A page cropped to the art is recognised in a UserUnit page's units too,
when the art's box is the page in those units and the artboard isn't the
page in points. Such pages are taken only when every page is the art's
box, and only the first page's type is read: the later pages are copies
whose type was moved by the first page's offset and could fill the
layers' text objects twice.
prune_unseen ran before the retry that moves placed files to their box's
corner, and that retry cloned the pruned layers: text an image covered
only at its first placement was already gone, so the file came in without
it or as its PDF part. Clone the layers before pruning and prune each
placement on its own.
Character attribute 12 of the editing data's text document is the case
(1 Small Caps, 2 All Caps). The reader skipped it, so type set in All Caps
came in as typed, in lower or mixed case where the page draws capitals.
Map it onto CharStyle::all_caps and CharStyle::small_caps.
Patch and triangle mesh shadings of more than 4,096 patches, and strokes
painted with a mesh shading, came in as a flat grey. Illustrator writes
meshes of up to 14,355 patches, and writes a freeform gradient on a
stroke as a stroke painted with a type 7 shading.
- MAX_PATCHES is 16,384, as in the .ai editing-data reader.
- A stroke painted with a mesh shading is the meshes clipped to the area
the stroke paints (stroke_outline: width, caps, joins, miter limit and
dashes, built in the path's space and then transformed); draw_meshes
always clips a stroke's meshes.
Two follow-ups to the empty-page fallback:
- A page counts as empty only when nothing on it was skipped: its only
notes are the page's own (art off the artboards, art past the page). A
page whose only content couldn't be read, such as an image that can't
be decoded, keeps its notes and goes through the usual check against
the editing data, instead of being replaced by it unchecked.
- With textAs: outlines, the document read from the editing data alone
has its type outlined. The usual path takes outlines from the page,
but an empty page (or a text-only placeholder, #1387) has none to give.
A mesh in the editing data that the reader couldn't read made the whole
file come in as its page or PDF part, without its layers. Three cases
were rejected that files write routinely:
- A mesh starts grey, and `/CS` is written only where the colour space
changes, before the corner it applies to, even within a patch. Meshes
with no `/CS`, or one that changed between corners, were unread.
- Named colours (`[c m y k /DeviceCMYK (name) /Process]` or `/Spot`) and
`/NChannel` inks: each corner gives the tint, 1 being the full colour.
- Grids over 64 patches a side or 4,096 patches (files have 97 x 72).
Each corner now takes the colour space last set before it; named
colours paint at their tint and inks overprint at theirs; the limits
are 128 a side and 16,384 patches.
A placed (linked) file is put where its matrix takes the canvas point of
art space's origin (the centre of %AI3_TemplateBox). In many files the
matrix takes the corner of the box the file is written with instead, so
the image landed far off its artboard, the layers differed from the
page, and the file came in as its PDF part.
The reader keeps that second placement for each placed file where the
two differ. When the layers don't draw as the page, they are compared
again with those files at their box's corner, and used if that draws
as the page. Files whose layers already match keep their placement.
A placed (linked) file written as an object's editable copy, on %_
lines after its drawn look, has its %AI5_BeginPlace and %AI5_EndPlace
markers on those lines too. The lexer dropped comments after %_ that
don't start a section, so the placed file never began, the ~ that ends
it was an operator the reader didn't know, and the whole file came in
as its PDF part. The lexer now returns those two markers from %_ lines,
and the preview reader skips the %_ its header line starts with.
PDF export (and the PDF-compatible .ai it writes) cut each gradient mesh
patch into 8 x 8 flat-coloured quads. Viewers showed colour bands and
hairline seams between them, and files with many meshes became very
large and slow to write.
An opaque mesh is now written as one Type 6 (Coons patch) shading
(ISO 32000-1, 8.7.4.5.7) filling its outline, through the PDF writer's
MeshGradient: each patch blends its corner colours bilinearly, as the
canvas draws it. Colours go through page_color as other fills do, and a
mesh mixing RGB with other models goes through RGB. A mesh with
see-through points, an invalid one, or one past the coordinate range
keeps the quads.
A PDF-compatible .ai whose PDF part's pages are cropped to its art (each
page the box of all the art, the editing data's bounding box) rather
than to its artboards came in as that PDF part: the editing data's
artboard wasn't the page's size, so its layers were refused. With
several artboards every page held all the art, so the document had a
copy of everything per page.
