CoreFoundation reads some XML plists that strict XML rejects, such as
documents with a double hyphen in a comment or a control character in a
string. inspect_plist refused them although plutil -lint accepts them,
and the outcome depended on how file(1) described the bytes.
When REA's XML reader refuses a document's syntax, inspection now takes
the existing plutil conversion route. The result keeps format xml and a
limitation names REA's reason. A document plutil also rejects stays
refused with its diagnostic, and plist structure errors in well-formed
XML are still refused without conversion.
Fixes#12234
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.
* 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(javascript): cover the isolated-world probe outside Electron identity
* fix(javascript): expand Electron call paths only with an Electron reference
* test(javascript): cover Electron reach through resolved names, aliases and worker URLs
* fix(javascript): attribute call-path unknowns only where the walk can reach Electron
* test(javascript): cover a hoisted read that precedes a loader alias
* fix(javascript): keep call-path unknowns for late-initialized loader aliases
* test(javascript): drop a duplicate loader-alias case
* test(javascript): cover loader aliases reached through bindings and calls
* fix(javascript): prove a late initializer loader-free before dropping its unknown
* test(javascript): cover loaders behind default parameters, wrappers and long alias chains
* fix(javascript): resolve identifiers inside late initializers and bound the loader proof
* fix(javascript): bound late loader proofs and preserve input uncertainty
* fix(javascript): fail closed for receiver and mutated loaders
* fix(javascript): close loader-free proof gaps
* fix(javascript): retain uncertain loader sources
* test(javascript): cover remaining loader proof gaps
* fix(javascript): bound alias writes and preserve loader uncertainty
* test(javascript): cover demand-driven loader mutation facts
* perf(javascript): index loader mutation facts on demand
* style(javascript): apply oxfmt to the loader proof changes
* test(javascript): cover inherited loaders and rebound alias writes
* fix(javascript): preserve inherited loaders and rebound aliases
* test(evaluation): update scoped feature coverage oracle
* test(javascript): cover inherited loader proof gaps
* fix(javascript): preserve projected and inherited loaders
* fix(javascript): prove safe parameter member reads
* test(javascript): cover loader proof review gaps
* fix(javascript): close remaining loader proof gaps
* fix(javascript): preserve qualified global replacements
* fix(javascript): preserve spread rest uncertainty
* fix(javascript): retain retried handler uncertainty
* fix(javascript): retain import.meta loader uncertainty
* fix(javascript): detach fresh constructed aliases
* fix(javascript): check inherited members of ambient loaders
* test(javascript): parse call-path gate fixtures as established CommonJS
* fix(javascript): prove projected locals inside late initializers
A function-local binding was accepted by scope alone, so a destructuring
projection such as `const { r } = {}` inside a late initializer never met
the inherited-member check that a top-level projection goes through. Route
projected locals through the binding proof; locals with inline initializers
keep the scope shortcut, since the walk already visits their expressions.
* fix(javascript): keep late results with changed prototypes or import.meta uncertain
The loader-free proof inspected member expressions inside a late function
but not what its caller does with the result. With `Object.prototype.load =
require`, `var load = () => ({})` and `load().load("electron")` reach the
loader through a member selected after the call, and a bare `import.meta`
returned after `import.meta.load = ...` does the same. Check each result
expression: a literal fails when its prototype chain gained members, a bare
`import.meta` when any of its members changed, and a result of unknown shape
when either happened. Promise and generator results inherit from
Object.prototype as well. `SemanticPropertyPathCoverage.touches` answers
whether any retained write or escape sits below a prefix.
* fix(javascript): detach replacements held by fresh bindings
`alias = replacement` with `const replacement = {}` selected an identifier
initializer, so the later `alias.other = 1` stayed a shared write on the
original alias group and the resolver treated `base` as overwritten,
dropping its bridge call path. Resolve the identifier to its binding and
accept it when that binding only ever received one fresh allocation.
Bindings that can hold the original object keep the shared write.
* fix(javascript): charge loader proofs through the wrapper-visit budget
#11319 replaced the wrapper-visit counter with a Budget, so the loader proof charge no longer compiled on current main. Spend it through the shared helper, which records the same limit.
* test(javascript): keep the alias bound off the quadratic write index
The 8,000 by 8,000 alias bound spent its time in the ambient write index, which walks every store to an object for each member write. That cost is the same on main and is not what the bound covers. Reach the same 64 million alias and write pairs with 32,000 aliases and 2,000 writes, which runs in about 2.5 seconds instead of 8.5 on the same host.
