Commit Graph
7707 Commits
Author SHA1 Message Date
morluto bcb068cc04 fix(rizin): refuse a start after shutdown began as an owner-closed cancellation (#11041)
OwnedDebuggerSessionRegistry owns the rule: closeStarts refuses later starts
as owner-closed (a caller that already aborted keeps its own reason) and
aborts in-flight ones. Rizin no longer reports an inferior_tty input error
for a start made after closeAll began; GDB's local copy of the rule is gone.

User-visible: start_rizin_debug_session called while REA is shutting down
now fails as a cancellation (owner closed) instead of an input error naming
inferior_tty, matching start_gdb_session.
2026-11-15 19:34:18 +08:00
morluto 423489a761 feat(evidence): name the side and issue path of comparison source integrity failures (#11049)
* feat(process): name the side and issue path of process-capture integrity failures

compare_process_captures reported which side failed validation, and where,
only in the message. EvidenceIntegrityError now carries an optional side
and issue path, projected as details.side and details.issue_path, and the
CLI's malformed-capture issue carries observed.issue_path. The first issue
path comes from the same selection describeValidationFailure names.

* feat(evidence): report source authentication issue paths from the Evidence owner

readEvidence and readImmutableEvidence now set the record-relative issue
path on their integrity failure, and readEvidencePayload sets it when the
caller says where the payload sits. compare_functions passes its side, so a
tampered function source reports details.side and details.issue_path like
compare_process_captures. Process capture result issue paths are now
relative to the Evidence record (normalized_result.*) on both adapters.
2026-11-15 19:33:47 +08:00
morluto 7f2141b568 refactor(ghidra): take the host from HOST_PLATFORM instead of reading process.platform (#11054) 2026-11-15 19:33:25 +08:00
morluto c0dd084376 refactor(verify): give seven verifier directories' fixture roots to VerifierFixture (#11055)
* refactor(verify): give the android, electron, firmware, ghidra-seed, harmony, inspector and managed verifiers' roots to VerifierFixture

Each root is removed only after the MCP clients, transports, owned processes
and file handles that may still use it confirm cleanup; a failed cleanup
retains and names the root. The verifier-fixture-root baseline shrinks from
90 files / 95 sites to 80 / 85.

* fix(verify): keep nested Electron and inspector children owned by the fixture, and plain roots where Windows creates directories

The Electron cancellation CLI and application, and each inspector target,
now register their stop with the fixture (onceConfirmed keeps the early stop),
so the root is removed only after they confirm. The android and inspector
verifiers keep a plain directory as before: tools create directories beneath
the root, which a private root blocks on Windows.
2026-11-15 19:33:03 +08:00
morluto 3bbdec3fc8 test(artifacts): compare ASAR block-integrity limitations with their owning constant (#11061)
Two tests matched three fragments of one limitation sentence, twice for the
same row. They now compare the row with BLOCK_INTEGRITY_LIMITATION from
AsarListing, beside the existing declared-integrity and hash-status facts.
2026-11-15 19:32:40 +08:00
morluto b3cba0d762 test(process): check each capture provider field once instead of per side (#11067)
The provider admission matrix ran each of the three provider fields for
both sides through CLI and MCP. Alternating sides keeps every field and
both side labels, and removing the version check or the side label still
fails a remaining row.
2026-11-15 19:31:44 +08:00
morluto 1151763582 fix(contracts): remove the duplicate advertisedOutputJsonSchema declaration (#11057)
#11005 left two exported declarations of advertisedOutputJsonSchema in
src/contracts/toolSchemaMetadata.ts, so main no longer typechecks
(TS2451). The second declaration takes an optional target that defaults
to the first one's fixed draft-2020-12, so it covers every caller. The
first is removed and its documentation moves onto the remaining one.
2026-11-15 06:21:39 -05:00
Ysopwnableandgensokyo-dev 08420be3bd fix(ghidra): explain Program reuse after startup timeouts
Explain Program reuse in cold-start timeout recovery and distinguish query
snapshots from saved Ghidra analysis. Preserve explicit startup-mode selection,
first-analysis requirements, and custom-analyzer choices.