When the artboard isn't the page's size but the editing data's bounding
box is, the layers are compared with the page on that box and keep
their artboards. The comparison of their art with the page still
decides whether they are used.
A pattern fill was written to EMF as every tile's art, each tile in a clip
of its own, with the brush made again for each part: a rotated two-colour
stripe fill of a 360 x 260 pt rectangle took 473 fills, 246 clips and
122 KB.
When the tiles can't overlap (a grid or brick tiling whose art stays in its
tile), each tile's clip is left out where what it clips lies inside it
already (a convex clip holding every control point of the fills), and the
n-th fill of every tile becomes one path. The picture is the same: the
tiles don't overlap, and the fill's own clip still trims them. That fill
is now 2 fills and 1 clip, 61 KB; unrotated, 99 KB becomes 33 KB.
Tiles that overlap, hex tilings, and tiles holding strokes, images or
clips that cut are written as before, their parts off the page left out.
A layered .ai or Illustrator EPS whose editing data (embedded rasters,
mostly) unpacks past 512 MiB opened as its PDF part or printed page, in
one layer, with "it decompresses to more than 512 MB" (#4628 on 0.17.0
hit the older 256 MiB limit the same way).
The limit is now an eighth of installed memory, from 512 MiB (what every
64-bit build read before) to 2 GiB: 1 GiB on an 8 GiB machine, 2 GiB on
16 GiB or more. Reading takes about as much memory again as the data
(1.1 GB of data with 24 embedded 4096 x 4096 images peaked 1.1 GB above
the start), so the ceiling keeps a small file that unpacks without end
from taking a large machine's memory. 32-bit and web builds keep 256 MiB.
vectorcraft_render::installed_memory exposes the reading the image cache
already sizes itself from.
Merges kept prepending each side's whole summary as "previously: …", so
target-app-parity.md's status line had grown to 66 MB (839 copies of its
Target clause), gaps.md's and ui-parity.md's to over 1 MB. Each status
line now names its latest change only, as the progress-docs standard has
it; the revision histories keep the earlier ones.
In Affinity Designer, Recolor Adjustment layers (RcRA) can be attached to
vector groups/shapes via AdCh. Lightness 1.0 recolors target elements to
white. Import visible RcRA layers as ColorOverlay effects and apply them
during node appearance evaluation.
A new reader, written from the ZSoft PCX Technical Reference Manual, decodes
the run-length encoded scan lines of 1-bit images in the header palette's
first two colors (black and white when the version has no palette or the two
colors are the same), 1-bit images in two to four planes and 4-bit images
with the header's palette (the EGA colors for versions 0 and 3), 2-bit CGA
images in the colors that the header's CGA background and palette settings
pick, 8-bit images with the 256-color palette after the image data (gray
levels without one) and 24-bit images in three planes. The image is stored as PNG at the resolution of the header's HDpi and
VDpi fields.
Open and Place list PCX, and pasting a copied .pcx file and Glass's Load
Texture… read it by its header. A file of another depth or plane count, one
cut short and one whose pixels as RGBA pass the pixel allowance are refused
with a message before its pixels are decoded. The desktop packaging does not
associate .pcx files, and no export writes them.
Docs: docs/file-format-parity.md, docs/target-app-parity.md, docs/mcp.md,
docs/releasing.md.
CodeRabbit follow-up for #4293. Add 2026-10-26 revision-history rows and
status-line updates for EPS Save and Save As in file-format-parity, gaps
(G9), roadmap and target-app-parity; move the completed EPS Save note in
roadmap's M4 row out of "Left:". Save-back is recorded as done; G9 import
fidelity and Illustrator editing-data writing (#868) are unchanged. Scores
and estimates unchanged. Docs only.
An undefined number (--) in `-- -- angle length Bh` is read as a radial
gradient with no highlight, and XC (a color space name after a linked
image's file name) is ignored like its neighbours. Either one used to make
the whole editing data unreadable, so the file opened as its PDF part and
lost its layers and live text. A -- anywhere else is still an operator the
reader doesn't know.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
* CI: line tables only for workspace crates, no debug info for dependencies
* CI: one integration-test binary per crate
Each crate's tests/*.rs files become modules of tests/it/main.rs. Files that change
process-wide state (Color Settings, the font database, installed plug-ins, the working
directory), the fuzz sweeps and the test targets the Affinity corpus workflow names stay
their own binaries.