Drop the 30 second overrides on both alias bounds too. They halved the hosted CI deadline the source project gives them, and the testing guide asks tests to leave deadlines to the project.
* refactor(javascript): ask the effect log for loader proof effects
Main now guards reads of the effect indexes outside javascriptSemanticEffectLog.ts. The loader proof read them directly in 15 places. Route those reads through the log: the existing ambientPathMayHaveChanged and bindingHasMemberWrites queries, plus three timeless queries for a path or any member below it, an ambient binding write, and an escape at a site other than the call being proved. The answers are unchanged.
* fix(javascript): retain uncertainty for changed ambient values
---------
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.
* fix(process): decode analyzer worker control frames as strict UTF-8
Refs #12145.
* fix(process): decode bounded control input without byte replacement
* fix(android): bound guardian readiness by the first UTF-8 frame (#12176)
Follows the readiness framing and lifecycle findings reported by @LcokHcak in #3048.
* refactor(process): share Linux resource limit admission with EVM
Refs #12128
* fix(desktop): bind process details to their Linux generation
Refs #12134.
* fix(desktop): bind open-file samples to their process generation
Refs #12134.
* fix(desktop): reject split ZIP volume metadata
Refs #12137
* refactor(zip): read Zip64 disk fields through BinaryReader
* fix(desktop): validate ZIP extra fields before selection
Refs #12146
* fix(desktop): keep bounded ZIP pairing compatible with Python 3.9
* fix(desktop): retain exact counts in bounded Zip64 views
* fix(desktop): follow the declared Zip64 directory frame
* fix(desktop): recognize empty ZIP state views (#12163)
* fix(desktop): recognize empty ZIP state views
(cherry picked from commit 263b25bd34)
* fix(desktop): reuse empty ZIP admission for wrapped documents
* refactor(zip): share the local-header signature with desktop state
---------
Co-authored-by: wunianze666-netizen <wunianze666@gmail.com>
* test(zip): inherit the portable ZIP count deadline
---------
Co-authored-by: wunianze666-netizen <wunianze666@gmail.com>
* feat(native): surface inferred prototype instruction conflicts
- Add evidence-backed x86-32 stack-read and stdcall cleanup diagnostics in ReaGhidraBridge.java without rewriting signatures or function bodies.
- Carry prototype conflicts through native API schemas, analysis summaries and retained dossier views so CLI and MCP share the observations.
- Update Ghidra analysis profile digests, documentation and MCP boundary tests for the new diagnostic contract.
* fix(native): follow framed paths and retain zero cleanup conflicts
- Traverse decoded entry-frame paths in ReaGhidraBridge.java instead of stopping at an address-ordered early return, so later reachable reads remain visible.
- Detect explicit, implicit and partial EBP writes and use decoded operands with p-code loads, preventing stale-frame warnings and recognizing real Ghidra stack reads.
- Compare immediate return cleanup with inferred stdcall stack bytes even when either side is zero, and allow zero in the native API contract.
- Add real Ghidra instruction probes and MCP/CI regressions for the reported cases and document the bounded checks.
* fix(native): attribute frame reads before register writes
- Check a decoded stack read before invalidating EBP in ReaGhidraBridge.java so instructions that read and replace the frame still report the conflict.
- Match each candidate operand's address p-code to its LOAD pointer so a POP from ESP into an EBP destination cannot masquerade as an argument read.
- Cover both decoded instruction cases in ReaPrototypeConflictProbe.java and await verifier-owned scratch cleanup on success and failure.
* fix(ghidra): recompute raw profile digests after rebase
- Recompute the 6502 and disabled-analyzer raw-binary profile digests in GhidraRawProfileDigest.test.ts so the pinned values commit to the merged shared policy on current main.
* fix(native): locate the entry frame by address and own verifier cleanup
- Find PUSH EBP; MOV EBP, ESP from the function entry address in ReaGhidraBridge.java so an owned block below the entry cannot hide a reachable argument read.
- Run analyzeHeadless in its own process group and stop that group before removing scratch in verify-ghidra-prototype-conflicts.mjs, so the JVM and decompiler children are addressed as one owned family.
- Cover the lower-owned-block case in ReaPrototypeConflictProbe.java with an entry frame preceded by a decoded lower block.
* fix(native): size prototype fixture and own Ghidra cleanup
- Allocate the instruction fixture through the lower-block offset in verify-ghidra-prototype-conflicts.mjs so the entry frame and decoded block exist for the probe.
- Check the detached process group instead of the leader exit code before cleaning up, so lingering decompiler children receive SIGTERM and SIGKILL before scratch removal.