Refs #10365

Co-authored-by: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
2026-11-15 19:14:41 +08:00
morluto 7181e7f515 perf(mcp): advertise large output union alternatives as shared definitions (#11005)
* perf(mcp): advertise large output union alternatives as shared definitions

inspect_analysis_view's advertised output schema held every view result inline in one union definition, which the MCP SDK's default validator compiled into one function: about 1.7 s CPU and 500 MiB peak. Large union alternatives now become definitions of their own and identical definitions are merged, at the owning output projection. Tests that check results against the advertised output contract now compile that schema instead of a separate inline projection, which overflowed the main-thread stack under forked workers.

* perf(tests): run source test projects in forked workers (#11016)

Thread workers of the source lane share one process, which grew to 3.8-5.0 GiB; forked workers keep each heap separate so the largest process stays near 1.2-1.8 GiB.

* test(mcp): select the analysis view contract from its family
2026-11-15 06:13:21 -05:00
morluto 2826592759 refactor(application): build application cancellations with a required reason (#10574)
* refactor(application): build application cancellations with a required reason

Migrate every AnalysisCancelledError construction under src/application
to cancelledBy and lower the guard baseline from 671 to 442.

Shutdown refusals in BinarySession report OwnerCancellation.ownerClosed
when the caller did not abort, instead of an undefined reason.
ReverseEngineeringService and DesktopBuildOwnedWork close their work
with the typed owner reason. SharedStartupAttempt returns the departing
caller's reason with a cancelled wait. One shared isAborted predicate
replaces eleven local copies and narrows a checked signal for
cancelledBy.

* refactor(desktop): drop an unused cancellation import from desktop capture state
2026-11-15 18:38:24 +08:00
morluto 0ddc732c69 docs: describe omitted output schemas as the default in README translations (#10658)
* test(mcp): check advertised schemas against a portable dialect in every profile

* fix(mcp): re-emit advertised patterns in a portable regex grammar

* fix(mcp): present discover_tools input through the selected profile

The hot-cold catalog registered discover_tools directly with the SDK, so
the default compact input profile still advertised its canonical schema,
including page-size defaults clients apply on their own (#5322). Route
every listed input schema, transport discovery included, through one
server-owned presentation, and pass the selected presentation to the
deferred server, which serves recorded tool schemas but not discovery.

* fix(mcp): keep the Evidence union on described borrowed inputs

borrowEvidenceInput installed its advertised projection on one schema
instance. A call site's .describe() rebuilds the schema from its
definition, so match_native_functions advertised left and right as
unconstrained values in every input profile. Install the projection in
the borrowed parser's constructor so derived schemas keep it, for both
borrowed inputs and borrowed arrays.

* fix(mcp): validate advertised dialect by selected profile

* fix(mcp): preserve schema widening through references

* feat(mcp): project compact input schemas into the portable advertised dialect

One projection (advertisedDialectProjection.ts) owns the compact input
advertisement: one type per node, items on every array, string literals
as enums, oneOf widened to anyOf, allOf merged, tuples as element unions,
exclusive bounds made inclusive, and every keyword outside the dialect
dropped. Each rewrite is exact or widening; server validation is
unchanged. The root description and sentences for dropped cross-field
constraints move to the tool description, which Codex and Zed keep.

The compact baseline falls to nullable values only (29), which the
dialect cannot express without narrowing an input. The gateway shape
repairs (typeArrayItemUnions, hasGatewayIncompatibleShape), the second
dependentRequired-to-prose copy and the strict-gateway test denylist are
removed; the dialect checker gates the inline presentation instead.

* feat(mcp): omit output schemas from tools/list by default

No surveyed model API reads an output schema, MCP client validators
repeat REA's own result validation, and complete output schemas were
about 92% of the default tools/list. REA_MCP_OUTPUT_SCHEMA_PROFILE now
resolves to off through one owner (resolveMcpOutputSchemaProfile);
compact and full remain selectable. structuredContent is still
delivered, and the deferred server serves both off and full from the
recorded catalog with the matching recorded identity.

Tests and artifact verifiers that validate results against complete
advertised output contracts select full explicitly.

