pdfcraft/vendor

Vendored crates

Permissively licensed crates we carry small patches for. Each directory keeps the upstream licence files unchanged. Every patch is marked PdfCraft patch: in the source, is covered by a PdfCraft test, and should be proposed upstream. Remove the vendored copy once upstream releases the fix. Vendoring copyleft code is never allowed (plan/adr/0001).

A PdfCraft patch in hayro-syntax::Page::base_dimensions expresses the existing one-point floor as a raw reciprocal of /UserUnit, rounded upward only when f32 multiplication would fall below one point. Five pdfcraft-render regressions with prefix raster::tests::raw_crop_ cover the observed dimensions, origins and rotations, units 41/47, bounded vector rasters and unchanged nearly-zero A4 fallback. Accepted page-unit policy and licences are unchanged.

In hayro-interpret, Type 3 glyph bounding boxes are transformed to device space before intersecting clip and shading bounds; raster::tests::shaded_type3_glyph_draws_at_multiple_positions covers translated glyphs (#3178).

Named form properties use the effective page dictionary when the form omits /Resources (#3574). hayro-syntax::Resources caches the page origin; hayro-interpret preserves the form dictionary's absence. An own dictionary stops fallback, including empty and malformed entries. Explicit null keeps its previous local-empty behavior. crates/render/tests/named_form_properties.rs covers text, Find, OCG identity, visibility overrides, page inheritance, and save/reopen. Other resource categories retain their existing parent-chain lookup.

