Admit exact string heap aliases and retain independent Module row facts for distinct declarations. Bind the optional partial row to the same scope checks as named Module Evidence.
* fix(managed): preserve duplicate GUID stream ambiguity
* fix(managed): preserve ambiguous metadata stream declarations (#12226)
Use one stream declaration owner for heap and table diagnostics. Keep user strings and row-layout uncertainty separate from independent PE/CLI observations.
* test(managed): separate heap entries from cache behavior
* fix(managed): preserve duplicate blob heap ambiguity (#12258)
Resolve #Blob in the shared stream layout. Preserve nil references and independent facts without emitting unsupported keys, signatures or hashes.
* fix(erlang): disclose BEAM bytes trailing the declared FOR1 form
A conformant FOR1 container sizes its form to the whole file, so bytes
past the form end are an unexamined appended or corrupt region. The
inspector decoded chunks only within the form but still reported
coverage.status "complete" with no mention of the surplus, so a caller
could not distinguish a clean BEAM from one carrying foreign trailing
bytes.
Compute the trailing byte count, report coverage.status "partial" with a
trailing_bytes field, and append a limitation naming the form end and the
unexamined count. Clean containers stay "complete" with trailing_bytes 0.
* fix(erlang): reparse retained BEAM Evidence without the trailing-bytes field
Evidence retained by a build predating coverage.trailing_bytes has no such
property, and projectErlangBeamView reparses that record with the strict
erlangBeamSchema. A required field rejected those records with an input error
before any summary or page could render, breaking the offline retained-Evidence
flow even for clean files.
Default coverage.trailing_bytes to 0 so historical records reparse while every
new parse still emits the concrete count and preserves each record's stored
status.
* fix(erlang): keep retained BEAM trailing coverage unknown, not zero
Defaulting coverage.trailing_bytes to 0 manufactured a "no trailer" reading for
Evidence retained before the field existed, even when its retained bounds
(artifact.bytes versus form.location.bytes) prove a trailer. Make the field
optional so a pre-field record reparses as unknown instead of a fabricated
zero, while every new parse still emits the concrete count.
* fix(erlang): admit chunk extents before applying count limits
* docs(erlang): explain trailing BEAM coverage
---------
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* feat(elf): decode FDO packaging and dlopen note metadata
Name FDO notes within their owner's namespace: type 0xcafe1a7e is
fdo_packaging_metadata and 0x407c0c0a is fdo_dlopen_metadata. One
elfNoteTypeName(owner, type) replaces the GNU-only gnuNoteTypeName.
Each entry gains an fdo field next to gnu. metadata holds the descriptor's
JSON object (packaging) or array (dlopen) when it is NUL-terminated UTF-8
JSON followed only by NUL padding; otherwise metadata is null and issue
names the broken rule, without failing the inspection. Each descriptor
range is decoded once and its retained JSON text is charged to the shared
note text budget per repeated entry. A limitation marks the metadata as a
producer declaration.
Fixes#7235
* fix(elf): key FDO note decoding by note type
The FDO decode cache was keyed by descriptor range only, but the note
type selects the required JSON shape. Overlapping PT_NOTE and SHT_NOTE
tables can give one descriptor range a packaging note and a dlopen note,
and the second reused the first note's verdict. Include the type name in
the key.
Refs #7235
* fix(elf): bound FDO note decoding before parsing JSON
The FDO decoder charged only descriptor source bytes before JSON.parse,
but a dense JSON value such as [{},{},...] decodes to about 50 times its
source size in heap before the service and Evidence copy it again. A
16 MiB descriptor took about 790 MiB of heap; one near the default
64 MiB file budget could exhaust the Node heap. Leave descriptors above
64 KiB undecoded with an issue naming the limit; real FDO notes are a
few hundred bytes.
Refs #7235
* fix(elf): share one FDO decode allowance across the report
The 64 KiB FDO decode bound applied to each descriptor, so raising
max_entries to 256 admitted 256 distinct dense 64 KiB notes: a 16 MiB
file added about 700 MiB of heap before Evidence copied it again.
