint, so the retained glyphs' data add up to at
most the table.  gvar::subset()'s avar2 culling pass used the offsets
as they are, so malformed offsets (0, D, 0, D, ...) could point
thousands of glyphs at one 128 KB block; the pass parsed, rewrote and
kept a heap copy of it once per glyph, repeated on each of the
serializer's buffer-doubling retries (26 s and 1.3 GiB from a 251 KB
font before it was rejected).

Restore the invariant where the pass is driven: charge each retained
glyph's byte length to a budget equal to the source blob length with
hb_budget_spend(), and on exhaustion raise HB_SERIALIZE_ERROR_OTHER
so _hb_subset_table_try() does not retry.  A spec-conforming table
can never reach the limit, so there is no tuning constant and valid
fonts subset byte-identically; a font whose retained glyphs' gvar
ranges overlap by more than the table length now fails avar2 partial
instancing.

Adds test_avar2_culling_aliased_gvar_data to test/api/test-subset-avar2.c.

The non-avar2 instantiate() branch trusts the same offsets and is
not touched here.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-NJZQ4XFP.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 17:04:43 -06:00
Behdad EsfahbodandClaude Opus 5.5 a1134fa3ea [subset] Budget kept glyf points and the avar2 reachable-range computation (#6640)
* [subset] Budget the glyf points a plan keeps for partial instancing

For partial instancing the subset plan decodes every retained glyph
through Glyph::get_all_points_without_var() and keeps all the points in
plan->new_gid_contour_points_map until the plan is destroyed. That
caller passed no work budget, and nothing capped the total. A simple
glyph of repeated flags declares ~65536 points in ~526 bytes, and a
loca whose offsets alias one another lets many glyph ids share those
bytes, so a 25 KB font drove plan creation to 2.2 GB.

PR #6256 added a charge-before-allocate budget to
SimpleGlyph::get_contour_points() and passed it from the draw/extents
path, but left this subsetter caller unbudgeted. Give the loop one
plan-wide budget, the larger of HB_BUDGET_SUBSET_CONTOUR_POINTS and the
source glyf length, and charge each glyph's decoded points after it is
decoded. One glyph's decode is bounded by the format (about 786 KB),
so the transient is proportional to the input; the budget bounds what
the plan retains. Exhausting it fails plan creation, the same outcome
as any other decode failure in this loop. Ordinary glyph data carries
at least one byte per point, so the size-derived bound does not
constrain well-formed fonts.

Adds an API regression test (test_subset_gvar_shared_glyph_data) with
an aliased loca over a 65536-point simple and composite glyph.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-O2ZFVVUI.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [subset] Budget the avar2 reachable-range computation

_compute_avar2_reachable_ranges() bounded only the breakpoint grid
(MAX_GRID = 16384 vertices), while its real work is
axes x vertices x active row columns x tents. A delta-set row may list
the same region tens of thousands of times, and a clamped
DeltaSetIndexMap lets every fvar axis share that one row, so a 50 KB
font cost about two minutes of CPU per hb_subset_plan_create_or_fail().

Charge one shared allowance per plan, HB_BUDGET_SUBSET_AVAR2_RANGES, at
the three places the work is created: scanning the row's columns into
the active set, collecting each active region's tents, and the exact
grid enumeration (charged up front as grid_size x (active + tents)).
Degradation reuses states the surrounding code already handles: an axis
whose exact enumeration does not fit takes the existing
interval-arithmetic bounds (still conservative, so the instance stays
correct), and once the allowance is gone the remaining axes are left
unconstrained. The hostile fonts now finish in well under a second;
real avar2 fonts use a small fraction of the budget and their
output is unchanged.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-3VDGZ019.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [subset] Guard against null glyf blob in contour-point budget

get_instance_glyphs_contour_points() sizes its op budget from the
source glyf table length.  When head.glyphDataFormat (or
indexToLocFormat) is out of range the glyf accelerator is disabled and
reference_glyf_table() returns no blob, so hb_blob_get_length() then
dereferenced a null pointer and crashed during partial instancing.

Take the length only when the blob is non-null.  Add a regression test
that builds a variable font with an invalid glyphDataFormat and plans a
partial instance.

Reported by Anthropic's OSS Scanner.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [subset] Document that the avar2 range budget fallback is safe

Running out of HB_BUDGET_SUBSET_AVAR2_RANGES returns early from
_compute_avar2_reachable_ranges() and records nothing for the
remaining axes.  That does not produce a wrongly instanced font:
avar2_reachable_ranges is only used to cull variation data that
cannot be reached, and an axis without a range is never culled
(_hb_avar2_region_is_dead() skips it, the same as the VARC path that
resets the map).  A pinned axis that is not detected as
self-contained stays in fvar as a hidden pin, which is what CFF2
fonts always get.  The output is correct, only larger.  Likewise the
exact-grid fallback uses interval bounds, which are conservative.

Say so in the comments instead of "left unconstrained".

Tested: meson test -C build (278/278); hb-subset --instance
wght=400 and wght=300:700 on GoogleSansFlex, NotoSansKR and
SourceSerif4 variable fonts, all glyphs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 17:03:53 -06:00
Behdad EsfahbodandClaude Opus 4.8 7b30ba988d [instancer] Budget VarStore rebuild, tuple splitting and per-point tuple expansion (#6639)
* [subset] Bound the dense delta matrix when rebuilding a VarStore

item_variations_t::as_item_varstore() allocates one int per
(row, region) pair for the whole rebuilt ItemVariationStore, every row
as wide as the full region list. In the font a row only pays for the
regions of its own VarData (possibly none, and several dataSets offsets
may share one VarData), so the row count and the region count are
independent and their product is unbounded: a ~120 KB font drove the
allocation to 2.4 GB, and a 972-byte font with region-less rows to
~800 MB.

Bound the matrix before allocating it, at the single site all
instancing callers (MVAR, HVAR/VVAR, GDEF, BASE, COLR, avar) pass
through: HB_SUBSET_MAX_VARSTORE_ROWS (1<<21, which also bounds the
per-row hb_vector_t overhead and catches the zero-width-row variant)
and HB_SUBSET_MAX_VARSTORE_DELTAS (1<<26 cells, 256 MB of int),
computed in 64-bit. These are flat ceilings on this one allocation
rather than a bound proportional to the source size: optimizer output
routinely has several VarData over disjoint region subsets, so the
dense matrix is legitimately much wider than any single VarData and a
proportional bound would reject it. No store in the test fonts comes
near the cell cap. Returning
false is the path a failed resize() already takes, and every caller
drops the table on it. Rows are fixed-width, so they are now allocated
with resize_exact(), removing the ~19% growth slack per row.

The cap bounds the amplification rather than removing it (a ~75 KB
font just under the cell cap still peaks around 265 MB); the
structural follow-up is to stop making every row as wide as the whole
region list.

Adds a unit test (test_delta_matrix_limits) covering both caps and an
under-cap case.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-0ZZNV0U6.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [instancer] Budget tuple splitting when narrowing axes

Narrowing an axis makes change_tuple_variations_axis_limits() replace
every tuple whose region uses that axis with up to six tuples
(rebase_tent solutions), axis after axis with no merging, so one region
spanning N narrowed axes becomes up to 6^N tuples, each with its own
copy of the axis map and deltas. A 364-byte font with a region on 8
narrowed axes drove one hb_subset_or_fail() call to 9 s of CPU and
1.7 GB. The only existing limit, the subset output buffer cap, is
checked after the whole instantiation has run.