* fix(native): own Ghidra probe cleanup with shared process owner
- Use REA's ProviderProcessSupervisor and token-verified cleanup in verify-ghidra-prototype-conflicts.mjs so the JVM and descendants are stopped before scratch removal, including failed launches.
- Preserve the source-owned x86 fixture with an entry frame and lower block while asserting verified cleanup.
- Build the pinned verifier dependency graph in the Ghidra bridge CI job and extend its timeout before running the real probe.
* fix(native): end verifier waits on interrupt before cleanup
- Install SIGINT and SIGTERM handlers before spawning Ghidra and abort the wait, so Ctrl-C no longer leaves the detached process group running.
- Keep cleanup handlers installed until the shared process owner reports verified cleanup, then remove scratch only after the process family is stopped.
* fix(native): preserve verifier cleanup interruption errors
- Keep interrupt handlers installed through process cleanup and workspace deletion, then check cancellation before reporting success.
- Preserve and report the failing cleanup step instead of replacing it with an interruption or deleting a tree that may still be unsafe.
* fix(native): retain probe failures across verifier cleanup
- Record the probe error before cleanup so a late SIGINT or SIGTERM cannot replace its stack and captured Ghidra diagnostics.
- Prefer the earlier probe failure when cleanup also fails, while preserving cleanup diagnostics and temporary state for diagnosis.
* test(kuna): raise caller matrix verifier budget for parallel CI
- The multi-target verifier spawn used a 10s timeout while sibling verifier cases use 25s.
- Under 8-way sharded CI load this intermittently exceeded 10s and produced ETIMEDOUT, failing shard 1/8 and portable tests.
- Raise to 30s to match the existing verifier budget and avoid load flakes.
* ci: retrigger checks after transient Ghidra SDK download 403
- The ghidra_bridge lane failed in upstream's untouched SDK acquisition step: curl returned 403 with 0 bytes downloaded in 0.046s from the GitHub release CDN.
- Push an empty commit to start a fresh run because rerunning on the upstream repository requires admin rights.
* fix(native): keep the cleanup lease on combined verifier failures
- Run the prototype-conflict probe through withVerifierFixture so a probe failure and a failed release surface as one AggregateError that keeps the probe error as its cause and attaches the retained fixture as cleanupLeases, instead of throwing the probe error alone and dropping the cleanup owner.
- Report the failed cleanup step and retained project only when the fixture was not released, and keep interrupt handlers installed until release finishes.
* test(native): use shared MCP prototype test deadlines
---------
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* feat(ghidra): admit plain MIPS32 PIC shared images
Add an elf32-dyn-o32-arch32r2-standard-v1 lane for ET_DYN images without
an ASE flag (shared libraries and ET_DYN executables), beside the MIPS16
lane. It reuses the complete dynamic-relocation inspection and admits
R_MIPS_NONE, symbol-bearing R_MIPS_REL32 and the TLS loader slots
DTPMOD32/DTPREL32/TPREL32, which stay unapplied with a limitation.
Ghidra's default moves an image linked at 0 to 0x10000 and then shifts
only nonzero local GOT entries, so page-zero GOT entries (how GNU-ld
libraries call static functions) keep unshifted values and calls through
them decompile to wrong addresses. The lane imports at the image's own
ELF image base instead (lowest PT_LOAD in the first 20 headers, as Ghidra
computes it), commits it in the analysis profile and requires the bridge
handshake to report it. Images with a prelink trailer are refused.
verify:ghidra:mips32-shared checks LE, BE and soft-float LLD fixtures laid
out like GNU ld output against an independent readelf oracle: every
REL32, local GOT entry (including a zero page entry), global GOT entry
(EXTERNAL or lazy-binding stub) and unapplied TLS slot, plus named calls
in the decompilation, snapshot round trip and CLI evidence.
Refs #2338
* style(ghidra): format link-base regression scenarios
* test(ghidra): require zero-valued local GOT coverage
* test(ghidra): cover TPREL32 in the shared lane and drop duplicated session suites
Add an initial-exec TLS variable so the MIPS32 shared fixture declares
R_MIPS_TLS_TPREL32 beside DTPMOD32/DTPREL32, and require all three slots
to stay UNSUPPORTED with unchanged bytes. Remove the soft-float and
mutation handshake suites that the main reconciliation duplicated in
GhidraSessionValues.test.ts.
* test(ghidra): read the section table in the MIPS32 shared-library oracle
parseMipsReadelf now confirms the fixture's section types from readelf -S
(main added -S to the executable, relocatable and PSP oracles). Request it in
the shared-library oracle too.