Charge each distinct descriptor's size to one 64 KiB allowance per
report; a range repeated by another table reuses its decode, and notes
past the allowance keep their entry with an issue.
Refs #7235
* test(desktop): exercise the shared JSON depth guard in the heap suite
---------
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* test(gpu): protect CUDA 0x2000 LZ4 payload identity
CUDA LZ4 entries were decodable through the shared block decoder, but no
public workflow selected a 0x2000 entry and checked the retained PTX bytes.
The fixture builder offered only none/zstd/legacy, so the profile dispatch in
readCudaFatbinary had no decoded-payload control.
Add an lz4 fixture profile backed by a single literal sequence, extend the
existing boundary matrix to it, and add boundary plus CLI/MCP regressions that
assert compression_id, decoded size, retained PTX text and the decoded digest.
A malformed decoded-length control covers the rejection path. Temporarily
disabling the 0x2000 profile selection makes all three new cases fail.
Verified with npm run check:fast and the GPU boundary, acceptance, inventory
and LZ4 suites (161 passed). No native liblz4 or CUDA compiler is required.
* style(gpu): apply repository formatting
* test(gpu): follow the shared CLI session rename in the code-object test
Upstream #10412 removed runFileCli from the CLI fixture and moved every
acceptance test onto the file-level shared process (cli.run). Rebasing
onto current main left this new LZ4 assertion calling the deleted helper,
so the whole file failed with a ReferenceError and the shared worker was
torn down for the rest of the suite.
* feat(rust): read cargo-auditable dependencies from WebAssembly modules
cargo-auditable embeds the same zlib-compressed `.dep-v0` JSON in a
WebAssembly custom section for wasm32 targets, but inspect_rust_binary
rejected every module as an unsupported target because only ELF, PE and
thin Mach-O containers were recognized.
Walk the section framing of version 1 core modules to locate the
`.dep-v0` custom-section payload and the data section, where rustc keeps
its constant /rustc/<hash>/ paths. Rust lookups now read one
format-neutral layout, so native and WebAssembly images share a single
dependency-section owner. WebAssembly declares no container word size, so
`bits` is null for that format only; the Evidence subject format is
`file`, as for inspect_wasm_artifact.
Truncated sections, invalid LEB128 sizes, custom-section names that run
past their section or are not UTF-8, duplicate `.dep-v0` sections and a
second data section are malformed. Components, such as wasm32-wasip2
output, embed core modules and are refused as unsupported, like universal
Mach-O containers, rather than inspected through one selected module.
* fix(rust): bound the WebAssembly section walk
The WebAssembly reader visited every section synchronously, and the format
sets no section count. A 256 MiB input of three-byte custom sections with
empty names walked about 89 million sections and held the event loop for
about 7.4 seconds, delaying cancellation and other MCP requests; empty type
sections took about 1.9 seconds.
Refuse a module with more than 65,536 sections with a typed cpu resource
constraint that reports maximum_wasm_sections, as the Go header search
reports its work budget. Producer output holds a few dozen sections, and the
same inputs are now refused within tens of milliseconds.
* fix(rust): accept WebAssembly Evidence in retained dependency views
Retained Rust views compared the parent subject with a Mach-O-only format
mapping, so WebAssembly Evidence, whose subject format is file, was refused
as a different artifact. The report domain now owns the subject format next
to the architecture mapping, and the service and the view both use it. The
real CLI/MCP view workflow now also covers a WebAssembly module.
* fix(rust): name the WebAssembly section-walk allowance in recovery
Fixed-allowance recovery mapped every Rust resource other than capacity and
transport to the parser allowance, so the section-walk refusal said host
memory does not raise it. Each Rust resource now maps to its own allowance,
and the cpu bound is a fixed work allowance. The MCP error contract test
checks that the refusal is not retryable and names that allowance.