Crate	Version	Licence	Patches	Test
hayro-interpret	0.7.0	Apache-2.0 OR MIT	(1) select /AP /N appearance states via /AS (checkboxes, radios, stateful icons) and honour the NoView flag; (2) Type3 Unicode fallback to /Encoding glyph names, then printable ASCII; (3) Type3Glyph::advance_width; (4) InterpreterSettings::ocg_overrides for viewer layer toggles; (5) soft masks whose group lacks /CS were dropped (Chrome gradient text); (6) paint nesting cap MAX_PAINT_NESTING for tiling patterns and Type 3 glyphs (self-referencing content overflowed the stack); (7) CID width ranges in /W and /W2 clamped to MAX_CID (a 0 4294967295 w range hung); (8) InterpreterSettings::hide_comments skips markup annotations but keeps widgets and links (View ▸ Hide all comments); (9) image_size_ok: images and masks over 2^28 pixels are not decoded (a fuzzed /W 4294967295 stencil mask allocated 4 GiB); (10) MAX_NESTED_PAINTS: nested paints (forms, Type 3 glyphs and tiling patterns inside one another) share a per-page budget of 50,000 (a glyph showing eight glyphs of its own font, sixteen deep, hung); (11) depends on skrifa 0.47 instead of 0.42 (upstream read-fonts fixed a Type 1 tokenizer that recursed through runs of numbers: a fuzzed font hung or overflowed the stack), with the one API change (Charset::glyph_id returns Option); (12) annotation appearances whose /BBox or /Matrix collapses to a line or a point are skipped (mapping them onto /Rect divided by zero and handed the device infinite or NaN matrices); (13) a Highlight annotation without a usable normal appearance is drawn from its /QuadPoints, /C and /CA with Multiply blending, as pdfcraft-annot draws highlights (it was not drawn at all); (14) simple-font /Widths (read_widths): entries only for codes 0–255 (one entry per code below /FirstChar was allocated: 4294967295 asked for 34 GB), an entry that isn't a number leaves its code missing instead of ending the array, and a /LastChar below /FirstChar gives no widths instead of discarding the font for a standard one; (15) the appearance box is mapped onto /Rect corner to corner: the translation now accounts for the scale (an appearance whose box didn't start at the origin and differed in size from /Rect was drawn shifted); (16) a clip set by W/W* takes effect after whichever path-painting operator ends the path (S, f, B, … as well as n; ISO 32000-2 §8.5.4): it was applied only on n, so re W* S clipped nothing and the clip stayed pending for a later, unrelated n; (17) InterpreterSettings::cancelled: once set, interpretation stops before the next content operator (a render pool sets it for every worker when dropped, after an edit or on closing, and for one worker when the watchdog gives up on its render, so renders whose answers nobody can receive stop instead of running to the end); (18) MAX_PAGE_CONTENT: the content a page holds at once (its contents, and the forms, tiling patterns and Type 3 glyph procedures being painted) stays within 512 MiB (decode_content; once it runs out, InterpreterWarning::ContentTruncated is reported (once per page by pdfcraft-render) so a truncated page is not a clean render; each level gives its bytes back when it finishes): forms nest fifty deep and each level kept its content while the next ran, so a form painting itself around a stream inflating to 256 MiB would have held over 12 GB; and Type 0 sampled functions take at most MAX_SAMPLE_ENTRIES (2^21) table points and MAX_SAMPLE_VALUES (2^23) samples, read no further than /Size asks (256 MiB of 1-bit samples asked for over 8 GB); (19) mesh shadings (types 4–7) are sampled only inside the area being drawn (ShadingPattern::encode_within, which the vendored hayro renderer calls with that area plus a pixel), within 8 visits per pixel of it plus 4 per triangle (TriangleSampler), and patches are triangulated one at a time: points far off the page were sampled over their whole bounding box, up to 65536 × 65536 entries (a fuzzed 147-byte patch stream; a synthetic page passed 6 GB in six seconds), ten thousand overlapping triangles ran for minutes, and every patch became 722 triangles before any limit; meshes hold at most MAX_MESH_PATCHES (2^16) patches and MAX_MESH_TRIANGLES (2^20) triangles, a lattice needs two to 2^19 vertices per row (/VerticesPerRow 0 pushed empty rows for ever; one huge row read the whole stream), and a page samples at most MAX_PAGE_MESH_TRIANGLES (2^25) mesh triangles in all (a mesh painted a thousand times was triangulated a thousand times); each cut logs a warning; (20) PDF function construction is limited to 64 levels and 10,000 total nodes with cycle detection; calculator parsing and evaluation share procedure depth/work budgets, integer division/bitshift/rotation/index are panic-free, and fixed-size stack errors propagate; (21) a Type 0 font drawn with a substitute that isn't CID-keyed (it isn't embedded, or its program doesn't load) draws .notdef for a CID of an Adobe character collection (Japan1, GB1, CNS1, Korea1) that it can't find by Unicode: hayro took the CID for a glyph index of the substitute, which numbers its glyphs its own way, so the character was drawn as an unrelated one (𠮷 as "Ｑ" in BIZ UDPGothic); (22) a Type 0 font drawn with a substitute finds a character given with a variation selector (/ToUnicode, or Adobe-Japan1-UCS2, which maps the JIS X 0208 forms of the kanji JIS X 0213:2004 redrew that way: CID 1247 is 噂 U+E0100) through the substitute's variation sequences, else as the character alone: such a CID found no glyph and was drawn as an unrelated one; (23) backport of hayro 0.8's gradient module: EncodedShadingPattern::as_svg_gradient turns an axial or radial shading into stops of an sRGB gradient (within 0.01 per channel, a stop at every stitching-function bound via Function::stitching_bounds) when a native gradient draws it exactly (extended past the box drawn in; a radial one with no background colour); (24) DecodedImageCache: decoded image pixels shared across renders through the interpreter cache within a 64 MiB budget (an image-heavy page decoded its thousands of images again per tile and per scale), keyed by every decode input (DecodedImageKey: the stream's place and dictionary bytes, its object, mask/stencil/raster output, interpolation, decode target size, preview or final); images under a transfer function and inline images are not cached, an image over the budget is never kept, and a poisoned lock decodes uncached; (25) Device::actual_text hands the device a marked content sequence's /ActualText (inline or named in /Properties), so text extraction keeps the Thai, Indic and Arabic clusters Chrome marks this way (their shaped tone marks and vowels have no Unicode of their own); (26) non-embedded simple fonts drawn with a standard-font stand-in build the Latin Extended-A letters the stand-in lacks (/Gbreve, /Idotaccent, /Scedilla, /aogonek, /ccaron…) from the base letter and accent it has (StandardKind::composite_glyph, ACCENTED_LETTERS): PDFium's Foxit stand-ins hold only the Standard/WinAnsi/Mac Roman glyphs, so iText letters with Times New Roman unembedded showed blanks for Ğ, İ, Ş, ğ and ş; the comma-below letters (/scommaaccent, /Tcommaaccent, /kcommaaccent…, found by character, so /uni0136 and the glyph list's /Kcedilla too) are the letter and the font's comma, smaller and under the baseline, since the stand-ins have no commaaccent (Romanian ș drew nothing and ț a cedilla in forms laid out from XFA; ģ, whose comma is turned and above, is not built); (27) an embedded TrueType font without a base encoding reads its codes as WinAnsi for text extraction (EmbeddedKind::code_to_unicode_name), matching pdftotext's output, instead of Standard: OCR text layers of scans write WinAnsi codes with no /Encoding, and Standard has no ü, ö or ß and reads é as Ø (glyph selection unchanged); (28) an image whose pixels hold at most 16 bits of samples (an 8-bit Separation or DeviceN image, a 1-, 2- or 4-bit gray or indexed one) is converted once per possible pixel value through a lookup table instead of once per pixel (rgb_via_lookup): a spot-colour tint transform is often a PostScript calculator program, run for every pixel, so a fillable character sheet's 3304 × 4105 DeviceN background took four seconds and 430 MB to draw a page; each entry is converted on its own (convert_each), so a tint transform giving the wrong number of components for a value the image never uses cannot shift the colours of the values it does; (29) Device::begin_annotations/end_annotations bracket the page's annotations, so a device can tell comment paints from document paints (reading colours recolour the document, never a comment, #1548); (30) the Annex D.2 alias codes read as the ink they draw: WinAnsiEncoding 0xA0 and 0xAD, and MacRomanEncoding 0xCA, carry the glyph names space and hyphen -- the names those tables also give 0x20 and 0x2D -- not nbspace and sfthyphen, whose glyph list values U+00A0 and U+00AD are invisible formatting characters. Extraction hid drawn ink behind them, so search and redaction matched neither (XiaoBiaoSong.pdf in the pdf.js corpus drew 104 soft hyphens; across the 989 files the soft hyphens we extract fall from 117 to 13 and the non-breaking spaces from 30 to 26, what remains coming from /ToUnicode maps, which say so explicitly). The base table only, in the text direction only: a /Differences name keeps its glyph list value and glyph lookup is untouched, both as pdftotext does; (31) a Type 0 sampled function keeps its samples in one flat array indexed by strides (Type0::entry) instead of a hash map keyed by each grid point: the map was built again, one hashed insert per sample, every time a shading using the function was painted (a 2,097,152-sample ramp painted 30 times took 38 s; bug1721218 in the pdf.js corpus spent most of a 1.5 s page on it); the interpolation is unchanged, so results are bit for bit the same, a sample missing from a short stream still gives no value, and the maximum sample value is computed without overflowing at 32 bits per sample; (32) the decoded-image cache (24) keys a decode by the size the image is drawn at only when a JPX stream decodes it or its soft mask or mask (decode_reads_target): JPX is the only filter that decodes a reduced resolution for a size, so a JPEG or Flate scan was decoded again at every zoom step and kept once per zoom level; one decode now serves every size; InterpreterCache::decoded_image_decodes counts the cache's decodes; (33) Coons and tensor-product patches copy their vertices' colours less: each point of a patch's 20 × 20 triangle grid is interpolated once (generate_patch_triangles; each cell interpolated its four corners itself, 1,444 interpolations a patch for 400 points), a cell's second triangle takes the vertices its first copied, and the mesh sampler moves a patch's triangles into device space instead of copying each vertex (TriangleSampler::sample_owned); the same values, 10–16% faster on mesh-gradient pages; (34) color::tint_transform_runs: how many colours the calling thread has run through a Separation or DeviceN tint transform (counted once per batch in convert_f32), so a test counts patch (28)'s once-per-sample-value runs instead of timing them (its 10 s deadline failed on a busy FreeBSD VM in CI); (35) EncodedShadingPattern::extent: where a mesh shading (types 4–7) can paint at all, the pixels its triangles touched when it was encoded (None for the types that paint everywhere, and when the shading has a /Background), so the renderer samples a texture there alone and skips a mesh with nothing on the canvas; (36) a mesh shading's pixel whose corner no triangle holds strictly (on an edge two triangles share, or on the mesh's side) takes the triangle that covers its centre, edges included (Triangle::covers): testing only the corner left a white seam along every shared edge, and the colour is still taken at the corner; (38) the /DefaultGray, /DefaultRGB and /DefaultCMYK colour spaces of a resource dictionary (ISO 32000-2 §8.6.5.6) replace the device space content selects: with g/G, rg/RG, k/K and cs/CS (by name, or through a /ColorSpace resource that is a device space), and as the /ColorSpace of an image XObject, an inline image or a shading painted by sh or as a shading pattern (DefaultColorSpaces, looked up once per content window in the resources and those they inherit, so a form XObject without a default of its own draws in its page's, as Poppler and MuPDF draw it); a default that is a Lab, Indexed, Pattern or device space, has another number of components or is no colour space is ignored and device colour is drawn as before; device spaces inside