Charge the copies to a cumulative budget held in optimize_scratch_t,
which every caller (gvar, cvar, CFF2, and item_variations_t for
GDEF/HVAR/MVAR/BASE/COLR) already keeps for one whole table. The budget
starts at HB_BUDGET_INSTANCER_MIN and earns HB_BUDGET_INSTANCER_FACTOR
times the cost of each source tuple; each split spends
out[0].get_cost() x (out.length - 1) through the existing
hb_budget_spend() helper, and the function returns false when the
budget is spent, which all callers already treat as failure. The cost
(32 + 8 per axis + 1 per row/point) scales with what a copy allocates
for both ItemVariationStore and gvar tuples, so the budget also bounds
a store with many VarData and a gvar with many glyphs.

An instancing request whose splits exceed the budget (roughly more than
64 copies per source tuple beyond a 4M-unit floor) now fails instead of
producing a store that could not be serialized anyway. Normal fonts
stay far below the floor, so the subset suite's partial-instance
tests are byte-identical.

Adds a unit test (test_tuple_split_budget) with a tent on 16 axes.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-X7UB9BVO.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [subset] Budget the dense per-point expansion of gvar/cvar tuples

create_from_tuple_var_data() expands every tuple to arrays with one
entry per point of its glyph (per cvt value), however few points the
tuple actually names, and keeps them until the table is serialized. The
tuple count (12-bit) and point count (16-bit) are each bounded per
glyph, but their product and its sum over the table are not, and neither
factor costs many font bytes: 15 bytes of gvar can expand to ~590 KB,
and a 31 KB font reached 2.1 GiB.

Give the decode a per-table points budget,
HB_BUDGET_SUBSET_TUPLE_POINTS_FACTOR per byte of the source table
floored at HB_BUDGET_SUBSET_TUPLE_POINTS_MIN, owned by
glyph_variations_t for gvar and a local in cvar::subset(), and charge
point_count per decoded tuple before expanding it. Exhausting it
returns false exactly like the function's existing allocation-failure
paths (the table is dropped from the subset). Real gvar tables
expand to only a small multiple of their byte size, so the factor
leaves ample headroom. The extra copies made when a range limit splits
a tuple are already bounded by the instancer split budget in
optimize_scratch_t, so they are not charged here again.

Adds a unit test (test_tuple_points_budget) covering the decode charge.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-8CH2PGW8.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 17:02:43 -06:00
Behdad EsfahbodandClaude Opus 4.8 a6beb11493 [subset/cff] Bound parsed subrs and share one charstring budget per pass (#6636)
* [subset/cff] Bound parsed subroutines by the table size before allocating

subr_subsetter_t::subset() resizes one parsed_cs_str_vec_t per Font DICT
to that DICT's localSubrs->count, assuming the sum is bounded by the
table.  It is not: any number of Font DICTs may point at the same
Private DICT and Subrs INDEX, so a 500 KB font with 16 Font DICTs all
sharing one 1000-subr INDEX cost about 1.6 GiB (measured: control 10
MiB, shared 1651 MiB; 2.8 GiB with hb_subset_preprocess).

Sum the global plus per-Font-DICT local subroutine counts in 64-bit
before any per-Font-DICT allocation, and fail the subset when the
records would exceed 256 bytes per byte of the source table, the same
ratio _hb_subset_table_try() uses to stop growing the output buffer.
A font whose Font DICTs do not share their Subrs needs at least one
table byte per subroutine, so it stays far below the limit; every
CFF/CFF2 font in the test corpus is at most a couple of bytes per table
byte.  Over the limit the subset fails, as any CFF planning failure has
since dde7da182e; the shared font now peaks at 7 MiB instead of 1.6 GiB.

The check sits before the allocations in the non-accelerator branch, so
it also covers the hb_subset_preprocess path and both CFF1 and CFF2
instantiations of the template; the --desubroutinize flattener path
does not allocate per subroutine and is unaffected.

Adds test_subset_cff1_shared_local_subrs: a 16-Font-DICT CID font with
1000 local subrs subsets when the subrs are unshared and fails when
every Font DICT shares them.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-HPFG1IAV.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [subset] Share one CFF charstring work budget across a subsetting pass

Every per-glyph charstring loop in the subsetter (subr_subsetter_t::
subset, subr_flattener_t::flatten, the two CFF2 partial-instancing
passes, the CFF2 instance-metrics loop, and the CFF1 seac closure in
the plan and in the depend builder) created a fresh interpreter bounded
only by HB_CFF_MAX_OPS per charstring.  Glyphs can share subroutines,
so a font where every glyph makes one cheap call into a deeply nested
shared subr cost num_glyphs x ~200,000 ops: a 121 KB, 30,000-glyph font
took 33 s to subset (control 0.05 s).

Give each pass one cumulative counter, max(HB_BUDGET_SUBSET_CFF = 2^24,
32 ops per table byte), threaded through cs_interpreter_t::interpret()'s
existing budget argument exactly as the outline and extents paths do.
When it runs out, interpret() fails as it does for a single over-limit
charstring, so the CFF/CFF2 table fails to subset (the subset fails, as
any CFF planning failure has since dde7da182e); the seac closure fails
the plan and the depend builder fails.  The 32-ops-per-byte allowance
was measured against the in-tree corpus, where real fonts interpret at
most a few operations per CFF byte, so they stay far below it and
produce byte-identical output; the attacker font now fails in 0.17 s.

Adds test_subset_cff1_shared_subrs: four shared-subr glyphs subset
fine, while a hundred of them exhaust the budget and the subset (and
the dependency builder) fail.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-Z5NLRFA6.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 16:58:48 -06:00
Behdad EsfahbodandClaude Opus 4.8 db38d0c7e0 [layout] Charge the skip, skipped-record and ligature walks to max_ops (#6634)
* [test] Force the ot shaper in the layout budget tests

test_ot_layout_gpos_mark_base_rescan_budget calls hb_shape() with no
shaper forced, then asserts an OT-specific max_ops trip via
hb_buffer_allocation_successful().  Under the HarfRust shaper
(HB_SHAPER_LIST=harfrust, as the rust CI job re-runs the api suite)
it aborts, because harfrust shapes independently and never reaches
the C-side max_ops paths it tests.

Shape with hb_shape_full(..., {"ot", NULL}) so it exercises the OT
shaper regardless of the configured shaper list, as
test_shape_lookup_budget already does.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [layout] Charge the skip, skipped-record and ligature walks

The per-pass `len / 16` charge does not bound per-position inner work:
the skipping iterator's rescans over default-ignorables, apply_lookup's
walk over a rule's skipped LookupRecords, and match_input /
ligate_input's walk over a ligature's components and marks. Each is a
CPU-DoS a crafted font triggers under the default max_ops.

Charge each on its work path only, so honest shaping is unaffected:
skipping_iterator_t::next/prev charge one op per skip (a first-try
match charges nothing); apply_lookup charges each skipped record;
match_input charges the ligbase backward walk and ligate_input the
mark re-adjust walk, once each.