* test(ghidra): follow main's verifier contracts in the MIPS32 shared lane
Main made the Ghidra installation probe asynchronous behind the shared
inspectVerifierGhidraInstallation helper, and analyze_function now returns a
summary unless detail "complete" is requested. Use both, as the other MIPS
lanes do.
* refactor(ghidra): keep Ghidra's image-base rule out of shared ELF parsing
The shared MIPS ELF reader computed Ghidra's ElfHeader.getImageBase rule
(lowest PT_LOAD address among the first 20 program headers) and the domain
type described it as Ghidra's base. Record source facts instead: every
PT_LOAD virtual address with its program-header index, plus the Android
prelink trailer. ghidraMipsElfImageBase in GhidraMipsProfile now applies
Ghidra's rule to those facts. The committed elf_image_base is unchanged.
---------
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
* fix(desktop): honor env working directories in launch commands
* fix(desktop): create the replay working directory before launch
---------
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
trace_dylib_resolution Evidence named the slice the session happened to
open, not what was traced. On Apple silicon, `rea trace-dylib-resolution
/bin/ls --architecture x86_64` returned a subject selecting the arm64e
slice (offset 65536, architecture "arm64") while its parameters and
roots said x86_64. Whole-file traces of universal images and app bundle
traces also carried that host slice.
Root cause: the artifact provider built the execution without a
subject, and bindExecutionTarget binds such an execution, like any
subject that names the target's own bytes without a selection, to the
open target's selection. A provider had no way to state that an
observation covers the whole file.
EvidenceSubjectTarget.selection now accepts null, meaning the
observation selects nothing even though a session opened one slice.
bindExecutionTarget already inherited a selection only when none was
stated; its comment now says so. The trace states selection: null. This
matches inspect-binary-hardening and the other whole-file inspectors,
which already record no selection for a universal image.
Validation (macOS arm64, Node 24.22.1):
- New CLI boundary regression: a universal arm64 + x86_64 tool, traced
whole and with --architecture x86_64, has a subject with no
architecture or selection and the expected roots. Before this change
the subject carried the opened slice.
- dylibResolutionCli and dylibResolutionMcp: 44 passed, 2 skipped.
- Suites that exercise subject binding (binarySession,
analysisEvidenceAdapterParity, artifactMcp, bundleResourceRootCliMcp,
appleBundleTraversalCoverage, AppBundleResourceRoot,
ArtifactComparisonEvidence): 144 passed, 1 skipped.
- evidence, BinarySession and ArtifactProvider unit tests: 186 passed.
- Real /bin/ls trace with --architecture x86_64: subject selection and
architecture null, roots [x86_64]. Format, typecheck, lint and
architecture guards pass.
Not changed: inspect-macho, inspect-signature and
inspect-native-dispatch-metadata also record the opened slice for a
universal image; whether each examines only that slice was not checked.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
* feat(ghidra): admit RISC-V position-independent executables (#4487)
Admit little-endian RISC-V ELF32 and ELF64 position-independent executables
(ET_DYN with PT_INTERP) into the pinned generic Ghidra lane.
- Collect `PT_INTERP` during the existing validated ELF table walk in
`validateElfDeclaredRanges` and record `hasInterpreter` in target metadata,
supporting arbitrary table offsets and extended program counts without duplicate scanning.
- Admit ET_DYN images when `hasInterpreter` is present under lane identities
`elf32-pie-le-double-float-generic-v1` and `elf64-pie-le-double-float-generic-v1`.
- Preserve refusal for non-PIE ET_DYN targets (shared libraries without PT_INTERP)
explaining that shared libraries require separate load verification.
- Record load model facts (`load_model: "pie"`, `expected_default_image_base: "0x10000"`
or `"0x100000"`, `has_interpreter: true`) in `riscv_elf` analysis profile parameters.
- Record default image base analysis coordinates limitation for PIE targets.
- Update `docs/ghidra-riscv.md` with PIE admission boundary and evidence semantics.
- Add unit tests for PT_INTERP collection, high-offset program headers, and PIE profile resolution.
- Add PIE fixture generation capability to `tests/conformance/ghidra/riscv-fixture.mjs` and
pass `-Wl,-pie` to ensure bare-metal drivers generate ET_DYN.
- Extend `scripts/verify-real-ghidra-riscv.mjs` to test both ET_EXEC and ET_DYN.
Fixes#4487
* fix(ghidra): check structured facts and rethrow typed errors in PIE tests
* style: format GhidraRiscvProfile.test.ts
* fix(ghidra): restrict PIE admission to verified double-float RISC-V lanes