* fix(rust): require WebAssembly container facts in reports
The report schema tied a null word size to WebAssembly, but it accepted a
wasm report that named a native architecture or big-endian byte order, and
a native report that named the wasm architecture. Retained views parse
imported Evidence with this schema, so such a contradictory report was
accepted. The architecture is now wasm exactly when the container is
WebAssembly, and a WebAssembly report must be little-endian.
* docs(rust): record the WebAssembly dependency-metadata feature
Feature work now carries a spec, plan and task list under docs/specs.
This adds that record for reading cargo-auditable metadata from
WebAssembly core modules and indexes it. The record states that it was
written during review and lists the verification that remains open.
* fix(rust): bind section names to containers and share the LEB128 reader
The report schema accepted a dependency section name that does not belong
to the report's container, such as a WebAssembly report naming the Mach-O
section. rustDependencySectionName now owns that name for the reader and
the schema, which refuses a mismatch for every container.
The WebAssembly reader decoded section sizes with its own loop. It now
uses the shared readLeb128, and an integer cut off by the end of the file
is truncated rather than malformed.
* fix(rust): classify a cut-off WebAssembly integer by its boundary
The reader called an unfinished LEB128 integer truncated whenever its
limit equalled the file length. A name length cut off inside the final
section was therefore truncated, and the same bytes followed by another
section were malformed. Each caller now states whether the file or the
enclosing section bounds the integer.
* fix(rust): report a duplicate WebAssembly dependency section first
The section walk kept two .dep-v0 matches but walked on, so a module with
two dependency sections and more than 65,536 sections after them was
refused for the section bound instead of the duplicate it had already
established. The walk now ends at the second match.
* test(rust): assert WebAssembly container facts instead of limitation prose
The CLI acceptance test matched an English limitation sentence. It now
checks format, architecture, bits and byte order on the parsed report.
* fix(rust): state container word size as 32 | 64 | null
Output contract: inspect_rust_binary bits, including the summary view, is
32 | 64 | null. Null is valid only for WebAssembly and is not a number.
ELF, PE and Mach-O reports stay 32 or 64. The schema description, guide,
feature record and skill state that, and CLI renderers keep null instead
of substituting a word size.
---------
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
* feat(ghidra)!: verify the bridge on Ghidra 12.2
Admit 12.x from minor 2 upward, with 12.2 and JDK 25 as the verified
pair. The bridge now calls the four-argument storage API added in 12.2,
so 12.1.x cannot compile it and is no longer accepted.
BREAKING CHANGE: Ghidra 12.1.x is rejected. Use Ghidra 12.2 or another
12.x build at or above minor 2, and a JDK inside that installation's
declared range (25 or newer for the verified 12.2 build).
* docs(ghidra): point remaining lanes at Ghidra 12.2
The Windows, 6502, and m68k prerequisites still named 12.1.4 and JDK 21,
which the verified-build check now rejects. The bridge compile lane names
the 12.2 jars it actually compiles against.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(ghidra): admit the JDK the verified install declares
The Java home framing probe still reported JDK 21 while the verified
installation fixture now declares a minimum of 25, so doctor admission
failed before the path framing was checked.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(ghidra): derive admitted install fixtures from the verified release
Boundary hosts and assertions still named Ghidra 12.1.4 after admission
moved to 12.2, so list_procedures evidence and completed property reads
disagreed with the install the provider actually reported.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(ghidra): record the exact 12.2 Clang switch-recovery defect
Ghidra 12.2 stays the verified build. Sessions and decompiler results on
that exact version name upstream issue 9996 and the 12.2.1 fix commit.
Other version strings, including 12.2.1, do not carry the limitation.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(ghidra): check switch-defect facts instead of the warning sentence
The policy test keeps the structured defect record and checks that the
owning formatter delivers its id, commit, release, and scope fields.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(browser): name the Playwright launch boundary the lifecycle catalog checks
Main retitled that case with a platform placeholder. The coverage catalog
still required the old literal, so the merged full suite reported a missing
title.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(browser): point the lifecycle catalog at main's launch title
The merged suite uses main's platform-specific Playwright launch test.