Reproducers drop from 20 s / 6.8 s / 3.8 s to 0.05 s / 0.003 s /
0.02 s; in-house and aots shaping tests pass with no output change
(the pre-existing text-rendering-tests failures are unchanged).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [layout] Stop matching once the skip/walk budget is spent

The previous commit charged the walks, but the skipping iterator's skip
arm only decremented max_ops; within one pass a tiny aliased font could
still re-run a walk per candidate, per rule or per subtable without bound
(dfrg's harfrust #604 review found the same gaps here).

Flag the buffer unsuccessful when the skip arm spends the budget -- it
still never early-returns, so *unsafe_to stays set for the kern machine --
and stop the loops that repeat walks once it clears: apply_backward,
Ligature::apply, the reverse-chain and mark-base/lig subtable applies,
and context_apply_lookup / chain_context_apply_lookup. apply_forward
already checks successful.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [layout] Enforce the budget in the shared subtable loop

dfrg's re-review (harfrust #604): the per-kind entry guards miss paths
that scan before reaching them -- context format 3, the context/ligature
pre-match scans, cursive and mark-to-mark -- so a lookup aliasing thousands
of subtables restarts a buffer scan at each one. Guard the shared
accelerator apply instead (hb_applicable_t::apply), and drop the
now-redundant reverse-chain and mark-base/lig entry guards it subsumes.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 16:57:53 -06:00
Behdad EsfahbodandClaude Opus 4.8 8d01a1b148 [subset/cff2] Budget the region transforms built for partial instancing (#6635)
cff2_instancing_plan_t::create() builds k identity tuples of k cells
(then keeps a k-float column per surviving solver tuple) for every
entry of the CFF2 ItemVariationStore's VarData offset array, and
nothing bounded k (a 16-bit region count), the number of entries, or
their aliasing of one VarData.  The work is therefore quadratic in a
font field, so a 12 KB font cost 2.2 GB.

Charge those cells, before they are allocated, to a per-table budget
HB_BUDGET_SUBSET_CFF2_CELLS (2^24 cells) via the existing
hb_budget_spend helper, at both allocation sites (the k*k identity
tuples and the surviving solver columns).  Exhausting it fails create()
like any other planning failure, so the partial instance fails safely
(bounded at ~175 MB worst case) instead of exhausting memory.

Adds test_instance_cff2_vardata_budget: a 6.8 KB font with 40 offsets
aliasing one unused 800-region VarData must fail to partially instance
rather than build a region transform for each; it succeeds (keeping
CFF2 at hundreds of MB) before this change.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-I7KWU2X6.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 13:12:41 -06:00
Behdad EsfahbodandClaude Opus 4.8 79146f902e [depend] Bound lookup recursion depth and whole-table path count
hb_depend_context_t::recurse() only refused to re-enter a lookup already
on the call stack, so a chain of distinct Context lookups recursed once
per lookup (stack overflow around 8600 lookups on an 8 MB stack) and
lookups invoking each other in every order were walked once per ordering
(a factorial number of interned paths, gigabytes of memory).

recurse() is where every sibling context (closure, closure-lookups,
apply) enforces HB_MAX_NESTING_LEVEL, so add the same nesting_level_left
guard here; shaping never applies lookups nested deeper, so no edge the
closure can reach is lost (a chain's final edge drops at exactly the
depth the closure stops, 65->66).  Bound the number of distinct
recursion paths with a counter kept over the whole table (not cleared by
reset_recurse_cache, which runs per top-level lookup) against a new
HB_MAX_DEPEND_RECURSE_PATHS limit in hb-limits.hh, so it can be tuned
on its own; exceeding it fails the build so
hb_subset_depend_from_face_or_fail() returns nullptr instead of emitting
a truncated graph.  The largest whole-table path count among the test
fonts is 744 (NotoNastaliqUrdu-Bold), so 35000 leaves ample headroom.

Builds only with experimental_api.  Adds a small permutation-font fuzz
seed and a test-ot-depend test; test-ot-depend, the depend fuzzer and
the depend-closure-parity tests pass.  Related to PR #6185.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-N28924W8.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-11 12:58:33 -06:00
Behdad EsfahbodandClaude Opus 4.8 c611a0589d [depend] Release oversized recursion maps in reset_recurse_cache()
reset_recurse_cache() runs once per top-level lookup and cleared
completed_recursions and recurse_path_to_index with hb_hashmap_t::clear(),
which walks the whole table and keeps its allocated capacity.  A single
lookup that grew these maps large therefore made every later top-level
lookup pay to walk that full capacity, however little it recursed,
giving a font with one busy lookup followed by many trivial ones a CPU
DoS in hb_subset_depend_from_face_or_fail().

Add clear_or_release(): when a map is large but was left mostly empty by
the lookup that just finished, swap in fresh storage before clearing, so
later lookups start from a small table.  Adds a fuzzer regression font.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-PNJPHF76.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 12:58:33 -06:00
Behdad EsfahbodandClaude Opus 4.8 0c8b00908b [GPOS] Charge propagate_attachment_offsets() base-to-mark walk to max_ops
propagate_attachment_offsets() turns a mark's stored attachment into an
offset by summing the advances of every glyph between the mark and its
base.  That distance can be up to 32767 glyphs (the range of
attach_chain), and position_finish_offsets() runs the pass once per
attached glyph, so the total work is the sum of all mark-to-base
distances and is not bounded by the font; it was never charged to
buffer->max_ops.  Attacker-chosen text over a cooperating font (or the
kerx path) could thus make hb_shape() spend several times the budgeted
CPU.

Thread the buffer into propagate_attachment_offsets() and charge the
base-to-mark distance to buffer->max_ops before the summation loops,
failing the shape when the budget runs out (the same budget the
unsafe_to_break walk over this range already uses).  Normal text, whose
marks sit next to their base, pays a negligible amount.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-4JGOSACZ.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 12:43:10 -06:00
Behdad EsfahbodandClaude Opus 4.8 0aa7da1772 [GPOS] Charge out-of-order MarkBasePos/MarkLigPos base rescans to max_ops
MarkBasePosFormat1 and MarkLigPosFormat1 find a mark's base with a
cache (c->last_base / last_base_until) that is only amortized while
buffer->idx moves forward.  A (Chain)Context lookup that recurses into
the mark lookup at input positions out of order resets the cache, so
each application rescans the glyphs from last_base_until to idx from
scratch; on a long run of marks with no base this is quadratic and was
uncharged.

Charge the rescanned span (idx - j) to buffer->max_ops after the
backward search in both lookups, failing the shape when the budget runs
out.  A real mark-heavy font pays a negligible amount.  Adds a
regression test shaping a Context lookup that recurses to MarkBasePos
at positions 1,0,1,0,...

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-FJEXEDJ4.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 12:43:10 -06:00
Behdad EsfahbodandClaude Opus 4.8 339aa688fa [layout] Charge each GSUB/GPOS lookup pass to max_ops
Lookups can share subtables, features can share lookups, and each
shaping stage applies its lookups anew, so the number of passes over
the buffer is not bounded by the font size.  An earlier lookup can
also grow the buffer far beyond the text the budget was sized from:
a font that grows one character into a long buffer and then walks it
with thousands of lookups (as in harfbuzz/harfrust#571) made hb_shape()
spend CPU out of proportion to the text, and none of it was charged
to buffer->max_ops.