* docs: describe omitted output schemas as the default in README translations

* fix(mcp): select schemas in verification workflows

* fix(mcp): refuse zero-padded repeat bounds and record host-specific path patterns

* fix(mcp): keep projected schemas permissive

* fix(mcp): keep references valid when presentation widens their targets

* fix(mcp): keep a single union branch's closure scoped to its properties

* fix(mcp): select complete output schemas in the Electron verifier

* fix(mcp): select complete output schemas in standalone verifier probes

* fix(mcp): keep an allOf member's tuple tail with its own positions

* fix(mcp): relax references into negations and conditionals presentation may remove

* test(mcp): select the borrowed Evidence contract from its family

* fix(mcp): keep astral literals and dynamic references widening in every mode

* fix(mcp): omit astral literals and relax untracked references when presentation widens

* fix(mcp): select complete output schemas in every verifier transport

* fix(mcp): read reference pointers by schema position, not segment spelling

* fix(mcp): protect references into collapsing boolean unions and keep referencing branches out of the collapse

* fix(mcp): keep reduced schemas open when their root admits names by pattern

* test(mcp): keep the reduced-schema pattern case outside the size-limited block

* fix(mcp): omit patterns whose classes name astral characters

* fix(mcp): decide union collapse and pointer protection from canonical branches

* docs(mcp): document non-u astral matching and pin BMP agreement across V8 modes

* fix(mcp): keep prototype-named properties as own properties in the dialect projection

* test(mcp): select full output schemas in packet and signal schema checks

* test(mcp): expect the advertised base64 schema to omit the unbounded pad-bit probe

* fix(mcp): protect references into unevaluated keywords lost evaluation removes

* fix(mcp): validate the doctor probe reply against the canonical output contract (#10655)

* fix(mcp): validate the doctor probe reply against the canonical output contract

* test(mcp): replay doctor probe validation under every output profile

* test(mcp): accept typed-branch nullable unions in the advertised dialect (#10649)

* test(mcp): select full output schemas in HTTP session workflows

* fix(mcp): emit stack-safe classes with non-ASCII members and protect draft-07 reference siblings

* fix(mcp): protect references into literal siblings of relaxed draft-07 references

* feat(mcp): omit output schemas from tools/list by default (#10646)

* feat(mcp): omit output schemas from tools/list by default

No surveyed model API reads an output schema, MCP client validators
repeat REA's own result validation, and complete output schemas were
about 92% of the default tools/list. REA_MCP_OUTPUT_SCHEMA_PROFILE now
resolves to off through one owner (resolveMcpOutputSchemaProfile);
compact and full remain selectable. structuredContent is still
delivered, and the deferred server serves both off and full from the
recorded catalog with the matching recorded identity.

Tests and artifact verifiers that validate results against complete
advertised output contracts select full explicitly.

* fix(mcp): select schemas in verification workflows

* fix(mcp): select complete output schemas in the Electron verifier

* fix(mcp): select complete output schemas in standalone verifier probes

* fix(mcp): select complete output schemas in every verifier transport

* test(mcp): select full output schemas in packet and signal schema checks

* fix(mcp): validate the doctor probe reply against the canonical output contract (#10655)

* fix(mcp): validate the doctor probe reply against the canonical output contract

* test(mcp): replay doctor probe validation under every output profile

* test(mcp): accept typed-branch nullable unions in the advertised dialect (#10649)

* test(mcp): select full output schemas in HTTP session workflows

* fix(mcp): keep the Evidence union on described borrowed inputs (#10597)

* fix(mcp): keep the Evidence union on described borrowed inputs

borrowEvidenceInput installed its advertised projection on one schema
instance. A call site's .describe() rebuilds the schema from its
definition, so match_native_functions advertised left and right as
unconstrained values in every input profile. Install the projection in
the borrowed parser's constructor so derived schemas keep it, for both
borrowed inputs and borrowed arrays.