* Docs: name the integration tests' new paths
* CI: xtask test with nextest and doctests, one wasm check, clippy --keep-going
cargo xtask ci runs the tests with cargo nextest (cargo test when it is not installed),
builds them with --lib --bins --tests, runs the doctests as their own step, and checks
the wasm crates in one cargo check --keep-going call.
Many tools note things on SVG objects in attributes of their own XML
namespace: a plotter's pen, a cutter's passes, a game engine's
collision flag. Import dropped them, so a file that went through
VectorCraft lost what other software had set on its objects. #522
asked for unknown and data-* attributes to survive a round trip; #580
kept the data-* ones, and this keeps the ones in other applications'
namespaces.
- Import keeps every attribute in a namespace that isn't SVG's,
XLink's, XML's, Inkscape's, Sodipodi's or VectorCraft's own, on the
object or layer it belongs to, in order, with its namespace URI and
the prefix the file gave it (ObjectAttributes::foreign, beside the
data-* attributes and the event handlers). The file is untrusted: at
most 64 per object and 64 KiB per value, and what is left out is
warned about. Files that declare no other namespace skip the scan.
- They are the object's own, as its data-* attributes are: a pasted
copy and a duplicate keep them, and deleting the object takes them
with it. The native file keeps them in the object's attributes, a
key written only when an object has some.
- Export writes them back on the same elements and declares each
namespace once, on the root, under the file's prefix. A prefix the
export uses itself, or that another namespace already has, becomes
prefix2, prefix3...; one XML keeps for itself (xml...) becomes ns. A
name or value XML can't hold is left out with a warning.
Nothing in VectorCraft reads them, and nothing in the UI lists them.
svg/tests/foreign_attributes.rs: a layer's and a shape's attributes
survive import, export and import again, Inkscape's and Sodipodi's
own aside and with no warnings, and the native form round-trips;
prefixes are made unique, and bad names and values are left out with
a warning; the caps hold and are warned about once.
engine tests_attributes.rs: a pasted copy and a duplicate keep them, a
deleted object leaves none behind, and the native file and SVG export
keep them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HhvnnnD5fTFupK8WWq35un
Cause: since #3742 place_each measures a line's spacing from the file's
own widths, so tracking is only the file's Tc/Tw/TJ. Where the widths
are systematically off the installed font's advances (up to the 1% that
still counts as the same font), those differences add up along a line
and nothing made up for them. With no kerning and every tracking value
0, place_each returned the untracked layout early, without the 0.02 em
placement check (CodeRabbit's review of #3742): a line of embedded
Source Sans 3 with /Widths 0.8% wide opened 3.7 pt off. With the early
return removed the check rejects the line and the line-wide fit takes
over: one whole-thousandth tracking value with the font's kerning,
0.30 pt off (as before #3715). Before #3742 main placed it exactly but
with a run per character (49 and 62 runs).
Change: the early return is gone; every line goes through the check.
Before it, on_course() keeps the file's spacing where the glyphs stay
within 0.01 em of the file's; otherwise each stretch of the line takes
one share after every character (the longest stretch one share keeps
within 0.01 em), so ordinary text stays one run, and a line whose
differences change course (30 m then 30 i) takes a run per stretch (2).
Verification: new test
a_long_line_of_widths_a_little_off_the_fonts_advances_lands_in_one_run
(embedded Source Sans 3, widths x1.008 and x0.992, a kerned and an
unkerned line): one run each, glyphs under 0.24 pt (0.02 em) from where
the file draws them. Without on_course it fails after #3742, with and
without the early return ("glyphs 0.300 pt"); before #3742 it fails on
the run count. It replaces an Inter test drafted for the review: Inter
isn't embedded there, so the file's outlines are the interpreter's
stand-in stretched to the widths, not Inter, and the contours differ by
0.6 pt even with exact widths. tests_text_placement and
text_becomes_point_type pass; cargo xtask ci: all 10 steps passed.
Thanks to the CodeRabbit review on #3742 for the early-return finding.
Cause: place_each (crates/pdf/src/import_text.rs, #3715) works out each
character's tracking as the file's distance between glyph origins minus
the installed font's advance. A PDF gives widths to the thousandth of an
em (the standard-14 AFM metrics or /Widths), so with a font of 2048 units
an em (Helvetica on macOS) every character is a hair (~0.0002 em) off.