Charge each pass in apply_string() len / 16 for walking the buffer,
even if it applies at no glyph (skipping a glyph is cheap, hence the
division; applying at one is charged only through the buffer
operations and nested lookups it does), and stop the stage loop in
hb_ot_map_t::apply() once the buffer has failed.  Nothing is charged
in the matching hot loops (skipping iterator, RuleSet, ChainRuleSet,
LigatureSet): charging there cost real fonts most of the budget in
HarfRust (Gulzar spent about half of max_ops on ordinary Urdu prose,
Noto Sans Duployan ran out on its own sample text).  Aliased rules and
ligatures stay bounded by the sanitizer, including the ChainRule array
charge.  Matches harfbuzz/harfrust#601.

test_shape_lookup_budget grows one 'a' to 65025 glyphs before applying
aliased lookups, and checks that the pass charge stops the shape.

With the font from harfbuzz/harfrust#571 (one 'A' grown to 30,000
glyphs, then 2,000 lookups), hb_shape() takes about 10 ms instead of
0.6 s, stopping after about 35 passes with the glyphs kept.  Gulzar
and Noto Nastaliq Urdu spend under 0.5% of max_ops on Urdu and Persian
prose.  Noto Sans Duployan spends up to 45% on its sample text, as on
main: nearly all of it is the _infos_set_glyph_flags() charge from
#5522; the per-pass charge adds about 5%.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-6A9E6OEX.
Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-K2T2LOJH.
Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-DVG5XAU0.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-11 12:43:10 -06:00
Behdad EsfahbodandClaude Opus 4.8 8f93b6d21b [layout] Charge GSUB/GPOS sanitize-time arrays to max_ops (#6631)
* [sanitize] Charge ChainRule backtrack/input/lookahead arrays to max_ops

ChainRule::sanitize() checks the length of each of its three arrays but
only sanitizes the trailing LookupRecord array, so it never charges
max_ops for the bytes it skipped over.  A table with many ChainRuleSet
offsets all pointing at one large ChainRule could therefore be
sanitized in time proportional to offsets x rule size, escaping the
sanitizer's work bound (restoring the accounting removed by 385896ff6).

Charge the skipped span (from this to the LookupRecord array) to
check_ops before sanitizing the records.  Adds a fuzzer regression
font.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-EFNNHRHV.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [GPOS] Charge zero-size ValueRecord arrays to the sanitizer op budget

When a ValueFormat is empty its records take no bytes, so
ValueFormat::sanitize_values() and PairPosFormat2::sanitize() pass
check_range() without charging max_ops for the records, while later
code (the subsetter, variation-index collection) still walks count
records.  A tiny table can thus declare a huge valueCount / class
matrix and cost the subsetter CPU out of proportion to its size.

Charge count ops when the record size / stride is zero, following the
918193ebf precedent.  Adds two fuzzer regression fonts.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-HFGOLC64.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-11 12:43:06 -06:00
Behdad Esfahbod 83a47049a3 Merge pull request #6519 from harfbuzz/scanner/depend
[depend] Bound dependency-graph edge count and dedup probe walk
2026-11-10 22:44:19 -06:00
Behdad Esfahbod 0eaceda537 [limits] Resolve work-budget conflicts with main
Merge main's active-glyph fast paths and precomputed ClassDef caches
with the dependency and GSUB closure budgets. Charge the fast paths
for the glyphs they visit and the class caches for their ClassDef
scans, preserving both the optimizations and the work limits.

The full Meson suite passes with experimental APIs enabled (300 tests)
and with the default configuration (281 tests). Also syntax-checked
hb-ot-layout.cc with HB_NO_BEYOND_64K.

Assisted-by: Codex
2026-11-10 19:31:06 -06:00
Behdad EsfahbodandClaude Opus 5.5 86c1bbb3b4 [cff1] Cache std_code_to_glyph() for seac component lookups (#6562)
std_code_to_glyph() resolves a seac component (bchar/achar) with
Charset::get_glyph(), a linear scan of the whole charset, and every
load of a seac charstring does it twice.  The scan is not charged to
the glyph work budget, so a COLRv0 glyph whose layers all name one
seac glyph costs seconds of CPU while the budget stays almost
untouched (the attacker font: 32 s).

Charging the scan is not a good fit here: Charset::get_glyph() does
not report how far it scanned, so there is no exact cost to charge,
and a flat per-lookup charge against the shared subset budget would
make large fonts with many accented seac glyphs fail to subset.
Instead remove the redundant work: the key space is the 256
StandardEncoding codes, so on first use build a 256-entry
code-to-glyph table in one pass over the charset, recording the
first glyph carrying each StandardEncoding SID, as get_glyph() does.
This fixes all three callers (path, extents and the subsetter's seac
closure) with byte-identical output.

The table is allocated lazily and installed with cmpexch, like the
accelerator's glyph_names, and freed in _fini(): the accelerator
grows by one pointer, HB_NULL_POOL_SIZE is unchanged, and nothing is
allocated for fonts without seac.  On allocation failure the lookup
falls back to the scan.

Tested by checking the table against get_glyph() on every lookup
across the full test suite (temporary assert), and the existing seac
tests (draw, extents, subset).

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-FH71HHTD.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 19:25:00 -06:00
Behdad EsfahbodandClaude Opus 4.8 8b0e9da51f [layout] Collect intersected ClassDef classes once per ContextFormat2 visit
The class-based (Chain)Context closure and depend paths called
ClassDef::intersects_class() for every rule-set index / class, and each
call walks the whole ClassDef, so a subtable with many classes sharing
one large ClassDef was quadratic per closure visit (and thus in
subsetting).

Collect ClassDef::intersected_classes() of the glyph set into a set
once per visit and answer the per-class query from that set.  All eight
call sites across ContextFormat2 and ChainContextFormat2 (closure and
depend) use the precomputed set; results are identical.  Adds a fuzzer
regression font.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-ZCMLEQFJ.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 19:15:42 -06:00
Behdad EsfahbodandClaude Opus 4.8 0cf5addc3e [GSUB] Add small-active-set closure/depend fast path to Coverage-keyed lookups
SingleSubstFormat2 already picks, per closure visit, between walking its
whole Coverage and iterating the (usually small) active glyph set.  The
Multiple/Alternate/Ligature/ReverseChain closure and depend paths, the
format-1 (Chain)Context closure, and Coverage::intersect_set lacked that
choice, so a nested lookup visited many times over a small active set
re-walked a large Coverage every time, giving quadratic GSUB closure
(and thus subsetting) time.

Add the same get_population()-based fast path to those paths and to
CoverageFormat1/2::intersect_set.  Output is byte-identical; only the
per-visit work is bounded.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-KNHDK379.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 19:15:42 -06:00
Behdad EsfahbodandClaude Opus 4.8 fb350edda9 [ot] Collect class-based contextual rule classes once per ClassDef in collect_glyphs
Rule::collect_second_glyphs() resolved the second input glyph of a
class-based contextual rule by scanning the whole input ClassDef once
per rule, so a RuleSet with many rules sharing one large ClassDef was
quadratic in collect_glyphs() (and thus in subsetting).

