Declaration binding instantiation reserves an upper bound of what a script's,
eval's or function's variables and function declarations add to the heap
before making them, and each declaration and parameter takes a unit of fuel.
A long declaration list now ends with the heap limit without growing the
estimate past it.
Realm::set_limits recomputes when the next collection runs, so a lowered heap
limit applies to the next run instead of after the old threshold.
ToNumber skips the same white space and line terminators as the lexer, so
U+180E is no longer white space in a numeric string.
A new property or binding now adds its name's code units to the heap
estimate when it is inserted, so names declared by eval or Function
source move the estimate toward a collection. Function parameter and
catch bindings no longer count their entries twice.
A host function's error message reserves its size in the heap before
the error object is made, so a message larger than the heap ends the
run with the heap limit.
The Function constructor sums its arguments' lengths and checks the
source limit before flattening or copying any of them, and builds the
source from their code units, so a lone surrogate in an argument is
kept.
In script and eval code a finally block that ends normally leaves the completion value of its try or catch block (ECMA-262 5.1 section 12.14), so try { 1; } finally { 2; } completes with 1. A new GetCompletion instruction saves the value when the finally block starts, and SetCompletion sets it back when the block ends normally; a break, continue or throw out of the finally block still replaces it.
A string returned by a host function is charged to the heap, and one that does not fit after a collection ends the run with the heap limit.
A collection counted each property or binding at a fixed 64 bytes and
never counted the name's own buffer. Once the string that produced a long
name became garbage, the name kept its buffer alive while the estimate
dropped it, so a script could hold megabytes of names past the heap
limit. trace_object and trace_env now add each name's buffer, counting a
buffer that a flat string value or another name shares once.
The constructor compiles its source as (function anonymous(...) {...}), and the
compiled function bound anonymous to itself as a named function expression does.
That binding shadowed a global variable of the same name. ES5.1 section 15.3.2.1
creates the function with the global environment as its scope and no name binding,
so the constructed function now leaves the name unbound.
Listing a String object's index names for for-in built one name per UTF-16 unit before any
limit check, so a string within the heap limit could make the loop allocate many times that
limit. The heap budget for all the names is now checked first, after a collection if needed,
and each name takes a unit of fuel. A wrapper made by ToObject stays rooted while the names
are listed.
Realm::run compiles the script with the realm's Options::strict instead of always as sloppy code, so a realm made with strict: true rejects with statements and calls functions with an undefined this. Indirect eval and Function keep their own strictness (ES5.1 10.4.2, 15.3.2.1).
Function.prototype.apply drops the temporary GC roots it holds for the argument values when an index getter throws, as it already did when reading succeeds. Before, a caught throw left them rooted until the run ended, so a loop of such calls could reach the heap limit.
A symbol whose first instance is written in the drawn look of an object
with a style (through a member's scatter brush) was defined while the
look's flag was set, so the warning for the effects its art loses was
left out, and the cached definition kept it out for later instances.
The definition now writes its art with the flag cleared and restores it
afterwards.
Realm::run now resets the exception that bind_last_thrown binds, so a run
that completes, reaches a limit or does not parse leaves it undefined. The
script source, the text built by Function and the strings given to eval are
checked against the run's remaining max_source_units before they are encoded
or flattened, so an oversized source is rejected without being copied.
vectorcraft-js gains a Realm that runs ECMAScript 5.1 scripts. The parsed
script compiles to stack-machine bytecode, a verifier checks every jump
target, handler, constant index and stack depth before the code runs, and
the interpreter keeps JavaScript calls on its own frame stack. Closures,
this, arguments, with, direct and indirect eval, new Function,
try/catch/finally, labeled break and continue, and for-in in insertion
order follow sections 10 to 14 of ECMA-262 5.1. Variable references are
resolved before the right-hand side of an assignment runs (11.13).
Objects and environments live in an arena with a mark-and-sweep
collector and index handles. Built-ins are limited to Object and
Object.prototype's basic methods, Function with call and apply, eval
and the error constructors.
Every run is bounded by fuel, heap bytes, frames, nested host calls and
an interrupt flag another thread can set, with smaller defaults in the
wasm build. A run that reaches a limit ends with a limit error that a
script's try cannot catch. Source passed to eval and Function counts
against Options::max_source_units.
cargo xtask test262 also runs the suite's ES5 test/language tests that
need only these built-ins and prints their counts apart from the parse
counts. Nothing in the app calls the interpreter yet.
A stroke-only path or compound path that uses the even-odd rule keeps
its live effects in the editing copy; the art-style writer leaves out
only paths that fill by the even-odd rule.
Noto Sans Arabic draws a letter's dots as zero-advance GDEF marks, so the tests expected a beh's dot to take its kasra's colour. Compare against the glyphs of the bare letters instead. Strict colour, geometry and kashida ownership assertions stay as they were.