When the line isn't evenly spaced as a whole (a TJ word gap, a space read
from a gap), `even` keeps every character's own value, rounding noise
included, and restyled() splits the line into a run per character.
text_becomes_point_type failed on macOS; Linux CI passed because Helvetica
isn't installed there, its stand-in's widths differ by more than 1% and
place_each falls back to the line-wide fit.
Change: measure the spacing from the file's own widths: each character's
pen is moved by how much wider the file says the characters before it are
than the font's advances, where the file gives the character's width and
the layout sets it as a glyph of its own. The tracking is then the file's
(Tc, Tw, TJ), whatever font is installed; the layout check that the glyphs
land within 0.02 em of the file's places is unchanged.
Verification: text_becomes_point_type and all of tests_text_placement pass
on macOS. New test widths_rounded_off_the_installed_fonts_advances_keep_one_tracking
uses built-in Inter (non-embedded, widths rounded to thousandths), so it
doesn't depend on system fonts; without the fix it fails on any platform
with runs N/e/x/t/ /li/n/e/ /h/e/re. cargo xtask ci: all 10 steps passed.
The SVG writer always wrote the document's title in <title>
(crates/svg/src/export.rs), so every file of Export As with Use
Artboards, Export for Screens or an `artboard`/`range` export was titled
after the document ("Icons"), whatever its artboard was called. The
writer's `artboard` option can't tell those apart from Save: Save writes
the first artboard too, standing for the whole document.
ExportOptions has a new `artboard_title` (set by the caller, as
`artboard` is, not deserialized): the file is that one artboard, exported
on its own, and its <title> is the artboard's name (XML-escaped as
before), or the document's title when the artboard has no name. The
engine's SVG encoder (fileio/svg.rs) sets it exactly when the export
names its artboards, which is also when a file holds only that
artboard's art (#550): Export As per artboard, document.export /
serialize with `artboard` or `range`, MCP export, and Export for Screens'
artboards, assets, layers and selections (each written as an artboard
named after it). Save and Save As keep the document's title. Export for
Screens' Full Document draws the art bounds as one artboard, which is now
named after the document, so its SVG keeps the document's title too.
docs/file-format-parity.md's SVG row says so.
Tests: title_control_chars::an_artboard_exported_alone_is_titled_after_it
(crates/svg: two named artboards, the second named with markup;
exporting either alone gives its escaped name, the default and an
artboard without artboard_title give the document's title, an unnamed
artboard falls back to it) and
tests_svgenc::an_svg_of_one_artboard_is_titled_after_it (the engine:
`artboard` 0 and 1, useArtboards range all, Save's SVG, Export for
Screens per artboard, an asset and Full Document). Both fail without the
change (the title was "Icons") and pass with it.
Exporting a CMYK document to PDF/X-1a or PDF/X-3 flattens transparency
first. Where the flattener has no single colour (an image with see-through
pixels over a flat fill, gradients, masks, raster effects) it renders the
area into an image, and that image was always an RGB PNG. The PDF writer
then separated it back to CMYK, so CMYK art behind the image came back with
other inks: a C100 M6 Y41 K3 panel printed C75 M14 Y35 K5 under a
transparent logo, with a hard edge against the rest of the panel, and K
moved into C, M and Y.
For files without transparency the flattener's images are opaque and
clipped to their regions anyway, so in a CMYK document it now draws them
as ink amounts over the paper (Renderer::render_region_inks) and stores a
CMYK TIFF, which the PDF writer keeps as DeviceCMYK samples. An area too
large for a CMYK image falls back to the PNG it made before. In-app
Flatten Transparency and other flattening, whose images keep alpha, are
unchanged.
Test: tests_pdfx::pdfx_1a_flattening_keeps_the_inks_behind_an_image_in_a_cmyk_document
exports a CMYK panel under a half see-through PNG to PDF/X-1a, opens the
file again and checks the inks behind the see-through half.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
**What:** Illustrator type styled through the Appearance panel came in as its drawn look (outlined compound paths). Comic and "text effect" templates are the common case, and the type couldn't be retyped. This keeps it as editable type with the style as a live VectorCraft appearance.
**Changes**
- **Art styles parsed as a tree.** Compound filters close at their `;`, basic ones at `/Part`. Each fill and stroke in a stack keeps its own effects, opacity and blend mode, and object-level effects stay on the object.