Collect the needed classes into a set and walk each ClassDef once for
all of them: add ClassDef::collect_classes(), and have
collect_second_glyphs() accumulate the classes and resolve them in a
single pass.  The collected sets are identical; only the work is
reduced.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-5FCXNOQF.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 19:15:42 -06:00
Behdad EsfahbodandClaude Opus 4.8 fd57025171 [syllabic] Record HB_MAX_SYLLABLE_LENGTH cuts so every pass sees the same pieces (#6558)
Since 12b686dad, _hb_next_syllable() cuts syllables longer than
HB_MAX_SYLLABLE_LENGTH at whatever glyph index the walk reaches, and
nothing records that cut.  The Indic shaper's initial and final
reordering (with GSUB features shortening the buffer in between) then
walk different pieces, the deferred pre-base-matra cluster merge never
happens, and hb_shape() returns non-monotone clusters at the default
MONOTONE_GRAPHEMES level for an over-long Indic syllable.

Add hb_syllabic_record_cut(), next to the other shared syllabic
helpers, and call it on each piece from the foreach_syllable() loops
that already exist where syllables are set up in all four syllabic
shapers (Indic, Khmer, Myanmar, USE), and once more in Indic initial
reordering since dotted-circle insertion can push a piece back over
the limit.  Its inline fast path is a single length compare, so it
costs nothing unless a piece reaches the limit.  Only a piece of
HB_MAX_SYLLABLE_LENGTH glyphs that is followed by more of the same
syllable takes the slow path: it cuts the rest of the syllable into
the pieces foreach_syllable() would make, gives every piece but the
last one of two serials that differ from the preceding syllable and
from the last piece (which keeps the ragel serial), so substituted
glyphs inherit it and all later walks see identical pieces, and
merges the whole oversized syllable into one cluster
(merge_clusters at monotone cluster levels, unsafe-to-break
otherwise) so the cut cannot become a line-break opportunity
mid-syllable.  Serials of all other syllables are untouched, so
numbering and wraparound are exactly what the ragel machines
produce.  The cut in _hb_next_syllable() stays as the backstop.

Behaviour change: syllables longer than HB_MAX_SYLLABLE_LENGTH become
a single cluster, where previously each piece got its own clusters.

A regression test with a 68-codepoint Devanagari input is added to
indic-syllable.tests; it fails (non-monotone clusters) on the
unpatched tree.

Tested: meson test (all pass).  hb-shape on hi-words / a 175KB
Devanagari paragraph shows no measurable slowdown.

Reported by Anthropic's OSS Scanner as ordinary bug ANT-2026-W15T9EXZ.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 19:15:38 -06:00
Behdad Esfahbod edca03eadb [bit-vector] Use bounded bitvectors in GDEF and the repacker (#6630)
Use hb_bit_vector_t and hb_min_max_t as Coverage sinks for the GDEF
mark-set cache. Preserve the exact-cache span limit and fall back to
the digest and table when allocation fails. Pack the repacker's
Dijkstra visited flags into one bit per vertex.

Fix singleton ranges and compute bit masks relative to the vector's
minimum. Avoid overflow when sizing inclusive ranges, expose allocation
failure, and add the Coverage sink methods needed by GDEF.

Add regression tests for biased ranges, singleton and allocation
boundaries, failing allocation, moves, and GDEF cache/table parity.

Testing: full meson test suite (303 passed, 1 skipped); bitvector tests
with AddressSanitizer and UndefinedBehaviorSanitizer; HB_TINY
amalgamation syntax check.

Assisted-by: OpenAI Codex
2026-11-10 17:12:12 -06:00
Behdad EsfahbodandClaude Opus 5.5 df02261eca [subset] Charge Coverage and ClassDef glyph walks to a work budget (#6628)
* [subset] Charge Coverage and ClassDef glyph walks to a work budget

Coverage::subset() expands a format 2 Coverage glyph by glyph, up to
numGlyphs, every time an offset to it is subset, and nothing bounded
that work.  A 10-byte range Coverage referenced by 30,000 two-byte
offsets in a subtable listed 8 times costs 1.6e10 iterator steps,
while the sanitizer only bounds work by table bytes; a 60 KB font
keeps hb-subset busy ~50 s.  GDEF's MarkGlyphSets reaches the same
walk with no feature or lookup needing to survive planning.

Give hb_subset_context_t a per-table work budget
(HB_BUDGET_SUBSET_TABLE, in the hb_budget_spend() idiom of
hb-limits.hh) with a check_ops() helper that fails the serializer once
exhausted, and charge Coverage::subset() its capped population before
the walk.  Charging inside Coverage::subset() covers every caller
(ReverseChainSingleSubst, (Chain)ContextFormat3, GDEF MarkGlyphSets,
...) without touching them.  The same unbounded per-reference pattern
in ClassDefFormat1_3 / ClassDefFormat2_4::subset() (the range or
glyphset walk and the glyph_filter count) is charged the same way.

A hostile table that exceeds the budget now fails the subset cleanly
(hb_subset_or_fail returns NULL) in a fraction of a second instead of
succeeding slowly; valid-font output is byte-identical and the
heaviest local table spends a small fraction of the budget.
test_subset_shared_coverage_budget builds the reverse-chaining GSUB in
memory and asserts 256 shared references subset while 2048 fail.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-LLIWV9UK.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [subset] Charge layout glyph walks to the existing subset budget

Drop the separate HB_BUDGET_SUBSET_TABLE budget and the second
hb_subset_context_t field it needed; check_ops() now charges
budget_remaining (HB_BUDGET_SUBSET_GLYF, 1 << 28), the per-table
budget glyf instancing already uses.  Both are per-table and shared
across buffer-size retries, and no table spends both.

The larger budget still bounds the attack (fails after 2^28 glyph
steps, ~3 s), so test_subset_shared_coverage_budget now expects the
failure at 8192 shared offsets instead of 2048.

Subsetting all glyphs with --layout-features='*' of Gulzar,
NotoSansKR-VF and GoogleSansFlex-VF keeps GSUB/GPOS/GDEF with output
byte-identical to before; the heaviest table (Gulzar GPOS) spends
~98K units, under 0.04% of the budget.  Full meson test: 278 OK.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:50:50 -06:00
Behdad EsfahbodandClaude Opus 4.8 70df3ff279 [layout] Charge GSUB glyph closure to a per-session work budget
hb_closure_context_t bounded GSUB glyph closure only by counting lookup
visits (HB_MAX_LOOKUP_VISIT_COUNT per round), and forgets which lookups
are done whenever the glyph set grows, so what one visit walks -- through
aliased offsets, many times the table size -- was charged to nothing.  A
LigatureSubst with 9000 ligatureSet offsets aliased onto one 64-component
Ligature, re-walked after every one of 1700 Context triggers that each
add a glyph, makes hb-subset --unicodes='*' of an 86 KB font spend ~110 s
of CPU.  A ReverseChain/ChainContextFormat3 with aliased backtrack
Coverages is a second way in.