The editing copy of a PDF-compatible .ai left out the Graphic Styles
panel's styles and objects' links to them, with a warning. They are now
written as art styles:
- The panel's styles come first in %AI9_BeginArtStyles and are listed in
its order in %AI9_BeginArtStyleList after the palette, with their
fills, strokes, transparency and effects. Fills and strokes may have
transparency and effects of their own, and a style may also hold Warp,
Transform, Free Distort, Pucker & Bloat, Twist, Zig Zag or Gaussian
Blur. A style of an object's own holds the same effects as before.
- A path, compound path or group linked to a graphic style names it when
it has the style's appearance, the style is written, and the reader
reads that style on that object. Objects linked
to a new document's styles keep their link: the panel is written for
them too.
- The writer reads each style back with the reader before it writes
it. A style with gradient or pattern fills, with anything the
reader does not read back, with a name another style has, with a name
over 1,024 bytes or past 4,096 styles is left out with a warning. Other
links are left out with a warning, and a panel with no style written
warns that it opens with a new document's styles. No graphic style is
written on a large canvas, with a warning.
Graphic styles:
https://helpx.adobe.com/illustrator/desktop/special-effects-styles/apply-graphic-styles/apply-graphic-styles.html
Tests: the round trip of the panel's styles and of links; each effect a
style can hold; the forms written; the list's limits; duplicate names; an
empty panel; links left out; a style's opacity at the written precision;
a large canvas; Twist and Transform on an object with no link; a link
to a style whose Transform moves the outline, with and without an
opacity mask; the
reader's reading of warp numbers; fuzz cases for damaged written styles.
Docs: docs/file-format-parity.md.
The editing copy keeps live effects on qualifying objects only when the art-style writer supports their appearance and effects. README, ROADMAP and the progress docs now say so and link to docs/file-format-parity.md for the limits.
The .ai editing copy writes an art style for paths, compound paths and groups with live effects only when they have at most a fill and a stroke, fill by the non-zero rule, and are not on a large canvas. README, ROADMAP, gaps, roadmap and target-app-parity say so.
The fallback for objects that can't use an art style keeps their vector
art with geometry effects applied and leaves out their other effects,
such as Drop Shadow, the glows and Feather. README, ROADMAP, gaps,
roadmap and target-app-parity no longer say every live effect is left
out, and target-app-parity lists the round-trip tests.
Current test262 (ECMA-262 2017 Annex B.3.6) accepts `for (var x = 1 in o)` only outside strict
mode code. The parser now rejects the initializer in strict code, with the error at `in`.
Paint-only harakat runs stay inside one shaping cluster so letters keep
their joining forms and marks their attachment. When a character-level
shaping pass matches the monotone glyphs, a glyph may take a later
in-cluster character's run only if GDEF classifies it as a mark (class 3)
and that character is a harakat; otherwise ownership stays at the cluster
start. Conservative: a mark may miss its colour; a base letter must never
take a mark's. Fonts without GDEF and AAT faces fall back the same way.
Insert Kashidas tatweels take the base letter's paint, not a trailing
mark's. Private harakat module; type-parity, target-app-parity and roadmap
M7 evidence updated (scores unchanged). Tests cover paint_only / ranges /
own_chars (wide mark, equal advance, multi-component, single-glyph
ligature, review counterexample), colour, hide, repeated glyph ids,
kashida, forced mark-last emit_kashida, and installed fonts.
A new standalone crate, vectorcraft-apischema, describes a command's
parameters: type (number with range and unit, integer, boolean, string,
enum, id, id list, color, point, object, array), whether each is
required, the default of an omitted value, whether null counts as
omitted, the API levels it was added and deprecated at, and the lenient,
cost and writes fields. It validates a params object and names the
parameter of the first problem, writes the params text, writes the
schema in the shape of a JSON Schema 2020-12 object schema, and hashes a
set of schemas.
CommandSpec gains an optional schema (both builders: the cmd! macro and
the text style specs). Session::execute checks the params of a command
that has one before anything else, refusing a value of the wrong type,
out of range or not in its list, a missing parameter or an unknown key
with an error naming the parameter, and fills in defaults; the journal
keeps the params as given. engine.commands lists each schema.
The six commands in cmd/views.rs have schemas, and their params text is
the text the schema writes. schemas::API_VERSION is 1, and a test fails
when a schema changes without a new row in SCHEMA_HASHES.
Docs: docs/mcp.md (Parameter schemas), docs/target-app-parity.md.
New crate vectorcraft-js reads ECMAScript 5.1 source into a syntax tree
with source positions: a UTF-16 lexer (line and column tracking, automatic
semicolon insertion, the regexp and division goals), preprocessor lines
read as directive tokens, a recursive-descent parser to an arena tree, and
the ES5 early errors, strict mode included. Nesting deeper than
Options::max_depth (200 by default) gives a SyntaxError with its line and
column rather than overflowing the stack. Nothing runs scripts yet.
cargo xtask test262 downloads test262 at a pinned commit, checks its
SHA-256, unpacks it into the gitignored corpus/ folder and runs the ES5
syntax tests. A differential test in the engine compares the lexer with
scriptcheck on line counts, directives and unterminated literals.