other spaces (an Indexed base, a Separation or DeviceN alternate, a Pattern's underlying space), shading dictionaries, masks and transparency group spaces are unchanged, and the decoded-image cache (24) keys a decode by the default it was drawn in. hayro drew device colour as it was, so pages that calibrate it this way (common in PDF/A) came out in the wrong colours; (39) knockout groups (#2971): Device::set_knockout right after a /K true group is pushed; Device::knockout_image for an unmasked image before its group is pushed, when Device::in_knockout_group: the device clears its placement rectangle only if it will draw the image (on the canvas, not a preview's tiny image, and it decodes), decoded at its draw size so the draw reuses that decode; and, when Device::knockout_erase_wanted, a nested group's content drawn once between begin_knockout_erase and end_knockout_erase before the group itself, which then always gets a layer of its own; (40) an optional content membership dictionary's visibility expression (/VE: /And, /Or, /Not, nested up to 32 deep) decides its visibility ahead of /OCGs and /P (ISO 32000-2 §8.11.2.2); it was ignored, so content under it stayed visible with its layers hidden; a malformed expression, including one naming an object the document doesn't list in /OCProperties /OCGs, falls back to /OCGs and /P (#2986); (41) a transparency group that is neither isolated nor knockout, entered at full opacity with no soft mask and Normal blending, is drawn straight onto the page instead of into a layer of its own (§11.4.7: its elements composite with the backdrop; the layer made a /BM /Multiply highlight appearance an opaque block, #1968); (42) text in a Type 3 font is drawn filled whatever the text rendering mode, except mode 3 (invisible): only that mode has any effect on Type 3 fonts (ISO 32000-1 §9.3.6). Stroking ran the glyph procedure a second time, so 2 Tr made bitmap fonts look bold, 1 Tr painted an uncoloured glyph in the stroking colour, and 7 Tr hid it; (43) Lab colour spaces honour their own /WhitePoint: L*a*b* goes to XYZ relative to that white, is Bradford-adapted to D65 and converted to sRGB (ISO 32000-2 §8.6.5.4) instead of through LAB.icc as if every white were D50 (a D65 blue came out 45 levels too red; the profile is removed); components normalised to [0, 1] (manual_scale, an /Indexed palette over Lab) map onto L* 0..100 and the a*/b* /Range; (44) a Type 0 font whose /ToUnicode is the name /Identity-H or /Identity-V (not a CMap stream; WSCAD writes it) reads each code as its Unicode value, as Acrobat shows it (Type0Font::identity_to_unicode); the identity CMap the name resolved to maps codes to CIDs only, so such pages had no text to select or find; (46) an image XObject's group applies the fill opacity, so the fill alpha is reset to 1 before a stencil mask takes the fill colour as its paint, and a pattern paint drops the opacity it captured at scn (Pattern::without_opacity) (it was applied twice: a mask at ca 0.5 was drawn at a quarter, #2975); (47) path_as_rect: a five-element path is a rectangle only when its four points are distinct, each step from one to the next runs along a single axis, and the last returns to the first. The length test admits one MoveTo and four LineTo as well as one MoveTo, three LineTo and a ClosePath, and asking only that every corner of the bounding box was visited says nothing about the order or the steps between, so a five-point polyline whose ends reached the corners was painted as that box -- a fourth side the path does not have -- and a clip of that shape was discarded; the coordinates are compared exactly, since the matrix is applied after this check and a cm of 10,000 turns is_nearly_equal's 256th of a unit into tens of device units (#3327); (48) an /ICCBased colour space is cached by what its profile is read from (IccProfileKey): the stream object it names, looked up without reading the stream, and then the stream's dictionary and its own bytes (decrypted, not decoded), so a byte-for-byte copy of a profile in another object is read once, and a profile stream written directly in its array, which reports the object holding the array as every other stream written there does, is told apart by its bytes. hayro keyed it by the stream dictionary's bytes alone, so two different profiles with the same /N and /Length shared whichever was read first, and colours in the second were drawn through the first. InterpreterCache::icc_profile_reads counts the profiles read; (49) OpenTypeFontBlob::new also accepts a post table version 1.0, 2.0 or 3.0 written with its 16-bit halves swapped (00 00 00 03), as ScanSoft PDF Create! embeds every font; such fonts were refused and their text drawn with a substitute (Hebrew came out garbled or missing, #2946); glyph_names reads such a table from a copy with the version written correctly, so a 2.0 table's names serve /Differences. Any other unknown version is still refused (pdf.js 9462); (50) CffFontBlob::glyph_index_by_name looks a glyph name up in a name to glyph map built on first use (shared between clones) instead of searching the whole charset, with a string lookup per glyph, for every character code of a font; the first glyph with a name still wins; (51) at most 4096 graphics states saved at once, holding at most 64 MiB together (each q copies the whole state, dash array and clip list included, so short operators multiplied one large state); a q past either limit is counted but not copied, and its Q still removes the clips added since (from storytold/craft-libs#239); (52) a /W2 run whose first CID is past MAX_CID stops there, as in /W (7), checked before the run's metrics are read: a first CID of 4294967295 inserted one entry and then overflowed, discarding the whole /W2 table (from storytold/craft-libs#254)	(1) raster::tests::widget_appearance_states; (2, 3) xtask text-oracle on corpus/pdfjs issue918; (4) raster::tests::layer_override_hides_content; (5) raster::tests::soft_mask_without_group_colour_space_is_applied; (6) raster::tests::self_referencing_tiling_pattern_terminates, raster::tests::self_referencing_type3_glyph_terminates; (7) raster::tests::huge_cid_width_ranges_terminate; (8) raster::tests::hiding_comments_keeps_fields; (9) raster::tests::absurd_mask_and_fax_sizes_are_refused_before_decoding; (10) raster::tests::type3_glyphs_that_fan_out_into_themselves_finish; (11) raster::tests::type1_font_programs_with_long_runs_of_numbers_load_quickly; (12) raster::tests::degenerate_annotation_appearances_are_skipped; (13) raster::tests::highlights_without_an_appearance_are_drawn; (14) raster::tests::malformed_simple_font_widths_keep_later_widths_and_the_font, raster::tests::huge_simple_font_first_char_terminates; (15) raster::tests::annotation_appearance_box_maps_onto_rect; (16) raster::tests::clipping_path_applies_after_any_painting_operator; (17) raster::tests::dropping_the_pool_cancels_its_renders, raster::tests::the_watchdog_stops_the_render_it_gives_up_on; (18) raster::tests::page_contents_and_nested_forms_share_one_budget, raster::tests::content_painted_again_and_again_keeps_drawing, raster::tests::sampled_functions_with_billions_of_samples_are_refused; (19) raster::tests::mesh_shadings_far_off_the_page_render, raster::tests::many_overlapping_mesh_triangles_finish, raster::tests::mesh_shadings_with_many_patches_or_no_rows_finish; (20) raster::tests::function_limits_calculator_arithmetic_and_depth, raster::tests::function_limits_public_construction_work_and_depth, raster::tests::function_limits_stitching_cycles_are_refused_and_shared_children_work, raster::tests::function_limits_invalid_calculator_keeps_the_page_renderable; (21) raster::tests::cids_a_substitute_lacks_are_not_drawn_as_other_glyphs; (22) raster::tests::variation_sequences_draw_their_kanji_in_a_substitute; (23) raster::tests::axial_and_radial_shadings_draw_as_native_gradients; (24) raster::tests::tiles_match_full_render, raster::tests::skipping_paths_off_the_canvas_changes_no_pixel (both decode through the cache), raster::tests::decoded_images_with_identical_dictionaries_keep_their_own_pixels, raster::tests::decoded_images_under_a_transfer_function_are_not_shared; (25) text::tests::actual_text_replaces_the_glyphs_it_marks_in_every_writing_direction, text::tests::actual_text_follows_the_pdf_rules_for_marked_content; (26) raster::tests::accented_letters_the_standard_stand_ins_lack_are_built_from_their_parts, raster::tests::comma_below_letters_in_standard_fonts_draw_a_comma_under_the_letter; (27) text::tests::truetype_without_an_encoding_reads_its_codes_as_win_ansi; (28) raster::tests::spot_colour_images_run_their_tint_transform_once_per_sample_value, raster::tests::images_converted_through_a_table_keep_every_samples_colour, raster::tests::a_malformed_tint_for_an_unused_sample_keeps_the_used_colours; (29) raster::tests::reading_colors_keep_highlight_annotations_readable; (30) text::tests::the_annex_d2_alias_codes_read_as_an_ordinary_hyphen_and_space; (31) raster::tests::shadings_with_large_sampled_functions_paint_quickly, raster::tests::sampled_functions_interpolate_their_grid_as_before; (32) raster::tests::decoded_images_serve_every_size_they_are_drawn_at; (33) raster::tests::patch_meshes_keep_their_corner_colours; (34) raster::tests::spot_colour_images_run_their_tint_transform_once_per_sample_value; (35) mesh_extent_tests::{a_mesh_paints_its_triangles_alone, a_mesh_keeps_the_graphics_state_alpha, a_mesh_off_the_canvas_paints_nothing} (cargo test -p hayro); (36) raster::tests::mesh_triangles_sharing_an_edge_leave_no_seam; (38) raster::tests::default_colour_spaces::device_rgb_operators_draw_in_default_rgb, raster::tests::default_colour_spaces::device_cmyk_operators_draw_in_default_cmyk, raster::tests::default_colour_spaces::device_gray_operators_draw_in_default_gray, raster::tests::default_colour_spaces::device_images_draw_in_the_default_colour_spaces, raster::tests::default_colour_spaces::device_shadings_draw_in_the_default_colour_spaces, raster::tests::default_colour_spaces::an_image_drawn_with_and_without_a_default_keeps_both_colours, raster::tests::default_colour_spaces::forms_inherit_the_default_colour_spaces_they_do_not_override, raster::tests::default_colour_spaces::unusable_defaults_leave_device_colour_alone; (39) raster::tests::knockout_groups_replace_what_their_elements_cover; (40) raster::tests::visibility_expressions_decide_what_a_membership_dictionary_shows; (41) raster::tests::a_multiply_highlight_inside_a_transparency_group_keeps_the_text; (42) raster::tests::text_rendering_modes_other_than_invisible_draw_type3_glyphs_as_filled; (43) raster::tests::lab_honours_the_space_white_point, raster::tests::indexed_palettes_over_lab_keep_their_colours
hayro	0.7.1	Apache-2.0 OR MIT	(1) RenderSettings::{x_offset, y_offset} to render one tile of a large page; (2) images over MAX_IMAGE_PIXELS (2^28) are skipped (a fuzzed /W 4294967295 hung in resampling); (3) tiling_cell_scale keeps a tiling pattern's cell pixmap within 3000 px a side whatever its /XStep//YStep (a fuzzed step far beyond /BBox allocated 2 GB); (4) stroke widths clamped to 4 × (canvas width + height) device pixels (a fuzzed /LW 9223372036854775807 allocated 10 GB in stroke expansion); (5) usable_dash_pattern: a stroke whose dash pattern could cut it into more than MAX_DASHES_PER_STROKE (250,000) pieces, counting kurbo's restart of the pattern at every subpath and ClosePath, is drawn solid (a fuzzed /D [[11] 0] on a line from x = 92234775807 allocated over 5 GB, [0.000001] 0 d on a 20 pt line over 10 GB); (6) a dash pattern with a negative or non-finite entry, or a period that isn't positive, is drawn solid ([-1 -1] 0 d looped for ever in kurbo); (7) render_size and render_into: render into a buffer the caller allocates (PdfCraft renders into pixel-aligned memory that the GUI takes over as texture data); render is render_into on a new pixmap; (8) checked image source/target/RGBA, alpha-premultiplication scratch and intermediate sizes before resampling, finite transforms and exact source/mask buffers; sources wider than u16 may shrink before backend limits, with a 2^20 source-axis cap and conservative 128 MiB filter-planning estimate; an over-limit or failed resize keeps the original image resolution and geometry when the source fits the backend, otherwise skips it, with a logged warning; staging, resampling output/scratch, RGBA conversion and padding buffers use fallible reserves; (9) may_paint_canvas: a fill or stroke whose control-point box, grown by how far its paint reaches past the path (half the stroke width times the larger of the miter limit and √2) plus 0.5 user units