Give the context an int64_t budget (HB_BUDGET_LAYOUT_CLOSURE records, in
the hb_budget_spend() idiom) spent as each closure() implementation walks
its arrays: dispatch() charges a visit its coverage record count;
ligature sets / ligatures / components, sequences, alternates,
single-subst intersections, rule sets and rules, context Coverages (via a
spend_coverages() helper) and the per-rule-set ClassDef walks are each
charged.  Coverage and ClassDef gain a get_record_count() helper because
intersects()/intersect_set() are linear in records, not population.  Once
the budget is spent, should_visit_lookup()/lookup_limit_exceeded() stop
the closure early, exactly as they do at the visit limit.

The budget lives in the context, so both public closure entry points (also
reached from hb_ot_shape_glyphs_closure() and the subset plan) are covered
with no API/ABI change.  Both attack shapes drop to ~1 s and the control
font's closure output is byte-identical; the heaviest real font in the
corpora uses about 1% of the budget.  A hostile GSUB now gets a truncated
closure, as one hitting the visit limit already does.
test_ot_layout_lookups_substitute_closure_budget builds the hostile table
in C and drives hb_ot_layout_lookups_substitute_closure() over it.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-J9K92HA9.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:42:16 -06:00
Behdad EsfahbodandClaude Opus 4.8 9b4e3df364 [depend] Clip context glyph sets to the face and charge a build budget
While extracting GSUB dependencies, find_or_create_glyph_set() turned
every Coverage (format 3 contexts) and class (format 2 contexts) that
a contextual rule mentions into an hb_set_t holding every glyph id the
table lists, whether the face has it or not, with no budget on the
expansion or on the second copy kept in the returned graph.  In
experimental builds, where the depend API and 24-bit Coverage formats
live together, a 6.5 KB font with 400 backtrack Coverage format 4
tables each covering [0, 0xFFFFFF] drives a build to ~2 GB.

Clip the cached sets to the face in find_or_create_glyph_set(), the
one place every context position goes through (the first input
Coverage and ReverseChainSingleSubst already clip this way), so the
copies kept in the graph shrink with them.  For real fonts this
changes nothing; only broken fonts stop carrying glyph ids past the
face.

Clipping alone is not enough: the temporary is still expanded in full,
and the glyph count itself is cheap to inflate with a MAXP table.  So
add a work budget for the whole build in the existing hb_budget_spend()
style: HB_BUDGET_DEPEND, budget_remaining on hb_depend_data_builder_t,
and charge_set(), which spends HB_BUDGET_4 per page while expanding a
temporary and HB_BUDGET_32 per page for every set the graph or context
keeps (new_set(), find_or_create_glyph_set(), find_or_create_recurse_set()).
Running out fails the build and hb_subset_depend_from_face_or_fail()
returns NULL, as it already does on allocation failure.  A
get_page_count() accessor is added to the bit-set types so the charge
does not reach into set internals; no public API change.

The 400-position font now peaks at ~11 MB instead of ~2 GB, the 64 KB
/ 4000-position font is refused, and hb-subset-depend output is
unchanged for real fonts.  Regression tests build the face with
hb_face_builder_create(): one asserts no glyph above the face count is
kept, the other that a 600-range face is refused.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-856K54QG.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:42:16 -06:00
Behdad EsfahbodandClaude Opus 5.5 3b517b36f5 [depend] Drop the edge probe budget and raise the edge cap
With edge keys hashed by fasthash32 over all fields, a font can no
longer steer its edges onto one slot, so probing is bounded by the
table's load factor; remove HB_DEPEND_MAX_EDGE_PROBES and its
per-probe charge.

HB_DEPEND_MAX_EDGES at 1<<22 was too close to real fonts: macOS's
DecoType Nastaleeq Urdu builds a graph of 1,198,826 edges, 3.5x below
it.  Raise the cap to 1<<25, 28x above that.  Other large fonts
measured: Noto Nastaliq 234,098; Hiragino Kaku Gothic 120,851; Gulzar
30,377; NotoSans-merged 29,968; NotoSansKR 7,797.

Reaching the raised cap takes 3.7 s and 1.7 GB, too much for a test,
so test_depend_edge_limit now only checks that a modest fan-out is
recorded in full.  test_depend_gsub_high_bit_tags still covers the
hash.

Tested: meson test -C build-exp (experimental API, 296/296 OK).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 16:38:50 -06:00
Behdad EsfahbodandClaude Opus 4.8 0f9c31a671 [depend] Hash edge keys with fasthash32 and bound the dedup probe walk
The dependency-graph edge dedup table in
hb_depend_data_builder_t::add_edge() was indexed by the low bits of
an FNV-style hash whose chain of XORs and odd-constant multiplies
only ever carries upwards, so bit j of the result depended only on
bits 0..j of each field.  A font controls feature tags (free 32-bit
values) and substitute glyph IDs, so it can pick values that differ
only in their high bits and pile every edge of a lookup onto one
slot.  The open-addressed probe loop then does about N^2/2 work and
each resize_edge_slots() repeats it, so a 15 KB font with 680k
colliding edges spent ~10 min of CPU while an ordinary font with the
same edges takes 0.1 s.

hb_depend_edge_key_t::hash() now runs fasthash32() over all identity
fields, so every bit of every field reaches the index; equality and
the probe sequence are unchanged, so real fonts produce the same
graph at about two probes per edge.  The now-unused
hb_depend_edge_t::hash() is removed.

A mixed hash alone is not enough: it is unkeyed, so colliding edges
can still be found offline (each one then costs font bytes instead
of coming free from a product of tags and pairs, but the walk stays
quadratic in them).  Every occupied slot stepped over in add_edge()
and resize_edge_slots() is therefore charged to a new per-build
counter edge_probes_remaining, seeded from HB_DEPEND_MAX_EDGE_PROBES
(1<<26, in hb-limits.hh and overridable like the other HB_*_MAX_*
limits, following the HB_RASTER_MAX_DRAW_EDGES precedent); exceeding
it fails the build the way an allocation failure does, so
hb_subset_depend_from_face_or_fail() returns NULL.  At about two
probes per edge the cap leaves more than 100x headroom over the
largest test font.  Honest fonts are already bounded well below it
by the HB_DEPEND_MAX_EDGES cap added in the previous commit; the
probe budget is what stops a crafted all-colliding chain from doing
quadratic work before that cap is reached.

A regression test builds a GSUB with 2048 features whose tags differ
only in bits 20 and up and checks that all 2048 edges are recorded.
test-ot-depend, the api suite and the depend fuzzing/closure-parity
tests pass on the ASan experimental build, and the finding's
reproducer now builds the hostile font in well under a second.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-4XW1PGV5.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:38:50 -06:00
Behdad EsfahbodandClaude Opus 4.8 d377bc81f2 [depend] Cap the number of stored dependency-graph edges
hb_depend_data_builder_t::add_edge() stored one record per
(source, target, feature tag) with no bound.  In GSUB a lookup
reached by F feature tags multiplies its N*M substitutions by F, so
a lookup whose covered glyphs all share one AlternateSet, selected
by many feature tags, produces N*M*F edges from a font that grows
only with N+M+F.  A 5 KB font yields 33M edges and ~1.4 GB of
builder memory; COLRv0 overlapping layer slices reach the same sink
without any GSUB.  The builder also kept walking after fail(), so an
allocation failure still cost the full N*M*F iterations.