- **Type keeps any plain stack.** That covers several fills and strokes, pattern and gradient fills, Warp, Offset Path, Transform with copies, Drop Shadow, Outer Glow, Feather, Gaussian Blur and Bas Relief. Bas Relief settings are left at their defaults, because Illustrator stores them in a Photoshop plug-in's binary block. Gradients and patterns are placed as the drawn look places them. Other objects keep the existing plain-style rule, since their drawn look is the more faithful option.
- **Point type without a frame matrix.** Illustrator 24 omits the matrix when point type is neither rotated nor scaled. Such stories were left out entirely; they're now read with an identity matrix.
- **Page comparison.** The page's rendering of a style isn't held to match VectorCraft's live rendering where the replaced drawn look lay, and styled made type isn't pruned as unseen.
- **EPS override.** `editingData: "always"` / `convert --editing-data` now applies to Illustrator EPS as well as `.ai`.
**Tests:** 6 new tests, written from scratch; `vectorcraft-eps` passes all 298 on current main. clippy `-D warnings` and rustfmt are clean for eps, engine and cli. `docs/file-format-parity.md` has been updated, including its revision history.
**Clean-room note:** I found the bugs with a user's licensed stock EPS, used locally only; nothing from it is committed. Filter and key meanings come from the public Illustrator Effect docs and public LiveEffect scripting examples (`jntp`, the Warp style order).
**Not included:** VectorCraft's pixel effects treat transparent content as empty instead of using Document Raster Effects Settings › Background (white by default). Because of that, Bas Relief on a transparent pattern gives flat grey. That needs a separate renderer change.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Text drawn in rendering mode 3 (3 Tr), which is how a scanned page carries
its recognized OCR text, was left out on open (since #2988, with a warning),
so a scanned PDF opened and saved again lost its searchable text.
Import: an invisible glyph now comes in as type with neither fill nor
stroke. It skips the glyph-outline checks that guard how visible type looks,
and it joins only other invisible text on a line (Look::takes no longer
lets it into a stroke-only run). INVISIBLE_TEXT_NOTE now only reports text
that can't be type (text opened as outlines, Type 3 fonts), and
tests_reader_compat::invisible_text_is_reported_when_left_out expects the
invisible line as unpainted type when text opens as type.
Export and print: characters with no fill and no stroke are written as real
text inside a form marked Mark::Invisible; forms::finish sets that form's
text rendering mode to 3, so it draws nothing but stays selectable and
searchable (PDF/X-1a and PDF/A-2b too). The writer has no invisible mode of
its own, so this reuses the marked-form mechanism used for layers and
overprint.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Honor the requested text mode in the single-page importer, as Open does, instead of forcing outlines for Place and PDF clipboard import.
Add synthetic regressions for bytes, unlinked paths, selected pages and explicit outlines. Update format parity and milestone evidence without changing scores.
Saving a version 7 .ai wrote the title, spot color names and gradient
names as the UTF-8 bytes of their characters: a spot color named Café
was written (Caf\303\251) in the header's custom colors and in each x.
Type and, since #2722, layer names are written in Windows-1252, the
encoding the file installs with TE.
These names now go through #2722's Windows-1252 path, so Café is
written (Caf\351). A character outside Windows-1252 is written as ?,
with one warning for all of these names. A spot color keeps one name
in the header, in each x and X and in gradient stops, and a gradient
keeps one name in its definition and in each Bg. Two spot colors or
two gradients whose names become the same that way are numbered (? 2).
Docs: docs/file-format-parity.md.
file.saveAsTemplate with format ait and aiVersion: "7" wrote the
legacy PostScript .ai of version 7 under the .ait name: for a
one-artboard test document, a 7,556-byte file starting %!PS-Adobe-3.0.
The web's Save as Template dialog showed the .ai Version menu for
.ait, whose Legacy (version 7) choice set aiVersion 7.
aiVersion 7 for a .ait is now an error, "aiVersion 7 writes legacy .ai
files; a .ait template is always PDF-compatible", and nothing is
written. The 17 catalogs that translate every message have a row for
it. The dialog shows no Version row for .ait, and aiVersion "pdf"
still writes the PDF-compatible template.
Docs: docs/mcp.md, docs/target-app-parity.md, docs/file-format-parity.md
(two stale rows for .ai and .ait removed).