and 2 pixels, misses the canvas is not drawn, so a tile no longer strokes and fills every path of its page (item 5 of #623); set_skip_offscreen_paths turns this off for the regression test; (10) /Interpolate governs the pre-resize too (#624): a minified image that does not request interpolation is pre-resized with nearest-neighbour, planned without the Catmull-Rom filter limits (image_resampling_size takes the flag); a Type 3 glyph image is sampled bicubically only when interpolated; a minified image with a matching soft mask is nearest-sampled straight into premultiplied RGBA; (11) may_paint_rect: an image whose stream-size box, mapped to the device, misses the canvas is skipped before decoding (decoded data can only shrink against that box), so a tile of an image-heavy page no longer decodes, resamples and rasterizes every image on it; (12) preview renders (InterpreterSettings::preview: backdrops, thumbnails, grid cells, signature-drag backgrounds) skip images under PREVIEW_MIN_IMAGE_AREA device pixels, while final renders always draw them; (13) reading colours (#897): RenderSettings::color_replacement takes a ColorReplacement (text colour, background, only_black, line_art) that remaps solid text and line-art paint while rendering, glyph outlines counting as text and stencil-mask ink as line art; images, image alpha masks, soft masks and pattern cells keep their colours; render_pdf gains a color_replacement: Option<ColorReplacement> argument (pass None for true colours); the upstream-only [dev-dependencies.log] entry is removed from Cargo.toml (log is a normal dependency since the logging patch), so cargo test -p hayro runs this crate's unit tests; (14) a shrunk, interpolated, axis-aligned image is snapped to whole device pixels: shrunk once to its rounded device rectangle and drawn 1:1 at the rounded corner, instead of a ceil-sized raster resampled again at a fractional scale and offset (snap_minified_image; scans looked soft next to Acrobat); (15) a soft mask of another size than its image is filtered when shrunk, whatever its /Interpolate (#1649): such a mask carries the shape on its own (Word saves text as a tiny black image with a glyph-bitmap mask), and nearest-neighbour dropped every stroke thinner than the shrink factor; (16) under a soft mask or a blend mode, axial and radial shadings are sampled instead of drawn as native gradients (#2128): vello_cpu 0.0.8 copies an opaque gradient over fully covered tiles without the mask or blend mode, so PowerPoint's faded gradient fills came out solid; (17) the renderer keeps annotation paints out of the reading-colours mapping between begin_annotations and end_annotations (#1548); (18) the page renders as one transparency group; under reading colours its background is painted inside it as the initial backdrop, so a Multiply highlight blends with the dark background instead of staying an opaque block, and otherwise the background stays outside, so a blend mode over an empty part of the page keeps its colour instead of blending with the white paper (#2970); (19) set_paint samples a mesh shading's texture over its triangles alone (EncodedShadingPattern::extent, padded rather than repeated) and skips a mesh with none on the canvas: a BricsCAD plan's 310 one-triangle sh each covered the whole tile, 2 s per 1024 px tile at any zoom (the watchdog gave the page up on a busy machine), now about 70 ms; (20) a tiling pattern whose step is too large for a cell pixmap, where what is painted (a fill, or a stroke with its width) can only show the first cell (PowerPoint's /XStep 2147480100 for a picture fill drawn once), has its /BBox drawn at full scale with a transparent border and padded instead of tiled (shows_one_cell; a cell whose scale isn't finite and positive, from a /BBox too thin to divide by, keeps the old path: usable_cell_scale); shrinking the cell to fit the 3000 px limit of (3) left the picture under a pixel, so network diagrams lost their pictures (#3349); (22) knockout groups: while one is innermost, fills, strokes, glyph outlines and shadings are drawn with Compose::Copy (weighted by coverage), so each element replaces what earlier elements of the group drew instead of compositing over it; an image or nested group, drawn into a layer of its own, first has what is under its shape cleared (knockout_image for an unmasked image's rectangle, only when draw_image will draw it, decoded at the size it is drawn; a knockout erase pass that draws a nested group's content with every paint clearing instead of painting, a shading fill's clear clipped to the shading's /BBox as its paint is) and then composites over it (#2971); (23) backport of hayro 0.8: shadings that as_svg_gradient can represent are painted as a vello_cpu linear or two-point radial gradient instead of a texture sampled pixel by pixel through the shading function (InDesign fades are luminosity soft masks of one full-page axial sh: 40 on a report page took 15 s at 144 dpi, now under 1 s); without upstream's half-pixel shift, which suits vello_cpu 0.3 but moves 0.0.8's gradients off pixel centres	(1) raster::tests::tiles_match_full_render; (2) raster::tests::absurd_image_dimensions_are_skipped; (3) raster::tests::tiling_cells_stay_small_whatever_the_step; (4) raster::tests::absurd_predictor_columns_and_line_widths_render; (5) raster::tests::dash_patterns_that_cut_a_stroke_into_billions_of_pieces_render; (6) raster::tests::invalid_dash_patterns_terminate_and_are_drawn_solid; (7) raster::tests::pixels_are_a_plain_renders_bytes_in_a_buffer_a_gui_can_take_over; (8) raster::tests::image_resampling_rejects_anisotropic_target_growth, raster::tests::image_resampling_bounds_intermediate_before_planning, raster::tests::image_resampling_rejects_invalid_sources_and_scales, raster::tests::image_resampling_keeps_valid_sizes_and_exact_limits, raster::tests::resampling_fallback_preserves_original_geometry_and_pixels, raster::tests::image_resampling_wide_sources_shrink_before_backend_side_limits, raster::tests::image_resampling_wide_transparent_sources_still_shrink, raster::tests::image_resampling_padded_backend_dimensions_stay_checked, raster::tests::image_resampling_mismatched_alpha_mask_keeps_pixels_and_placement; (9) raster::tests::skipping_paths_off_the_canvas_changes_no_pixel; (10) raster::tests::non_interpolated_images_are_not_smoothed_when_minified, raster::tests::non_interpolated_gray_images_are_not_smoothed_when_minified, raster::tests::non_integer_minification_samples_the_same_columns, raster::tests::non_integer_minification_samples_colour_and_mask_from_the_same_columns, raster::tests::non_interpolated_alpha_images_keep_their_colours_and_mask_when_minified, raster::tests::non_interpolated_type3_glyph_images_are_not_smoothed_when_minified, raster::tests::image_resampling_nearest_accepts_wide_sources_the_filter_refuses, raster::tests::wide_non_interpolated_sources_draw_after_shrinking; (11) raster::tests::skipping_paths_off_the_canvas_changes_no_pixel; (12) raster::tests::preview_renders_skip_tiny_images_but_finals_draw_everything; (13) color_replacement_tests::{black_maps_to_text_colour, white_maps_to_background, midgrey_interpolates, alpha_copied_verbatim, only_black_threshold, line_art_false_skips_paths_but_not_text}, color_replacement_render_tests::{color_replacement_maps_text_and_line_art, color_replacement_maps_stencil_mask_ink}; (14) raster::tests::interpolated_minified_image_edges_snap_to_whole_pixels (cargo test -p hayro); (15) raster::tests::shrunk_soft_mask_larger_than_its_image_keeps_thin_strokes; (16) raster::tests::soft_mask_applies_to_a_shading_pattern_fill; (17), (18) raster::tests::reading_colors_keep_highlight_annotations_readable; (19) mesh_extent_tests::{a_mesh_paints_its_triangles_alone, a_mesh_keeps_the_graphics_state_alpha, a_mesh_off_the_canvas_paints_nothing} (cargo test -p hayro); (20) raster::tests::a_tiling_pattern_drawn_once_shows_its_cell, raster::tests::a_large_step_still_tiles_where_the_next_cell_shows, raster::tests::a_tiling_cell_too_thin_to_scale_keeps_the_old_path
hayro-jbig2	0.3.0	Apache-2.0 OR MIT	(1) at most 2^28 pixels per bitmap, page bitmap, halftone grid and grayscale image (each side could be 65535: 4.3 billion pixels decoded one by one); (2) at most 2^26 symbol pixels per symbol dictionary (a fuzzed dictionary decoded for minutes); (3) a symbol dictionary's height classes are capped at one per new symbol and its export runs at 2 × symbols + 1, each plus EXTRA_EMPTY_STEPS (1024): an empty height class or a zero-length export run doesn't advance, and past the end of its data the arithmetic decoder can decode them for ever (one changed byte hung the renderer); (4) a segment header's long-form referred-to segment count is capped at u16::MAX and a larger count is refused before its retention or referred-to bytes are read: the 29-bit count went straight to Vec::with_capacity, a reservation of about 2 GiB whose failure aborts the process (from storytold/craft-libs#176)	(1) raster::tests::jbig2_regions_of_billions_of_pixels_are_refused; (2) the nightly cargo xtask fuzz finding, which now finishes in about a second; (3) raster::tests::jbig2_symbol_dictionaries_that_never_advance_terminate; (4) segment::tests::test_referred_to_count_is_capped (cargo test --manifest-path vendor/hayro-jbig2/Cargo.toml)
lopdf	0.45.0	MIT	(1) PNG predictor rows longer than the data are refused before their buffers are allocated (a fuzzed /Columns 4294967295 allocated two 4 GiB rows), and empty frames return without row buffers; (2) predictor row widths use saturating arithmetic or reject overflow; (3) sub-byte TIFF Predictor 2 scratch is bounded by the samples present in the input instead of the declared /Colors, aligned with lopdf#589; (4) the object lookup keeps its cycle history through compressed objects, requires an ordinary in-file container, and caps the chain of indirect lookups at 16 (a chain of new objects, each a /Length compressed in a container whose own /Length names the next, or a stream whose /Length names the next stream, overflowed the stack: a dev build's 2 MiB thread within 800 hops); (5) object-stream members are parsed within their own extent (up to the next distinct offset), an entry whose offset repeats an earlier entry's is skipped (so a repeated object number can resolve to its earlier entry), and offsets are processed in index order; (6) find_latest_trailer: a rebuild takes a /Type /XRef stream dictionary as the trailer when the file has no trailer keyword (the latest in the file wins, as a later trailer would), so a concatenated or truncated file keeps its /Info and /ID instead of failing to load; a candidate whose /Root is the catalog beats one whose /Root is another dictionary, and an in-file /Root must match its generation (#2365)	(1) inspect::tests::png_predictor_rows_longer_than_the_data_are_refused_before_allocating, inspect::tests::empty_png_predictor_frames_do_not_allocate_rows; (2) inspect::tests::tiff_subbyte_predictor_row_width_overflow_is_refused; (3) inspect::tests::tiff_subbyte_predictor_scratch_is_bounded_by_available_samples, inspect::tests::tiff_subbyte_predictors_keep_components_rows_and_padding; (4) inspect::tests::compressed_object_self_container_does_not_recurse, inspect::tests::compressed_object_mutual_containers_do_not_recurse, inspect::tests::object_stream_length_cycle_does_not_recurse, inspect::tests::compressed_object_container_chain_does_not_recurse, inspect::tests::compressed_length_chain_does_not_recurse, inspect::tests::plain_stream_length_chain_does_not_recurse, inspect::tests::compressed_objects_and_indirect_stream_length_still_load, inspect::tests::compressed_length_in_a_container_with_an_indirect_length_still_loads; (5) inspect::tests::object_stream_member_extents_bound_owned_strings, inspect::tests::object_stream_member_extents_preserve_index_order_and_whitespace; (6) inspect::tests::rebuilt_file_keeps_the_info_of_its_cross_reference_stream, inspect::tests::rebuilt_file_prefers_a_trailer_whose_root_is_the_catalog