The contour passes of the two-path boolean operations, Offset Path and
the many-shape boolean (`boolean_n`, used by compound shapes and
Intertwine) built a flag for every input shape for each winding they
tested. They now test the list of inputs that cover the face, ascending,
refilled in one buffer per pass, so a test costs the size of the cover
rather than the number of shapes. `boolean_n` and `try_boolean_n` take
that list; the compound shape and Intertwine tests read it directly.
The contours, their order and the output are the same: a new test
compares every pass's test with the flag form on overlapping, nested,
touching and apart squares and on 72 overlapping squares, and another
checks that a pass over 100 squares apart hands its test one input at
most.
The editing copy of a PDF-compatible .ai wrote an object's live effects
as their result (geometry effects) or left them out (the others).
Objects with effects the .ai reader reads are now written as art styles:
- An object is written as its drawn look (a group of its vector art as
its geometry effects leave it, without shadows, glows or feathers,
with the object's transparency after it), then the object itself on
%_ lines (its points painted with nothing, or with its gradient fill
placed on the outline it draws; a group with its members), its name
and opacity mask, and
1 (style) XW naming its style.
- Styles are written in a non-printing block after the symbols, in
%AI9_BeginArtStyles: a chain of the stack of fills and strokes (with
the contents where the appearance puts them), the effects and the
object's transparency. Objects of the same appearance share a style.
- The effects written are the ones the reader reads on objects, with
their settings in the units it reads them in: Drop Shadow, Inner Glow,
Outer Glow, Feather, Round Corners, Scribble, Convert to Shape, Offset
Path, Outline Object, Outline Stroke and Pathfinder > Add.
- Each style is read back before it is written, by a new reader
function, read_art_style, which runs the reader's style parser and its
reading of a graphic style with the object's own gradient. An object
whose style reads back differently (several fills or strokes, fills
or strokes with transparency or effects of their own, pattern fills,
other effects), a filled object that fills by the even-odd rule (a
style's fill fills by the non-zero rule), a group whose effects change
its members' outlines while a member is written on another outline
(strokes as filled outlines, pattern fills as tiles, gradients that
fit their object), or one in a container whose own
effects change its members, is written as its drawn look with the
warning, and a style that is not written leaves no gradient
definition behind. Graphic styles and links to them are left out with
a warning. No art style is written on a large canvas.
Appearance attributes:
https://helpx.adobe.com/illustrator/using/appearance-attributes.html
Tests: the round trip of objects with each effect, a gradient fill, a
group with transparency and a mask, Pathfinder > Add and a hidden
effect; the same bytes on a second write; the forms written; the
objects the reader would not read; a large canvas; the warning for
graphic styles; filled even-odd objects; a gradient placed on the
outline Offset Path draws; a style that is not written; effects in a
pattern a drawn look uses; the reader's check of a style; fuzz cases
for damaged written styles.
Docs: docs/file-format-parity.md.
A paragraph that ends in a soft return ends with the empty line after it,
so the line the soft return ends is justified like any other; it was taken
for the paragraph's last line and left aligned.
Key 30 of a paragraph's attributes holds its tab stops: each its position
(key 0) and alignment (key 1: 0 left, 1 center, 2 right, 3 decimal).
They were not read, so tabs fell on the default stops. At most 1,024 are
read a paragraph and 2^20 copied into a story's paragraphs, past which the
default stops apply with a warning. The fuzz test feeds hostile paragraph
attributes: alignments, auto leading, hyphenation options and tab stops.
In right-aligned area type, the text before a line's last tab sits at its
stops, measured from the left edge, and the text after it (none when the
tab ends the line) against the right edge. The line was laid out from the
left and aligned as a whole, which moved its start off the left edge and
its tabs' text off their stops. Illustrator's own layout in a user's .ai
sets its price lists so. Centred and justified lines are aligned as a
whole as before.
The line names trapping, Intertwine and the Pathfinder effects next to
Divide, Hard Mix, Soft Mix, Shape Builder and Flatten Transparency, and
gives six alternating runs of each build of Divide on 3,000 overlapping
rectangles, timed on this branch's final code.
Finding the distinct regions of an arrangement built a flag for every
input shape at each edge and kept them in a sorted set, so it cost edges
times shapes in time and memory. Each coverage is now the sorted list of
the shapes that cover it, deduplicated in a hash set and sorted into the
order the flag masks had, so regions, their order and the output of
Divide, Hard Mix, Soft Mix, Shape Builder and the flattener are the same.
The face lookup and Pathfinder merging take the same lists.
Tests compare the lists and the regions with the flag masks for
overlapping, nested, touching and disjoint shapes and for 72 shapes.