* test(mcp): select the borrowed Evidence contract from its family

* docs(mcp): admit baseline growth only for restored contract declarations

* fix(mcp): state reduced multi-type properties as typed branches, not type lists

The reduced fallback derived a property's types and emitted them as a type
list, which Gemini, Moonshot and Ark reject. A property whose canonical
schema is anyOf string and number therefore traded a typed union for a
rejected construct when the inline projection fell back. Emit one type, or
anyOf typed branches, so [T, null] becomes the dialect's nullable union.
2026-11-15 18:37:55 +08:00
morluto af5c89a857 refactor(desktop): release retained screen-judge cleanup through its lease (#10809)
* refactor(errors): read cleanup completion from dependencies instead of pushing it to dependents

AnalysisError.cleanupIncomplete becomes a getter over the error's own unresolved state and its bound dependencies. The WeakRef dependents graph and resolveCleanup's push propagation are removed, and subclasses mark their initial state through markCleanupIncomplete instead of overriding a public mutable field.

* fix(errors): settle bound dependencies before binding more

A composite whose dependencies had released but whose status was never read
stayed incomplete when a later dependency was bound.

* refactor(errors): let errors read a lease group's release instead of resolving them (#10759)

AnalysisCleanupLeaseGroup no longer calls resolveCleanup() on its error.
The error holds the group as a release claim and reads cleanupReleased when
asked, so completion follows the lease owner without a push.

* chore(verify): count resolveCleanup and bindCleanupDependencies calls so they only shrink (#10763)

* refactor(errors): let errors read a lease group's release instead of resolving them

AnalysisCleanupLeaseGroup no longer calls resolveCleanup() on its error.
The error holds the group as a release claim and reads cleanupReleased when
asked, so completion follows the lease owner without a push.

* chore(verify): count resolveCleanup and bindCleanupDependencies calls so they only shrink

* refactor(process): let worker cleanup errors read their root epoch's release

OwnedWorkerRuntime's root epoch is now an AnalysisCleanupReleaseClaim. Errors bound to it read its release when asked instead of the epoch pushing resolveCleanup() through weak references, and a runtime without a retained root binds an already-released owner. Removes the OwnedWorkerCleanupFailure structural type and four resolveCleanup() pushes; the counted baseline shrinks accordingly.

* refactor(pcap): let pcap cleanup errors read command and snapshot release (#10792)

Retained pcap commands, retained snapshot failures and snapshot completion descriptors now own their release and errors read it: commands and snapshot failures implement AnalysisCleanupReleaseClaim, and a failed completion-descriptor close is retried through an AnalysisCleanupLease bound to its error. Removes four resolveCleanup() pushes; the counted baseline shrinks accordingly.

* refactor(process): let Darwin helper cleanup errors read their owner's release

Darwin helper cache and run-token reader root cleanup now record an AnalysisCleanupRelease that every failure reports; a successful retry confirms it. Replaces the first-failure anchor that later failures depended on and the resolveCleanup() push. Removes two resolveCleanup() and two bindCleanupDependencies() calls; both counted baselines shrink.

* refactor(ghidra): let Ghidra cleanup errors read the client's and checkpoint's release

GhidraClient and GhidraCheckpoint bind each cleanup failure to an AnalysisCleanupRelease and confirm it when a later close or retry releases everything, instead of keeping a set of failures and pushing resolveCleanup() to each. Removes two resolveCleanup() calls; the counted baseline shrinks.

* refactor(process): let debugger session cleanup errors read a confirmed stop

OwnedDebuggerSessionRegistry binds each retained gdb or rizin cleanup failure to the session's AnalysisCleanupRelease and confirms it when a stop is confirmed, keeping the latest failure for retry reporting, instead of pushing resolveCleanup() to a set of failures. Removes one resolveCleanup() call; the counted baseline shrinks.

* refactor(process): drop the debugger registry's unused closeOperation option

Returning the latest retained failure removed the last reader of closeOperation; remove the option and the values gdb and rizin supplied.

* refactor(server): let environment shutdown errors read a successful close

serveEnvironment binds each shutdown failure to an AnalysisCleanupRelease that a later successful close confirms, instead of pushing resolveCleanup() to a set of failures. The state-file retry test now also requires the earlier report to read the release. Removes one resolveCleanup() call; the counted baseline shrinks.