hayro-syntax	0.7.2	Apache-2.0 OR MIT	(1) Page-tree cycle guard in resolve_pages (visited set + depth cap): a /Kids loop among compressed objects overflowed the stack and aborted the process; (2) ccitt_size_ok: CCITT images wider than 2^20 columns or over 2^28 pixels are not decoded (a fuzzed /Columns 4294967295 allocated 4 GiB); (3) Flate/LZW predictor row widths are saturating usize, a zero width is refused and a row longer than the data is not allocated (a fuzzed /Columns 9223372036854775807 wrapped to a 2^61-byte allocation and aborted); (4) the inline-image end-of-data check (EI) doesn't nest inside its own look-ahead and examines at most 256 candidates and steps (50 images with "EI" in their data ran for over ten minutes); an EI followed by a delimiter (EI%..., EI[...) ends the image like one followed by white space; only a regular character after EI skips the candidate; (5) a stream whose /Length is wrong is recovered only when stream is the next token — scanning ahead to a later object's stream made an indirect appearance-state dictionary (Quartz checkboxes: /AP /N → << /Yes … /Off … >>) render as a non-form, so the widget was skipped; (6) JPX rescaling uses an exact integer sample range for 1–32-bit components (32 shifted past u32, and floating-point maxima at 25–31 bits wrapped to zero), and refuses zero-sized rows; (7) decode_limit: Flate (through flate2 and the fallback inflater), LZW, RunLength and ASCII85 never write past MAX_DECODED_STREAM (256 MiB), or, for the last filter of an image of known size, past twice its pixel bytes (a predictor tag byte per byte), at most 1 GiB; Stream::decoded_within for budgeted callers; a page's /Contents array shares one MAX_DECODED_STREAM (a 2 MB page whose content stream or 8 × 8 image inflated to 2 GB took the renderer past 3 GB in under a second; naming the stream twenty times in /Contents, 5 GB); (8) high-precision numbers (jsPDF MediaBox coordinates with over 19 digits) parse without overflowing the u64 mantissa: fractional digits past it are dropped, integer digits past it scale the value (a real, kept finite); (9) large dictionary indexes are built on first use and shared across repeated reads within each PDF (#307); the validated-byte cache is bounded, reader contexts remain separate, repairs and inline/content dictionaries bypass sharing, and small dictionaries stay eager; (10) decoded object-stream slots publish once under contention, cold loading rejects recursive dependencies, and xref repair serializes without holding recursive read guards; this permits one parser per native render pool with private worker caches; (11) sync.rs: a poisoned lock is recovered instead of unwrapped (a worker panic no longer cascades into every reader of the same PDF), and xref repair takes a blocking write lock instead of asserting a non-blocking one; (12) SegmentList (util.rs) returns None for an out-of-range index instead of panicking; (13) a page's /UserUnit scales render_dimensions and initial_transform, so the page and everything on it are sized in points, as Acrobat shows them; it is read as Acrobat reads it (the page's own entry, not inherited; a number from 1 up, at most 75,000; anything else counts as 1), and a page without it keeps exactly the transform it had; (14) revision 2–4 passwords are tried as given and then in PDFDocEncoding (legacy_passwords in crypto/mod.rs, which resolves the upstream TODO: Convert to PDFDocEncoding): passwords arrive as UTF-8 text, so a password with characters outside ASCII (Çok güzel, ılık€) written in PDFDocEncoding, as ISO 32000-2 §7.6.4.3.2 asks and qpdf and Acrobat write it, was always wrong; (15) page content is read a piece at a time (page::ContentReader, Page::content_reader): plain streams and Flate streams without a predictor are decoded as they are read, with a budget of 64 times their compressed size and never less than MAX_DECODED_STREAM (a stream that expands past it is cut and reported); other streams are decoded whole, as before, and share MAX_DECODED_STREAM. The content tokenizer reads a window at a time (UntypedIter::new_window, TypedIter::needs_more, TypedIter::committed): an instruction that runs past the window is not decided until more data follows, which Reader records by flagging reads and at_end past the end of the data, including through the clones the parser makes. The Flate decoder keeps its own 32 KiB history, so it pauses at any output size. ContentReader::new reuses raw/decrypted stream bytes for budget accounting and piece creation, and decoded_raw_within applies the existing Whole-stream filters and limits to those bytes without decrypting twice; (16) revision 5–6 (AES-256) passwords are tried in their SASLprep form (RFC 4013, then UTF-8, at most 127 bytes, as ISO 32000-2 §7.6.4.3.3 Algorithm 2.A asks and pdfcraft-crypt writes them) and then, if that differs, as given (aes256_passwords in crypto/mod.rs, which resolves the upstream TODO: Do the above.; stringprep, already in the tree through pdfcraft-crypt, is an optional dependency enabled by std, and without std passwords are used as given): the renderer used the password as typed, so one that SASLprep changes (a decomposed café, a no-break space, the fullwidth ａｂ a Japanese IME types in full-width mode) was wrong, although pdfcraft-cos accepted it: protecting a document with such an open password was rolled back ("the password is incorrect"), and a file with such a password did not open with it.; (17) CCITT /Rows 0 means the height is not predetermined (ISO 32000-2 Table 11), not zero rows: scanned Group 4 pages written that way rendered blank; (18) The streaming Flate decoder (lzw_flate::flate::fallback) copies a back-reference in runs of up to its distance (extend_from_within), not byte by byte. The copy reads only bytes already in the output, and a copy whose source does not exist still writes nothing, so the bytes are those a byte-by-byte copy writes. Inflating the A1 sample's content stream (278.6 MB) takes about 14% less time (0.66 s to 0.56 s); the output is identical; (19) CCITT /Rows below the image's /Height decodes to the height: a plotter driver wrote /Rows equal to /Columns, so only the first rows were decoded and the rest drawn black (#3308) (20) stream followed by spaces or tabs and then the line break (stream \r\n, as Microsoft Reporting Services writes it) starts the data after the line break; the content stream was taken to start at the space and the page rendered blank.; (21) a JP2 colour specification naming an enumerated colour space the decoder refuses (CIEJab in a veraPDF fixture, YCbCr, PhotoYCC, YPbPr, bilevel…) no longer fails the image: it is decoded again from a copy whose colr names the space the decoder assumes without one, greyscale, sRGB or CMYK by component count, as MuPDF and Poppler draw it (without_refused_enumerated_colour; the page drew blank, #3306); (22) an object stream's index of objnum offset pairs is parsed once per stream and kept beside its decoded data (ObjectStreamIndexes, XRef::with_object_stream_indexes), and an offset past usize is malformed instead of wrapping: XRef::get_with parsed the whole index again for every object it fetched from the stream, so reading all of a stream of N objects took N² steps (a Firefox page whose 65,542 objects share one stream took 37 s to draw and 40 s to extract its text); (23) the inline-image end-of-data look-ahead no longer rejects a candidate EI because a byte above 0x7F lies between it and the next one: when the candidate is the end, those bytes are the content stream that follows the image, and its strings carry such bytes (WinAnsi (Grüße) Tj, the two-byte codes of any Type 0 font), so the operator check that follows decides alone, bounded as in (4) (a scanned page whose OCR text held an accented word and a later capital "EI" — EINGANG, BEI, KLEIN — drew and extracted nothing after the image: 40 of 275 pages across five real documents); (24) xref::prev_offset: a trailer /Prev written as a reference N G R is read as the offset that object holds when the section's own entries place it (iText's indirectPrev.pdf; each section's entries are read before its Prev is followed), and otherwise as offset N (tiff2pdf; JHOVE's PDF-HUL-157 files, #2366), as pdfcraft-cos and other readers take it, so the earlier sections' pages are seen instead of one; (25) direct arrays and dictionaries share a reader-local nesting cap of 128, counted while skipping and while parsing a dictionary (its lenient repair reads, rather than skips, an object found where a key belongs, so << << … recursed through that read); deeper objects return None instead of overflowing the stack (including unused trailer values, trailer entries, content operands and marked-content properties); a chain of 65 direct stitching functions now stops at this cap, before the 64-function guard; (26) an xref stream subsection must be backed by its decoded data: all-zero /W widths (an entry reads no bytes) and a declared count longer than the data are refused before any entry is inserted, where a tiny stream with a large /Size or /Index count inserted that many entries (from storytold/craft-libs#239); a subsection running past object number i32::MAX is refused instead of overflowing or turning negative; (27) an xref stream's type field is read big-endian at its full /W[0] width, where only its first byte was read and wider fields shifted every later entry, and /W[0] over 8 bytes is refused like /W[1] (from storytold/craft-libs#241)	(1) cos/tests/writer.rs::page_tree_loops_in_object_streams_do_not_crash_readers; (2) raster::tests::absurd_mask_and_fax_sizes_are_refused_before_decoding; (3) raster::tests::absurd_predictor_columns_and_line_widths_render; (4) raster::tests::inline_images_with_ei_in_their_data_parse_in_linear_time, content::tests::inline_image_ei_may_be_followed_by_a_delimiter; (5) raster::tests::indirect_appearance_state_dict_is_not_another_stream; (6) raster::tests::jpx_stream_bit_depth_boundaries_do_not_panic, raster::tests::jpx_32_bit_image_does_not_prevent_other_page_content_from_rendering, raster::tests::jpx_gray_rgb_and_alpha_keep_eight_bit_samples, raster::tests::jpx_streams_reject_out_of_range_bit_depths; (7) raster::tests::decompression_bombs_stop_at_the_stream_limit, raster::tests::page_contents_and_nested_forms_share_one_budget; (8) inspect::tests::jspdf_high_precision_mediabox_does_not_overflow_dimensions, object::number::tests::high_precision_float_does_not_overflow_mantissa (hayro-syntax); (9) object::dict::lazy_index_tests, raster::tests::large_resource_indexes_preserve_pixels_and_text; (10) data::tests, xref::tests::concurrent_cold_object_stream_reads_all_succeed, raster::tests::shared_parser_is_lazy_and_initializes_once_under_contention; (11) data::tests::panicking_initializer_wakes_waiters_and_prevents_retries, xref::tests::concurrent_cold_object_stream_reads_all_succeed; (12) util::tests::segment_list_rejects_out_of_range_indices; (13) raster::tests::user_unit_scales_the_page_and_everything_on_it; (14) crypto::pdfcraft_password_tests, pdfcraft-engine tests::legacy_passwords_with_turkish_characters_open; (15) content::tests::windows_keep_every_instruction_at_every_split, content::tests::windows_keep_every_instruction_in_pieces_of_any_size, content::tests::windows_match_the_slice_on_random_content, content::tests::windows_drop_blanks_and_keep_a_comment_that_may_go_on, page::content_reader_tests::content_past_the_cap_is_read_to_the_end, page::content_reader_tests::a_stream_that_expands_past_the_budget_is_cut_there, page::content_reader_tests::a_small_expansion_under_the_cap_is_not_cut, raster::tests::page_content_past_the_old_cap_renders, raster::tests::content_reader_returns_the_same_bytes_in_any_pieces, raster::tests::content_bombs_stop_at_the_budget_and_say_so, page::content_reader_tests::plain_and_whole_pieces_keep_their_bytes, page::content_reader_tests::repeated_content_reference_counts_compressed_bytes_once, page::content_reader_tests::whole_cap_still_counts_later_unique_streams; (16) crypto::pdfcraft_password_tests::aes256_passwords_are_tried_in_their_saslprep_form_then_as_given, crypto::pdfcraft_password_tests::aes256_passwords_open_as_typed_and_in_their_saslprep_form, pdfcraft-engine tests::aes256_open_passwords_that_saslprep_changes_protect_and_reopen_as_typed, tests::aes256_files_whose_passwords_saslprep_changes_open_as_typed; (17) raster::tests::group4_fax_with_rows_zero_decodes_to_the_image_height; (18) filter::lzw_flate::tests::back_reference_longer_than_its_distance_repeats_its_source, back_reference_cut_by_the_limit_resumes_in_the_next_call and back_reference_reads_bytes_written_by_an_earlier_call; each fails when the copy returns the wrong remaining count; (19) raster::tests::group4_fax_with_rows_below_its_height_decodes_to_the_image_height; (20) raster::tests::space_before_the_line_break_after_stream_still_renders; (21) raster::tests::a_jp2_image_in_a_refused_enumerated_colour_space_is_drawn, filter::jpx::tests::refused_enumerated_colour_spaces_become_the_component_counts_default, filter::jpx::tests::what_the_decoder_handles_or_cant_be_read_is_left_alone (cargo test --manifest-path vendor/hayro-syntax/Cargo.toml); (22) raster::tests::objects_in_one_large_object_stream_are_read_in_linear_time; (23) text::tests::text_after_an_inline_image_survives_a_non_ascii_byte_before_a_later_ei, content::tests::an_inline_image_ends_at_its_own_ei_when_later_text_holds_a_non_ascii_byte (hayro-syntax); (24) xref::tests::prev_as_reference_reaches_the_earlier_section, xref::tests::prev_as_reference_to_an_object_holding_the_offset_reads_that_offset, pdfcraft-cos document::tests::a_prev_written_as_a_reference_still_reaches_the_earlier_sections, document::tests::a_prev_reference_to_an_object_holding_the_offset_reads_that_offset; (25) raster::tests::deeply_nested_direct_objects_do_not_overflow_the_stack, raster::tests::dictionaries_nested_without_keys_in_content_do_not_overflow_the_stack, raster::tests::dictionaries_nested_without_keys_in_the_trailer_do_not_overflow_the_stack, raster::tests::dictionaries_nested_without_keys_as_marked_content_properties_do_not_overflow_the_stack, raster::tests::dictionaries_nested_without_keys_are_read_up_to_the_nesting_cap, raster::tests::nesting_at_the_cap_fits_a_512_kib_stack (native only), raster::tests::function_limits_depth_guard_refuses_a_chain_of_references; (26) xref::tests::xref_stream_entries_need_backing_bytes; (27) xref::tests::xref_stream_type_field_uses_the_full_w0_width, xref::tests::xref_stream_loader_rejects_oversized_w0 (cargo test --manifest-path vendor/hayro-syntax/Cargo.toml)