File > Place could not place a Word document: a .docx file was "not a
format VectorCraft reads". The reference app's File > Place reads
Microsoft Word files as text.
File > Place now reads .docx files, written from ECMA-376 Part 1
(WordprocessingML, Transitional and Strict), and sets their text as
area type the way a text or RTF file is placed. The package is read
through its relationships. The formatting comes from the document
defaults, the paragraph style (else the default paragraph style), the
character style (else the default character style), the styles each
is based on, and direct formatting. Bold, italic, caps, small caps,
strikethrough and hidden text are toggle properties: a style that
turns one on reverses the state the styles before it give.
Each run keeps its font (theme fonts from the theme part), size, bold
and italic, underline, strikethrough, all caps, small caps,
superscript or subscript (in Document Setup's proportions), raised or
lowered text and color. Each paragraph keeps its alignment, start,
end, first-line and hanging indents, space before and after, and
direction; the start indent of a right-to-left paragraph is its right
indent. Table cells are separated by tabs and each table row is a
paragraph. Text in insertions, hyperlinks, content controls and
fields' results is read. Hidden and deleted text, field instructions,
pictures, text boxes, headers, footers, notes and list numbers are
left out.
A ZIP archive named .docx, or one that holds word/document.xml, is
read as a Word document. fileio/zip.rs gains a reader for stored and
deflated files that checks each file's CRC. A package with more than
10,000 files, a part that unpacks to more than 16 MiB or nests
elements more than 256 deep, and a document with more than 4 MiB of
text or 100,000 paragraphs or runs are refused with an error. The
Place dialog lists Word files, and file.place, file.place.info and
file.place.queue report the format `docx`. .doc files are still not
read.
The Word document tests in tests_textimport.rs fail without the
change, and text_import_fuzz.rs places random and damaged packages.
Docs: docs/file-format-parity.md, docs/mcp.md,
docs/target-app-parity.md.
1-bit TIFFs compressed with CCITT Group 3 did not open, place or
relink: the image decoder reads Group 4 only and refuses Compression 2
(modified Huffman) and Compression 3 (ITU-T T.4).
They are now decoded with hayro-ccitt, the CCITT decoder that EPS
import uses. Compression 2 rows start at a byte and have no EOL codes.
Compression 3 rows follow an EOL code and are one-dimensional, or, with
T4Options bit 0, one- or two-dimensional as each row's tag bit says.
Fill bits, FillOrder 2, either PhotometricInterpretation, strips and
tiles are read.
A file whose pixels would take more than the decoding allowance as
RGBA is refused as too large to draw. A strip or tile whose data ends
before its last row is refused as damaged, and so is a tile larger
than the allowance. The strips and tiles together are read from at
most as many bytes as the file holds.
Part of #1466.
Docs: docs/file-format-parity.md, docs/target-app-parity.md.
File > Place read text files only as plain `.txt`, and an RTF file
could not be placed. The reference app lists RTF among the formats it
places as text, keeping the file's font and style settings.
File > Place now reads RTF files, written from the Rich Text Format
Specification 1.9.1, and sets their text as area type the way a text
file is placed. Each run keeps its font from the font table, its size,
bold and italic, underline, strikethrough, all caps, small caps,
superscript or subscript (in Document Setup's proportions), baseline
shift and its color from the color table. Each paragraph keeps its
alignment, left, right and first-line indents, space before and after,
and direction. Table cells are separated by tabs and each table row is
a paragraph. Unicode characters, Windows-1252 and Mac Roman bytes and
the special-character control words are read, and of a \upr group's
ANSI and Unicode text, the Unicode text is read. Hidden text, pictures,
objects, headers, footers, notes, field instructions and other
destinations are left out, and unknown control words are skipped.
A file is read as RTF when it starts with {\rtf, whatever its name,
except a .txt file, which stays plain text. RTF files share the 4 MiB
limit of text files. Groups nested more than 512 deep share the
formatting of the deepest one read, the font and color tables keep at
most 4,096 entries, and a file with more than 100,000 paragraphs or
runs is refused. The Place dialog lists RTF files, and file.place,
file.place.info and file.place.queue report the format `rtf`.
The tests in tests_textimport.rs fail without the change, and
text_import_fuzz.rs places random RTF token streams.
Docs: docs/file-format-parity.md, docs/mcp.md,
docs/target-app-parity.md.
Don Omondi