* refactor(desktop): release retained screen-judge cleanup through its lease

Each retained judge or screenshot cleanup owner is wrapped in an AnalysisCleanupLease bound to its failure; close() releases the lease, which records the release the failure reads, instead of closing the owner and pushing resolveCleanup(). Removes one resolveCleanup() call; the counted baseline shrinks.
2026-11-15 18:37:39 +08:00
morluto d61de7b1a7 refactor(captures): classify TShark, sigrok and pcap failures through the shared provider classifier (#10820)
* fix(pcap): report a refused tool launch as an unavailable tool

When a capture read's tool launch was refused with EPERM, pcap reported it
as unreadable tool output (AnalysisProtocolError) instead of an unavailable
tool. Pcap also never told "not executable" apart from "missing".

Add classifyLaunchFailure in src/process as the one owner of which spawn
errnos mean missing (ENOENT, ENOTDIR) and which mean present but not
executable (EACCES, EPERM, ENOEXEC, EISDIR). Pcap, apktool and gdb used three
different errno lists; all three now use it.

* refactor(process): classify owned command outcomes through one owner

Add ProviderOutcome, read once from the supervisor's facts, and classifyProviderOutcome, the one mapping from an outcome to a typed error. Migrate pcap and the shared artifact producer command failure (TShark, sigrok). Add the rea/provider-failure-classifier lint rule with a shrink-only baseline of per-site (cause: unknown) => AnalysisError functions.

* refactor(process): classify Java and adb launch refusals with the shared owner (#10598)

The Ghidra and JADX Java probes and adb's version probe now take
missing-versus-not-executable from classifyLaunchFailure instead of their
own errno checks. ELOOP and ENAMETOOLONG join the owner's missing set, so
adb keeps its existing classification and both Java probes now treat a
non-directory path component as a missing runtime.

* fix(pcap): preserve command launch attribution

* test(pcap): expect an unknown failing stage when a limited capture tool exits unsuccessfully

* refactor(process): classify owned command outcomes through one owner

Add ProviderOutcome, read once from the supervisor's facts, and classifyProviderOutcome, the one mapping from an outcome to a typed error. Migrate pcap and the shared artifact producer command failure (TShark, sigrok). Add the rea/provider-failure-classifier lint rule with a shrink-only baseline of per-site (cause: unknown) => AnalysisError functions.

* test(lint): cover baselined files in the provider-failure classifier rule

* fix(apktool): report the decoder JVM's heap exhaustion as memory, not a target refusal

The supervisor now watches for a managed runtime's out-of-memory report while output streams (JVM java.lang.OutOfMemoryError, V8 'JavaScript heap out of memory'), so a retention budget cannot hide it. ProviderOutcome gains a runtime-exhausted kind and classifyProviderOutcome maps it once. Apktool decode opts in and moves to the shared classifier; Ghidra's single-purpose output token becomes the JVM runtime option and its area matching reuses the shared table.

* fix(process): keep the runtime's out-of-memory report on the memory error it supports

* fix(pcap): attribute a limited capture's exit through the prlimit wrapper

* fix(process): keep a shortened out-of-memory line marked as shortened

* fix(process): identify the refused executable for every launcher shape

Name the executable on errnos Node throws synchronously from spawn
(ENOEXEC, ENOTDIR, ...) and read the Windows native launcher's
executable-launch constraint, so pcap attributes those refusals to the
selected tool or limiter instead of an unknown stage.

* test(pcap): prove synchronous launch naming with ENOTDIR on every POSIX host

* fix(apktool): advise the JVM limit that governs the reported memory area

* fix(ghidra): read the supervisor's parsed OOM area instead of reparsing lines

* fix(sqlite,capture): read V8 heap exhaustion from streamed output, not retained stderr (#10724)

The SQLite worker and the Node capture decoders now declare runtime: "v8", so the supervisor observes V8's heap report as output streams. Both classifiers read the runtime-exhausted outcome instead of matching retained stderr, so a report beyond the retention budget, or one whose process then overflowed its output budget, is still reported as memory exhaustion.

* refactor(evm): classify EVMole worker outcomes through the shared provider classifier (#10716)

* refactor(evm): classify EVMole worker outcomes through the shared provider classifier

The EVMole worker's cancellation, deadline, output budget and CPU/file-size signals now go through classifyProviderOutcome; only the prlimit launch refusal and the reserved EFBIG exit remain in its dialect. The worker is opted into V8 exhaustion scanning so a heap report names its fixed 256 MiB heap. providerOutcomeOf now reads a launch refusal only from a spawn errno, so a private-storage or carrier read errno is no longer reported as a limiter launch failure.