winit	0.30.13	Apache-2.0	(1) on Windows 11 (build 22000 and later), WM_DPICHANGED applies the rect Windows suggests. It used to nudge the window onto the monitor MonitorFromWindow names, which during a drag is still the monitor being left; the nudge sent another WM_DPICHANGED, so a window dragged to a monitor with another scale factor grew and stayed on the wrong monitor at the wrong scale (#324). Backport of winit master 488c036a (winit#4341), which no 0.30 release has; Windows 10 keeps the old adjustment, as on winit master. (2) on Wayland, a seat's text input is destroyed when the seat goes away, not whenever it loses a capability: losing any capability (a touch screen or tablet unplugged, a remote-desktop session ending) destroyed it, and only an added capability recreated it, so IME stopped working until restart. Port of winit#4747; (3) on Windows, an IME composition reports its cursor from GCS_CURSORPOS only when WM_IME_COMPOSITION says it is set, and at the end of the composition otherwise: the Microsoft Korean IME never sets it, so the cursor read as 0 (egui then drew no caret while composing; it now draws it after the syllable). Port of winit#4746; (4) on Windows, a hidden window is not maximized until it is shown, and a hidden window shown while maximized gets one ShowWindow(SW_MAXIMIZE): SW_MAXIMIZE shows a hidden window, so PdfCraft, which eframe keeps hidden until its first frame is drawn, appeared unpainted (white) when it restored a maximized window, and could lose its focus. Port of winit#4587 (the later, focus-safe form of winit#4575); (5) on X11, when allowing or disallowing IME replaces the focused window's input context (egui does it each time a text field gains or loses focus), the new context is focused: no focus event follows, so it stayed unfocused until the window lost and regained focus. Port of winit#4727; (6) on macOS, insertText commits through IME when a composition was started (pending_commit, set by setMarkedText) rather than when marked text is left: input methods that end a composition with unmarkText before insertText lost the text, and Apple Korean dropped the key typed after a syllable (한 then 5 or ;); other insertText calls stay KeyboardInput; a commit outside a key press (a candidate clicked with the mouse, dictation) clears the marked text and leaves no key to swallow, as in winit#4748 (without it the next key, such as Backspace, was lost). validAttributesForMarkedText lists the underline and marked-clause attributes, without which some input methods don't compose. Port of winit#4650, the 0.30 backport of winit#4651 (merged on winit master), with the underline attribute spelled as AppKit's NSUnderlineStyleAttributeName (NSUnderline); (7) on macOS, disallowing IME with a composition in progress (focus leaving a text field) tells the input method to discard its marked text (discardMarkedText): it kept composing it once IME was allowed again, so the abandoned syllable came back (하, then 가 typed in the next field, became 학, then 하 + 가; Japanese か then さ became 傘). Port of winit#4745; (8) on macOS, text inserted outside a key press without a composition (the emoji picker) is reported as Ime::Commit; it was dropped. Port of the emoji-picker part of winit#4749 (0.30 backport of winit#4748); its selectedRange change for dictation is left out because it switches on press-and-hold for every held key (an accent menu instead of key repeat, and the chosen accent added after the repeated letters); (9) on macOS, a key whose text was committed through IME isn't also sent as KeyboardInput unless AppKit forwards it as a command (Enter): Apple Korean answers Space with no syllable in progress with setMarkedText(" "), insertText(" "), setMarkedText(""), and the last call reset winit's state, so the space arrived twice. Same goal as winit#4478, whose changes don't reach this call order (macOS 14); to be proposed upstream there; (10) on macOS, when the first key in a fresh app reaches Apple Korean's insertText as a raw compatibility jamo instead of starting a composition, that key is interpreted once more, synchronously, before it is sent as KeyboardInput (g k s typed ㅎㅏㄴ instead of 한); only an unmodified, non-repeated first key whose one callback inserts that same jamo, with IME enabled and nothing composed, focus and input source unchanged, is retried (ime_startup.rs). Port of winit#4744, with its policy tests.; (11) on X11 without Xft.dpi, the scale factor comes from the X screen's configured DPI (its size in pixels and millimetres) before the output's physical density from RandR: Xorg can deliberately run a dense panel at 96 DPI, and only winit applications then looked zoomed; Xft.dpi still wins, WINIT_X11_SCALE_FACTOR still overrides, and a screen reporting no size keeps the RandR fallback. PdfCraft patch (#1150, replacing the launcher pin proposed in #908). (12) on macOS, WinitView answers mouseDownCanMoveWindow with the window's movable_by_window_background attribute (NO unless asked): NSView answers YES for a view that isn't opaque, so with the transparent title bar and full-size content view PdfCraft uses, AppKit moved the window for any drag that began in the 28 pt title-bar band before the app saw it, and document tabs (which live in that band) could not be dragged to reorder them. The empty strip still moves the window through ViewportCommand::StartDrag. PdfCraft patch. The published crate's asset files (docs/res/*.svg, examples/data/*.png) are left out, so its window example doesn't build; (13) on Wayland, pixel-only wheel, wheel-tilt and continuous-source scroll events report Moved rather than starting an unterminated finger gesture, so egui follows modifier release instead of keeping Ctrl zoom latched (#991). Finger gestures and unknown sources retain their existing phase policy	(1) pdfcraft tests::winit_carries_the_windows_11_monitor_scale_fix keeps the patch in the build; the drag itself needs two monitors with different scale factors on Windows 11; (2) pdfcraft tests/winit_wayland_text_input.rs; the capability change itself needs a Wayland session and an input method; (3) pdfcraft tests/winit_windows_ime_cursor.rs; composing itself needs the Microsoft Korean IME; (4) pdfcraft tests/winit_windows_hidden_maximized.rs; the start-up itself needs Windows and a window that was closed maximized; (5) pdfcraft tests/winit_x11_ime_focus.rs; switching fields itself needs an X11 session and an input method (IBus, fcitx5); (6) pdfcraft tests/winit_macos_ime_commit.rs; composing itself needs macOS and an input method; (7) pdfcraft tests/winit_macos_ime_discard.rs; composing itself needs macOS and an input method; (8) pdfcraft tests/winit_macos_emoji_picker.rs; the picker itself needs macOS; (9) pdfcraft tests/winit_macos_korean_space.rs; composing itself needs macOS and Apple Korean; (10) pdfcraft tests/winit_macos_korean_first_key.rs and ime_startup.rs's own tests (run them with rustc --test); typing itself needs macOS and Apple Korean; (11) pdfcraft tests/winit_x11_dpi.rs keeps the patch in the build, and util::randr tests the policy (run them in the vendored crate); the scale itself needs an X11 session; (12) pdfcraft tests/winit_macos_title_bar_drag.rs keeps the patch in the build; the drag itself needs macOS and a pointer; (13) pdfcraft tests/winit_wayland_scroll.rs compiles the production phase policy and exercises modifier release through egui; live COSMIC confirmation remains open
egui-wgpu	0.36.2	MIT OR Apache-2.0	(1) winit::surface_fit: a window's surface is configured within the device's max_texture_dimension_2d, both sides shrunk by one factor that egui then draws at (pixels_per_point × render_scale), so the whole window is drawn into the surface (on DX12 it is stretched over the window and lines up with the pointer; wgpu's GL backend copies it unscaled into a corner); the longer side becomes exactly the limit; the depth and MSAA textures follow, and a recreated surface starts again from the window's own size. egui-wgpu configured the window's size as is, and Surface::configure panics beyond the limit: a window restored at another monitor's scale (opening an Outlook attachment on a mixed-DPI laptop asked an 8192-pixel device for 8600 × 3423, #577) or stretched across monitors closed PdfCraft, before its window appeared when it happened at start-up (emilk/egui#8361). The published crate leaves out egui's licence files; LICENSE-APACHE and LICENSE-MIT are from the 0.36.2 tag; (2) winit::catch_errors: a new window's surface is configured inside wgpu error scopes (validation, out of memory, internal), and a failure is returned as WgpuError::ConfigureSurface (for the root window at start-up eframe returns it as Error::Wgpu, which PdfCraft retries with OpenGL); nothing is kept for the window. wgpu's default handler panicked instead, and eframe configures the root surface before the app is created, so a driver that can't configure one closed PdfCraft at launch (a 2015 Intel GPU through wgpu's GL backend timed out waiting for the GPU to go idle, GpuWaitTimeout, #519). Unchanged, so still a panic: an error in a later reconfigure (resize, a lost surface), a device lost during the first configure (wgpu reports it only to a callback), and a window shown at zero size (first configured on resize)	(1) pdfcraft tests::a_window_larger_than_the_gpu_limit_gets_a_surface_within_it, tests::surface_fit_never_empties_or_overflows_a_surface, and tests::egui_wgpu_carries_the_surface_size_fit, which keeps the patch in the build and in the paths that size a surface; (2) pdfcraft tests::a_gpu_error_while_a_window_is_set_up_is_returned_not_a_panic (a real device: an error inside catch_errors is returned, not a panic) and tests::egui_wgpu_returns_a_surface_it_cannot_configure_as_an_error, which keeps it around add_surface's configure
vello_common	0.0.8	Apache-2.0 OR MIT	(1) clip::intersect (the clip step every clipped path passes through) starts its row walk at the first row both paths have strips in, not the earlier of the two first rows. The walk visited every row from the top of the clip for each path (on the A1 sample's full-page canvas, about half its strip rows per stroke), and rows where one path has no strips produce no output, so the strips are unchanged (vello_common 0.1 starts at the later row too). Its cost grew with the page, and a full-page render of the A1 sample was about 30% faster; the output is byte-identical	(1) clip::tests::intersect_walks_only_rows_both_paths_have_strips counts the rows the walk visits (cargo test --manifest-path vendor/vello_common/Cargo.toml) and fails without the patch; the render crate's pixel tests run through the patched crate
hayro-cmap	0.1.0	Apache-2.0 OR MIT	(1) malformed parts of a CMap (bad bfchar/bfrange entries, unresolvable usecmap, unparseable tail) are skipped instead of discarding the whole map, so search and redaction find text the edit layer can read (#3177); (2) a binary CMap's usecmap counts against the 16-level nesting limit as a text one does, where it passed its own depth and a self-referencing binary CMap recursed until the stack overflowed (from storytold/craft-libs#254)	(1) hayro-cmap::tests::bad_bfchar_entry_is_skipped, unresolvable_usecmap_is_skipped; (2) tests::self_referencing_binary_usecmap_stops_at_the_nesting_limit

