Retain the referrer selected for each completed hop before applying the redirect response policy. A later permissive policy cannot restore an omitted referrer or stripped path. Apply this to buffered, HTML streaming, raw streaming and cached redirects, with native HTTP regressions for policy changes, manual redirects and request identity preservation.
Add the shared WebIDL octet converter and derive support for explicit octet and inferred u8 fields. Convert through JavaScript numeric coercion and modulo arithmetic. Regressions cover negative/fractional/large values, nonfinite numbers, thrown conversion identity, argument sequence order, dictionaries, and unsupported primitives. MIDI bindings are separate.
Vendor rust-url 2.5.8 with narrowly documented file-URL parsing fixes. Preserve leading empty path segments, retain non-local hosts while normalizing Windows drive letters, and apply the same rules to relative resolution and pathname setters. Include licensing, the upstream fixtures, native URL regressions, and JavaScript URL property regressions.
Build on the shared SVG animated-length getter ownership fix to validate geometry and graphics receivers through native brands. Keep producer-realm identity for cold borrowed getters and registered native proxies; reject forged and author-proxy receivers without invoking traps. Cover live, windowless and foreign-document geometry attributes, transforms and path lengths.
Validate SVGFilterElement and SVGMarkerElement getters through generated native receiver brands. Canonicalize registered native proxies before caching animated length, enumeration, and angle tear-offs, and create those values in the owner realm. Regression matrices cover forged and author proxy receivers, borrowed getters, windowless documents, and registered native proxy identity.
Use typed WebIdlArgs for destinations and port numbers, leaving receiver
validation in the shared AudioNode declaration. Complete disconnect
argument conversion before checking graph state or removing edges.
Move choose-_parent-001/002/004 to passed after receiver-based target
resolution. Each passed five isolated release runs on this main-based
branch. Refresh only these three cases and preserve all other recorded
outcomes; this is not a new full-suite run.
Resolve window.open _parent/_top targets from its receiver Window instead
of re-reading the ambient dispatch scope. Recognize child receivers on
these target paths, including borrowed Window.open methods. Preserve the
entered Window fallback for other receivers and the existing dispatch
scope used by element/form navigation.
Cover nested receivers, mixed-case _parent targets, borrowed methods, and
empty/nonempty URLs. Fix choose-_parent-001/002/004 without depending on
other popup or storage fixes.
Validation:
- cargo fmt --all
- cargo clippy --workspace --all-targets --all-features -- -D warnings
- cargo nextest run --no-fail-fast: 19,937 passed, 17 skipped
- release WPT: each affected case passed five isolated attempts
- nearby cases: 10 pass, 5 timeout; the five timeouts match the checked-in
case statuses at the main base, with no new failing classifications
Carry received HTTP reason phrases from the parser through buffered and streaming responses, disk and memory caches, page responses, Fetch, and XHR. Preserve custom and empty HTTP/1 phrases and report empty text for HTTP/2 and HTTP/3. Synthesized responses retain canonical fallback. Regressions cover whitespace, non-ASCII bytes, and cache round trips.
Validation on this branch: cargo fmt --all; cargo clippy --workspace --all-targets --all-features -- -D warnings.
Nextest and WPT reruns omitted at the user's request.
Includes the ablation-confirmed cleanup from e5491f4b65.
Validation on this branch: cargo fmt --all; cargo clippy --workspace --all-targets --all-features -- -D warnings.
Nextest and WPT reruns omitted at the user's request.
Validation on this branch: cargo fmt --all; cargo clippy --workspace --all-targets --all-features -- -D warnings.
Nextest and WPT reruns omitted at the user's request.
Propagate an unhandled nested worker exception to the parent worker with ErrorEvent.error set to null, as required by omitError reporting. Update the existing end-to-end parent onerror regression to inspect the null value while retaining filename, line, column, and message checks.
Convert getComputedStyle Element and nullable pseudo-element arguments through the shared WebIDL derive path in main, iframe and popup windows. Preserve conversion exception identity and prevent pseudo-element conversion for invalid element receivers. Regression fixtures cover cross-realm genuine elements, author proxies and thrown coercions. Retain current window receiver and scrolling infrastructure. Determine expected intrinsics from the actual binding function realm while retaining main, iframe and legacy popup document coverage. Treat disconnected child-frame and popup targets as empty before consulting cached context state. Retained style objects update across connection and removal; the 318-assertion matrix keeps all original conversion checks.
Validate DOMImplementation native receivers before argument conversion, and convert createDocument qualifiedName before its nullable DocumentType argument. Allocate factory results in the owner Document relevant realm after conversion, preserving the callee realm for conversion errors. Cross-realm regressions cover genuine and forged doctypes, author proxies, conversion order, and factory result ownership.