* fix(evm): keep a stream failure's exit status and update Linux worker expectations

* refactor(process): read launch refusals through the launcher identity owner

* refactor(wasm): classify WABT command outcomes through the shared provider classifier (#10727)

* refactor(wasm): classify WABT command outcomes through the shared provider classifier

WABT's cancellation, command deadline, output budget, and inherited-limit signals now go through classifyProviderOutcome; its dialect keeps only the configuration-probe and launch-refusal unavailability and the unclassified-exit output error. A tool ended by SIGXCPU/SIGXFSZ or the memory limit is now a resource constraint instead of an output error, and a refused launch in the producer phase is reported as unavailable. The shared timeout context gains the analysis setting that selected the deadline, and the output-limit context may name its exceeded budget.

* fix(wasm): keep setup guidance when a WABT version probe's output cannot be collected

* refactor(providers): classify raw provider failures through one owner

* refactor(process): build classified cancellations with cancelledBy and the carried caller reason

* chore(lint): count runOwned's close-failure mapper from main in the classifier baseline

* fix(process): carry the runtime's exhaustion report on every memory error that claims it

SQLite and historical-capture memory errors recognised a streamed V8 heap report but dropped it, so a report outside retained output reached callers without its source line or omission flags. One projection, runtimeExhaustionReport, now owns that shape for the shared classifier and both consumers. Also lowers the EVMole and WABT cancellation-constructor baselines.

* fix(process): keep the provider's recovery for a native-thread exhaustion report

The shared classifier replaced the provider-supplied recovery with generic capacity advice whenever the JVM reported it could not create a native thread, dropping Apktool's launcher guidance and flipping retryable to true. The provider's recovery now wins; the capacity advice remains the fallback for providers that supply none.

* refactor(pwntools): classify pwntools failures through the shared provider classifier (#10782)

* refactor(pwntools): classify pwntools failures through the shared provider classifier

pwntools' per-site classifier is replaced by pwntoolsFailureDialect for classifyProviderFailure. Command outcomes go through classifyProviderOutcome, launch refusals through launcherRefusalOf instead of a 'spawn ' syscall prefix check, the input limit through the file dialect's fileSize, and the cleanup report through commandCleanup. The shared read cancellation now carries the reader's partial observation, and resource-signal reports accept provider detail.

* fix(pwntools): report a failed command's output once and expect resource_constraint for overflow

The shared outcome owner already attaches the command's output as capturedOutput, so the pwntools adapter fallback no longer copies the same text into diagnostics.stdout, diagnostics.stderr and diagnostics.captured_output; diagnostics keep the exit facts. Update the overflow expectations the dialect migration missed (layout provider cleanup and over-budget rows, pwntools cleanup) to resource_constraint.

* refactor(wasm): leave WABT cancellation to the shared owners

WabtFailures matched any Error named AbortError as a cancellation. No WABT path produces one: reads, path checks and owned commands report cancellation as ArtifactReaderFailure or OwnedCommandFailure, which classifyProviderFailure and classifyProviderOutcome already classify with the caller's reason. The provider's two signal checks now build the cancellation with cancelledBy. Lowers the AnalysisCancelledError constructor baseline for both WABT files to zero.

* fix(wasm): report a cancelled snapshot write as the caller's cancellation

The private snapshot write passed the caller's signal to Node's writeFile, whose abort envelope reached the dialect's unclassified output failure once the name-based AbortError guess was removed. The write now turns the envelope carrying its own signal's reason into cancelledBy at the producer.

* refactor(firmware): classify firmware failures through the shared provider classifier (#10810)

Firmware producers now report their own deadline expiry: the start check, the output inventory and launcher admission throw the phase timeout where they observe the abort, and a read the phase signal cancelled reports it through the dialect's interruption. The classifier no longer rereads a cancellation's reason as a timeout. A new readCleanup hook keeps firmware's cleanup report shape.