Renderer startup errors: egui-wgpu::RenderState::create also wraps Renderer::new in winit::catch_errors when the winit feature is enabled. A rejected shader or render pipeline is returned as WgpuError::CreateRenderer, allowing PdfCraft's existing startup OpenGL retry. The regression pdfcraft tests::pipeline_creation_error_is_returned_and_device_remains_usable uses an invalid three-sample pipeline, then valid and invalid pipelines again on the same device. Later rendering errors, device loss, and non-winit renderer construction are unchanged.

Stencil masks filled with a pattern (#2975): hayro's renderer paints a pattern-filled stencil mask with the fill (non-stroking) paint. It passed is_stroke = true, so an uncoloured tiling pattern was drawn in its stroke colour (black by default) instead of the colour given with scn. Covered by raster::tests::stencil_masks_are_drawn_at_the_fill_opacity_once (the uncoloured tiling pattern case).

Temporary: hayro is the bootstrap renderer and lopdf the bootstrap inspector (ADR-0004); M2.6 replaces them with our own devices and cos/model.

TIFF predictor decoding: hayro-syntax aligns the output writer to a byte boundary after each row before splitting off its buffer. Packed RGB samples at 1, 2 or 4 bits per component can leave padding bits; without alignment later rows overwrite the preceding row or decode empty. Covered by raster::tests::tiff_predictor_packed_rgb_rows_keep_their_byte_boundaries (including horizontal differencing and an 8-bit control) and raster::tests::tiff_predictor_packed_rgb_image_renders_every_row.