This adds HB_DEPEND_MAX_EDGES (1<<22, about 4M, next to
HB_MAX_GRAPH_EDGE_COUNT in hb-limits.hh and overridable the same
way; 16x the largest graph in the test corpus) and enforces it in
add_edge(), the single choke point every table's extractor stores
through, so hb_subset_depend_from_face_or_fail() returns NULL
instead of building an unbounded graph.  add_edge() and
add_gsub_lookup() return early once the builder has failed, and
GSUB_accelerator_t::depend()'s lookup loop stops as well, matching
the existing !successful checks in the context walkers, so a failed
build does bounded work.

A regression test builds the fan-out font in memory with
hb_face_builder (no binary test font needed) and checks that a
16x16x4 fan-out is recorded in full and that the 256x256x512 one is
refused.  test-ot-depend and the api suite pass on the ASan
experimental build, and the finding's reproducer now returns NULL
in ~1 s.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-9IOHFFER.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:38:50 -06:00
Behdad EsfahbodandClaude Opus 4.8 7f65c5d820 [ot-map] Resolve a feature once in add_lookups() instead of once per 32 lookups (#6626)
hb_ot_map_builder_t::add_lookups() paged through a feature's lookup
indices 32 at a time with hb_ot_layout_feature_with_variations_get_lookups(),
which re-resolves the feature on every call.  Under a FeatureVariations
record that resolution is a linear search of the substitution records,
so a feature with many lookups and many variation records cost time
quadratic in the number of lookups, stalling hb_shape()'s plan build.

Add _hb_ot_layout_feature_with_variations_get_lookups(), which resolves
the feature once and returns its whole lookup-index array, and have
add_lookups() iterate that array directly.  The public API is
unchanged; the resulting map is identical.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-M3XZD80O.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:29:56 -06:00
Behdad EsfahbodandClaude Opus 4.8 26a6729680 [instancer] Merge repeated VarData region indices into one tuple (#6625)
create_from_item_var_data() deep-copied a region's axis hash map once
per entry of VarData.regionIndexes, so memory grew with
regionIndexCount x axisCount while the font pays only their sum, and
region indices may repeat freely. A 100 KB font (2000 axes, one region,
16000 copies of region index 0) made one hb-subset call allocate about
2 GB.

A region a VarData names again now adds its delta column to the tuple
it already owns, exactly what merge_tuple_variations() computes later,
so the instanced output is byte-identical. To also bound the sibling
case of many dataSets offsets each pointing at one VarData, the region
axes copied while converting one ItemVariationStore are summed across
its VarData subtables and capped at a new HB_VARSTORE_MAX_TUPLE_AXES
(1<<20); stores in the test fonts stay far below the
cap. Exceeding the cap fails instancing of that store, the same
path a failed allocation already takes.

Adds a unit test (test_repeated_region_index) covering the merge and
the cap.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-1CGIYCMY.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:24:19 -06:00
Behdad Esfahbod ca50d4ad91 Merge pull request #6624 from harfbuzz/split/gsubgpos-header
[layout] Write GSUB/GPOS subsets with a header that matches their layout
2026-11-10 16:21:48 -06:00
Behdad EsfahbodandClaude Opus 4.8 8aa3f2a887 [gpos] Skip non-increasing Coverage indices in PairPosFormat1::intersects fast path (#6623)
* [gpos] Skip non-increasing Coverage indices in PairPosFormat1::intersects fast path

The slow-population fast path of PairPosFormat1_3::intersects() walked
the glyph set and indexed pairSet by the Coverage index of each glyph.
In a well-formed font those indices increase with glyph id, but a
broken Coverage can repeat an index, so one PairSet could be scanned
once per glyph, giving quadratic time.

Track the last index used and skip any that does not increase (and any
NOT_COVERED), following the existing "Broken table. Skip." convention,
so each PairSet is scanned at most once.  Well-formed fonts are
unchanged.  Adds a fuzzer regression font.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-6GSU471N.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* [subset] Skip PairPosFormat2 classes outside the class matrix when collecting variation indices

collect_pair_variation_indices() added every class returned by the two
ClassDefs to class1_set / class2_set without bounding them by
class1Count / class2Count.  A ClassDef can map glyphs to classes far
outside the value matrix, so the collected sets could be enormous and
the subsequent nested loop quadratic, although apply() and the
serialization loops only ever touch classes inside the matrix.

Bound the collected classes by class1Count / class2Count, the same
bound the rest of the code already uses.  Output is byte-identical;
adds a fuzzer regression font.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-C60AYSA3.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:20:27 -06:00
Behdad EsfahbodandClaude Opus 4.8 7f1440eac1 [GDEF] Fill the mark-glyph-set cache straight from each Coverage (#6622)
The GDEF accelerator built its per-mark-set bit-page/digest cache by
materializing an hb_set_t for every mark glyph set and walking it glyph
by glyph.  A mark set's Coverage can name an enormous glyph range in a
few bytes, and several sets can alias one all-glyphs Coverage, so a
256 KiB font with 65535 such sets cost about 1 GiB and billions of
iterations in the accelerator constructor.

Fill each entry directly from its Coverage table via
collect_coverage() into lightweight sinks (gid extents, then a biased
bit-page or a digest), so the work is proportional to the coverage
records the sanitizer already budgeted for, not to the glyphs they
span.  Replace MarkGlyphSets::collect_coverage() with get_count() /
get_coverage() accessors.  Answers are identical.  Adds a fuzzer
regression font.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-OM3ZVHGK.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:15:17 -06:00
Behdad EsfahbodandClaude Opus 4.8 40d215957a [depend] Collect GSUB feature tags, substitutes and lookup variations in one pass (#6621)
GSUB_accelerator_t::depend() resolved each feature tag separately: it
ran hb_ot_layout_collect_features() once per tag and, for every tag,
walked the whole FeatureVariations table looking for substitute
features.  Tables with many features and many FeatureVariations records
were therefore quadratic, work that sanitize() bounds only for a single
pass.

Query all tags at once, then visit each feature once and each
FeatureVariations substitute once (via new visit_substitutes()
helpers), mapping every lookup to the tag of the feature it belongs to.
The emitted dependency graphs are unchanged.

Adds a regression test building a FeatureVariations GSUB whose features
have no direct lookups.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-5C2QXX5N.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:11:44 -06:00
Behdad EsfahbodandClaude Opus 4.8 c569b7fede [gsubgpos] Read contextual rule input glyphs through glyph_data_t in closure/depend (#6620)
context_closure_recurse_lookups() and context_depend_recurse_lookups()
already hold the rule's input sequence as a width-erased glyph_data_t
view, but for glyph-based rules (ContextFormat::SimpleContext) they
unwrapped the raw pointer and let intersected_glyph() index it as
HBUINT16.  That is wrong for the HBUINT24 arrays of
SequenceContextFormat4 / ChainedSequenceContextFormat4: every input
position after the first was read at the wrong bytes, so the nested
lookup was closed over the wrong glyph and its real output never
entered the closure (other glyph ids keep glyphs that can never
appear).  Only experimental / beyond-64k builds are affected.

Resolve the glyph via input[seqIndex - 1] for every format and make
intersected_glyph() a plain insert of the value, mirroring its sibling
intersects_glyph() which already takes the resolved glyph and ignores
data.  These two call sites were the only users of the
pointer-plus-index convention; 16-bit formats are unchanged, and there
is no API/ABI change.