Return Document.rootElement from the current native document element only when it is an SVG-namespaced svg root. Retain wrapper identity in the Document owner realm without reading author-overridable properties. Regressions cover root replacement/removal, adoption, cloning, namespace/case distinctions, cross-realm access and registered native Document Proxies.
Add an incremental XHR text/document decoder that distinguishes text, XML and HTML response kinds. BOM precedence, transport charset, XML declarations and HTML sniffing retain state across arbitrary chunk boundaries. Native unit tests cover all split points and truncated input. This PR adds the decoder API; switching every XHR producer to it is separate work.
Group ignored reasons, values and nonempty properties behind one optional box. Use nonzero backend IDs and serialize fixed role/name kinds on demand, reducing the common AX node from 296 to 96 bytes while preserving payload shape and field order.
Add reproducible CLI and native construction benchmarks. Generated 40000-row text dumps use 382.6 to 341.5 MiB peak RSS; 200000-row text dumps use 1753.5 to 1612.8 MiB. Record the roughly 5.5 percent native construction cost for dense ignored nodes alongside the memory gains.
Validation: cargo fmt --all; workspace clippy with all targets/features and warnings denied; cargo nextest run --no-fail-fast (19890 passed, 17 skipped). All 188 formal benchmark samples match their corresponding before/after outputs.
Large semantic dumps retain a JSON object graph with allocated field names
and string node references for every AX node before formatting the output.
Build compact typed nodes with numeric references and pass them directly
from the renderer to the CLI. Keep one AX traversal for both typed nodes
and CDP payloads, converting protocol nodes individually as they are built.
Preserve omitted/empty fields, field insertion order, list markers, and
frame-root attachment order. Render text iteratively and trim in place.
Add compatibility coverage and a standalone benchmark that stages both
executables on the same filesystem, alternates runs, records peak RSS and
wall time, and rejects changes to dump output bytes.
Against main at 5f5e1fff3 on a generated 40,000-row list (three interleaved
runs, medians):
- semantic_tree_text: 718.42 -> 390.96 MiB; 2.919 -> 2.204 s.
- semantic_tree: 815.71 -> 483.81 MiB; 2.484 -> 2.369 s.
- markdown control: 95.52 -> 95.59 MiB; 0.895 -> 0.913 s.
A generated 200,000-row list also completes with identical bytes (one run):
- semantic_tree_text: 3473.02 -> 1775.07 MiB; 13.245 -> 10.015 s.
- semantic_tree: 3803.45 -> 2137.26 MiB; 12.626 -> 10.397 s.
JSON/text comparisons also match the baseline on a mixed fixture containing
hidden/ignored nodes, form values, list markers, shadow DOM, and child frames.
These generated fixtures are independent of the external MOL-269 evidence.
Validation:
- cargo fmt --all
- cargo clippy --workspace --all-targets --all-features -- -D warnings
- cargo nextest run --no-fail-fast (19,885 passed, 17 skipped)
- cargo build --release -p moli
Refs: MOL-269
Use computed style facts for content-visibility locks instead of reading
inline declarations. Hidden-until-found and stylesheet-hidden contents
stay out of innerText until revealed or overridden by the CSS cascade.
Remove the obsolete inline-only readers and cover hidden-until-found,
boolean hidden, synchronous reveal, and important stylesheet overrides.
Fixes MOL-113.
Add borrowed AllowShared BufferSource conversion for algorithms that copy after dictionaries are converted, including native range writes and atomic shared-memory byte access. Reject resizable/growable storage and author proxies using native identity, including detached resizable buffers and views. Tests cover late snapshots, bounded writes, surrounding bytes, detachment and native resizable flags. Individual API migrations are separate.
Refresh all-profile WPT Cross lists from Moli 1.1.16 release at 23c6bb5754
against WPT db95fafd1f, using CDP with 50 workers. Each of 13,631 cases
ran once: 9,674 pass, 3,656 fail, 253 timeout, 37 harness-stalled,
10 error and 1 crash. Pass count rises by 86 (132 gained, 46 lost).
Review all 46 lost passes against the previous snapshot's binary:
41 stable version regressions, 2 existing flaky cases, 1 existing runner
readiness race and 2 cases failing in both versions. Preserve original
full-run outcomes in the lists. Validate exact report agreement, sorted
disjoint coverage and no missing cases; git diff --cached --check passes.
Validation on this branch: cargo fmt --all; cargo clippy --workspace --all-targets --all-features -- -D warnings.
Nextest and WPT reruns omitted at the user's request.
ldm0