JPEG 2000 palettes (#3179): hayro-syntax::filter::jpx::decode applies a JP2 palette (pclr) when the image dictionary has no /ColorSpace, since the colour then comes from the JP2 header (ISO 32000-2 §7.4.9). With a /ColorSpace the raw indices are kept, as an /Indexed space needs. A /ColorSpace that does not resolve counts as none: the JP2 colours are then the only ones. Without the patch a palette image with no /ColorSpace was drawn as its indices, in gray. The resolved colours are 8-bit samples whatever the precision of the indices (a 1-bit index no longer rescales its colours to 0 or 255 per channel); an image without a palette keeps its own. Covered by filter::jpx::tests::jpx_palette_is_applied_without_a_color_space, filter::jpx::tests::jpx_palette_indices_stay_raw_for_an_indexed_color_space, filter::jpx::tests::palette_colours_are_eight_bit_and_other_samples_keep_their_precision and raster::tests::a_jp2_palette_keeps_its_colours_whatever_the_index_precision.

Glyph names outside the Adobe Glyph List (#3169, #3170): hayro-interpret's glyph_name_to_text maps a name by the glyph list specification's full mapping (drop from the first ., split on _, each part through the list, uniXXXX… or uXXXX; a uni part with any group that isn't a character maps to nothing), so f_i, f_l and hyphen.case read as text. A name that only numbers the glyph (C39, c251) still reads as nothing: the code's StandardEncoding character would be a guess, wrong wherever the font's /Encoding puts the glyph elsewhere (review on #3239). Covered by text::tests::glyph_names_outside_the_glyph_list_still_read_as_text.

Text extraction: hayro-interpret exposes OutlineGlyph::is_vertical from a Type0 font’s CMap writing mode. PdfCraft follows negative glyph-space y for WMode 1 so upright vertical columns can be searched (#126).

Lab Bradford follow-up to #1237: Lab adapts XYZ to D65 with white-point ratios computed in Bradford LMS coordinates. CalRGB's existing flat-white and black-point normalization is unchanged. lab_bradford_tests checks independent D50/equal-energy neutral and chromatic oracles, D65 identity, and direct/Indexed consistency (including an asymmetric chromatic Range), and zero/overflowing cone-response fallback.

Clipping-only text reporting (#3109 / #3116)

hayro-interpret/src/interpret/text.rs reports mode 7 glyphs after installing their clip, and hayro-interpret/src/types.rs distinguishes GlyphDrawMode::Clip from invisible OCR text. hayro/src/renderer.rs does not paint either mode directly. This preserves clipping-only titles when TextFilter.images excludes invisible OCR. Coverage: text::tests::clipping_mode_filter_keeps_titles_and_excludes_ocr (plain and ActualText filtered extraction), and text::tests::clipping_mode_text_is_extracted_and_searchable_like_other_rendering_modes (extraction, search, glyph geometry, clip-through pixels and no-paint pixels).

Missing spacing glyphs (#3601): in hayro-interpret Type0Font::map_via_unicode, an OpenType substitute uses its ordinary-space glyph only when its own charmap lacks a spacing character (including Adobe-Japan1 U+2002). Native glyphs and variation forms retain priority; embedded fonts, PDF advances and Unicode decoding are unchanged. Renderer regressions japanese_substitute_missing_spaces_draw_no_ink and japanese_substitute_spaces_keep_pdf_widths_and_unicode use the synthetic MS-Mincho / UniJIS-UCS2-HW-H fixture with craft-fonts, including /W and /DW placement controls.