Adds a regression test under HB_NO_BEYOND_64K: a SequenceContextFormat4
rule "1 2" invoking a SingleSubst 2->3 must close {1, 2} to {1, 2, 3}.
It fails on the unpatched library and passes with the fix.

Reported by Anthropic's OSS Scanner as ordinary bug ANT-2026-IU58QGNT.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:10:13 -06:00
Behdad EsfahbodandClaude Opus 4.8 f2390b0a2e [subset] Merge CFF charstring commands in place, not by erasing (#6619)
Passes 2 and 3 of CFF::specialize_commands() called
hb_vector_t::remove_ordered() once per merged command, and each call
shifts every command behind it, so a charstring of A mergeable line
segments followed by T unmergeable commands cost A*T element moves,
bounded only by HB_CFF_MAX_OPS: 32 s of CPU for one glyph of a 1.6 KB
font with --downgrade-cff2, paid again on every serializer-buffer
retry.

Replace both loops with one helper, combine_commands(), that walks
backwards exactly as before (each command absorbs its already-merged
successor chunk while the predicate allows, matching fontTools order)
but keeps finished commands packed at the tail and compacts the vector
once at the end, so the work is linear.  The two merge conditions
(operator pair, args-length parity, combined args < maxstack) become
the predicates verbatim and the merge order is unchanged.

Adds test_specialized_many_commands() to src/test-cff.cc (100,000 unit
rlineto followed by 100,000 empty hhcurveto must come back as 2,128
line commands plus the untouched hhcurveto commands); it takes 0.08 s
with this change versus tens of seconds before.

Discovered by Anthropic's OSS Scanner, as vulnerability ANT-2026-7TEMF5SE.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-11-10 16:08:49 -06:00
ebraminio e4802120b6 [subset] Keep hinting for fonts that require it (#6588)
* [subset] Keep hinting for fonts that require it

Some legacy fonts have hinting programs that must be run for correct
rendering; dropping them changes glyph advances and outlines.  FreeType
refuses to load these fonts without hinting, so do the same in the
subsetter by ignoring HB_SUBSET_FLAGS_NO_HINTING for them.

Only TrueType-outline fonts qualify.  Identify them by family name, and
by the length and checksum of their cvt, fpgm, and prep tables.  The
lists and the name-to-ASCII semantics are ported from FreeType's
src/truetype/ttobjs.c; fontations has the same logic in
read-fonts/src/tables/glyf/hint_reliant.rs.  Keep the detection in
hb-subset-hinting.hh rather than the subsetter, and add tests for both
paths.

Assisted-by: DeepSeek

* [subset] Refine hint-dependent font detection

- Store family names in a fixed-size table to avoid relocations.
- Read the name table directly instead of building the name
  accelerator, and select the record with FreeType's platform and
  language priority.
- Use hb_vector_t::set_storage() for the name buffer; treat an
  allocation failure as a non-match.
- Pick the family name the way FreeType does: honor the OS/2 WWS bit,
  and do not fall through to a lower-priority family name when the
  preferred one exists but does not match.
- Add equality operators and use CheckSum::set_for_data() for table
  checksums, rejecting unaligned data and handling the trailing bytes.
- Use face->table.glyf and drop the beyond-64K guard.
- Use ISALPHA/TOUPPER for the PDF randomization tag check.
- Add a regression test for the OS/2 WWS family selection.

Assisted-by: DeepSeek

* [perf] Add NO_HINTING subset benchmarks

Add a per-test NO_HINTING flag and enable it for the static TrueType
font benchmarks, so subsetting through the hint-dependent-font
detection path is measured. Benchmark names get a /nohinting suffix.

Assisted-by: DeepSeek

* [subset] Allow unaligned and unpadded checksum data

CalcTableChecksum() can add a trailing partial word.  Use that so
_table_checksum() can pass arbitrary table data directly to
CheckSum::set_for_data() instead of rejecting unaligned data and
splitting off the padded prefix.

Add unit coverage for padded, unpadded, and unaligned data.

Assisted-by: DeepSeek

* [subset] Rename and cache the hint-dependent face check

Prefix the entry point with _hb_subset, and cache the answer in face
user data so repeated subsetting of the same face does not repeat the
name and table scans.

Assisted-by: DeepSeek

* [subset] Avoid allocating for the hint-dependent cache

Store the hint_dependent_cache_t value in the face user-data pointer
instead of allocating a bool.  UNCACHED is zero, CACHED_FALSE is one,
and CACHED_TRUE is two.

Assisted-by: DeepSeek

* [test] Validate checksums against Python-generated vectors

Generate deterministic buffers and expected OpenType checksums with an
independent Python implementation, covering unaligned offsets and
lengths that are not multiples of four.  The generator is committed as
src/generate-checksum-vectors.py and rewrites the marked block in the
unit test.

Assisted-by: DeepSeek

* [test] Drop the checksum vector generator

Keep the Python-generated vectors embedded in test-checksum.cc, but
remove the generator script and its markers.  The comment records that
the values were generated independently with Python.

Assisted-by: DeepSeek

* [subset] Use CheckSum::CalcTableChecksum directly

Drop the trivial _table_checksum() wrapper and call
CheckSum::CalcTableChecksum() from the table-ID builder.

Assisted-by: DeepSeek

* [subset] Release the sanitized name table

hb_blob_ptr_t does not destroy its blob in its destructor, so add a
scope guard to release it on every return path.

Assisted-by: DeepSeek

* [subset] Limit cache enum scope

* [subset] Use dagger iteration for hint lists

Replace the family-name and font-ID membership loops with
hb_iter/hb_map/hb_any, and name the family-name capacity used by the
table and its iterator.

Assisted-by: DeepSeek
2026-11-10 11:26:23 -06:00
Behdad EsfahbodandClaude Opus 4.8 5aeb5ec60d [ft] Reselect the COLR palette instead of copying it per paint (#6617)
* [ft] Reselect the COLR palette instead of copying it per paint

#6490 lets paint callbacks run with the FT face unlocked, so a
callback can paint another glyph that selects a different palette on
the same FT_Face, changing the colors FreeType's palette buffer holds.
It guarded against that by heap-copying the whole palette on every
COLRv1 glyph paint.

Instead, remember on the hb_ft_font_t which palette hb-ft last
selected, and when a palette color is about to be read (with the lock
held) and another palette has been selected since, select ours again.
The common case costs one compare and no allocation.

Testing: meson test -C build passes.  test-paint's
ft_reentrant_paint, which paints with palette 1 from inside the
custom-palette callback, fails with the reselect disabled and passes
with it.

Follow-up to #6490

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* [ft] Drop palette bookkeeping; rely on FT_Palette_Select's no-op

FT_Palette_Select already skips its work when the requested palette is
the one already selected on the face, and returns the palette pointer
either way (freetype src/base/ftcolor.c).  So the selected_palette
member added to track it, and the guard around the re-select, are
unnecessary: _hb_ft_get_solid_color re-selects our palette
unconditionally after a callback may have changed it, and FreeType
no-ops the common case where it is already current.

Removes hb_ft_font_t::selected_palette and both of its assignments.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-10 11:22:15 -06:00