The coverage catalog now names that title, and this branch keeps main's
test body so the merge stays clean.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(browser): cover the platform launch title on this branch
The lifecycle catalog names main's platform launch title. This branch
now has that quoted test too, so the check passes before and after the
merge.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(ghidra): record the exact 12.2 masked byte-switch defect
Ghidra 12.2 leaves an AArch64 and-0xff byte-offset switch unrecovered.
Exact 12.2 sessions now cite open upstream issue 10154, and the AArch64
lane expects that jump table to stay unrecovered.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(browser): keep one platform launch environment check
The second launch test repeated the selected environment check. The
remaining test is the parameterized title, and the launcher forwards
SYSTEMROOT on win32 so that assertion holds.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(ghidra): keep the open switch defect on admitted builds
The exact-version check cleared ghidra-10154 on every build except 12.2.
That defect stays on admitted releases until one is known to fix it.
ghidra-9996 still clears after the exact 12.2 build.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* style(docs): format the Ghidra testing-lane table
The AArch64 jump-table row wrapped past the formatter width and failed
static checks.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(ghidra): give the masked switch its full byte range
The unrecovered oracle now uses 256 readable table bytes, and the same
object must still recover its compare-bounded switch.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* docs(website): point Ghidra setup at 12.2 and JDK 25
The public setup pages still told new users to install Ghidra 12.1.4 and
JDK 21, which this release rejects. Changed setup sentences are extracted
and left untranslated so those editions use the English page.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* docs(website): keep catalog edits to the changed setup strings
The catalog check compares active message ids, not a full rewrite.
Restore the previous catalogs and update only the Ghidra 12.2 setup
strings, leaving their translations empty.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(ghidra): apply the closed switch defect from the parsed patch
12.2, 12.2.0, and 12.2.0 with a suffix all precede 12.2.1, so they keep
ghidra-9996. The masked-switch table lists all 256 byte entries.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(ghidra): share the AArch64 switch fixture compiler invocation
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(zip): corroborate local member extent claims
* fix(zip): verify descriptors of contained records
---------
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* fix(setup): write the recommended tool call timeout into dsh registrations
rea setup --client dsh wrote registrations without toolCallTimeoutMs, so
DeepSeek Harness applied its 60-second default while a native provider
startup alone can exceed it. Setup now derives the deadline from the same
provider allowance plus operation headroom used for Kimi and Codex,
preserves a positive caller-selected value, and doctor reports a
registration without a usable timeout as stale.
ProviderRequestBudgetResolver declares an optional budgetSettings read
set, so the fail-closed budget gate rejects unresolved Cordis expressions
only for settings the derivation may read plus the timeout inputs; an
unrelated expression such as GHIDRA_INSTALL_DIR still refreshes, and an
expression in REA_IDA_MCP_CONFIG fails closed.
* test(setup): record the dsh tool call deadline in the NUL repair fixture
The NUL refusal scenario rewrites a valid dsh registration and expects
doctor to report it healthy. That fixture omitted toolCallTimeoutMs,
which doctor now reports stale, so the repaired registration records the
deadline a current dsh registration carries.
* test(setup): separate dsh timeout repair coverage
---------
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* fix(sqlite): refuse one-byte files instead of reporting an empty schema
* test(sqlite): assert refusal facts instead of the full error sentence
The one-byte and corrupt-input acceptance tests froze the sentence that
SqliteDatabaseWorkerExecution builds from fixed wording, SQLite's
diagnostic and the selected path. Share one helper that keeps the code,
issue path, reason and single-issue shape exact, and checks the
diagnostic and selected path separately, so rewording the sentence no
longer breaks either regression.
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(artifacts): classify Gradle version selectors by their grammar
* fix(artifacts): match Gradle ASCII range whitespace
---------
Co-authored-by: ShuaixinHuang <197200901+ShuaixinHuang@users.noreply.github.com>