* refactor(captures): classify TShark, sigrok and pcap failures through the shared provider classifier

TsharkCaptureProvider and SigrokRecordingProvider replace #failure with a dialect, and PcapProvider replaces analysisFailure and the tool-identity classifier. ArtifactAnalysisResources exposes its command context and a readCleanupFailure composer so the readCleanup hook keeps each provider's cleanup report shape.

* chore(guards): shrink the cancellation construction baseline after merging the base
2026-11-15 18:37:06 +08:00
morluto 557bf41c2c refactor(macho): select slices through one owner with typed refusals (#10874)
* refactor(macho): name load-command codes through the shared LC table

* refactor(macho): name version-min and string commands through the shared tables

* refactor(parsers): report XML, web artifact and Mach-O failures through ParseFailure (#10760)

* refactor(parsers): report XML, web artifact and Mach-O failures through ParseFailure

* refactor(xml): type XML parse failures as malformed or unsupported only

* refactor(parsers): catch per-format document and source failure classes (#10794)

* refactor(parsers): report apktool strings and XML plist source failures through ParseFailure

* refactor(apktool): drop the unused ParseFailureError import

* refactor(native): select ELF sections through the keyed owner

* test(native): pin returned exports and versions for repeated version sections

* refactor(nativeaot): read neither of repeated frozen object regions (#10790)

* refactor(nativeaot): read neither of repeated frozen object regions through the keyed owner

* refactor(parsers): retire every first-match section lookup (#10793)

* fix(nativeaot): report an ambiguous frozen object region as partial (#10812)

* refactor(native): select program headers and load commands through the keyed owner (#10827)

* refactor(native): select the GOT base and dependency segments through the keyed owner (#10847)

* refactor(native): select the GOT base and dependency segments through the keyed owner

resolveGotBase selects the SHT_DYNAMIC section and the allocated .got.plt/.got sections with uniqueKeyed, so a repeat leaves the GOT base unknown instead of taking the first match; elfDependencies reads PT_LOAD segments through the same keyed program table as PT_DYNAMIC. A new keyed-first-match guard recognizes a find/findLast or filter-then-[0] whose predicate compares a key field, whatever the table is called; existing matches are baselined. The unwind fixture can declare PT_DYNAMIC headers and repeated .dynamic/.got.plt sections.

* test(native): pin the PT_DYNAMIC path and destructured keyed lookups

The keyed-first-match guard moves to scripts/verify/keyed-first-match.mjs with focused tests, and now recognizes a destructured key in any position and a destructured filter predicate. The single PT_DYNAMIC unwind test declares .dynamic as SHT_PROGBITS so only the program header supplies DT_PLTGOT, and asserts the resolved ranges.

* fix(macho): refuse repeated chained-fixup and dyld-info tables (#10850)

parseMachoLayout kept the last LC_DYLD_CHAINED_FIXUPS range and merged the rebase/bind streams of every LC_DYLD_INFO(_ONLY) command. A repeat now refuses the slice as malformed. NativeMachoInspection's chained import-name reader already declined repeated LC_DYLD_CHAINED_FIXUPS; that rule is now covered by a test.

* refactor(parsers): retire the Mach-O, source-map, NDJSON and managed-graph failure classes (#10818)

* refactor(parsers): retire the Mach-O, source-map, NDJSON and managed-graph failure classes

* refactor(plist): report property-list and keyed-archive failures through ParseFailure (#10822)

* refactor(plist): report property-list and keyed-archive failures through ParseFailure

* refactor(kuna): report Kuna document failures through ParseFailure (#10824)

* refactor(kuna): report Kuna document failures through ParseFailure

* test(kuna): pin the limit and malformed facts of Kuna document failures

* refactor(plist): report plist decode budgets through ParseFailure (#10842)

* refactor(plist): report plist decode budgets through ParseFailure

The plist preflight owner throws a ParseFailure limit (plist-decode-bytes) for its byte budget, the shared PlistNestingLimitError for nesting depth, and a malformed failure for a reference cycle, instead of a per-format PropertyListDecodeBudgetFailure whose meaning each caller re-mapped through a fail callback. The fail option, PropertyListDecodeBudgetFailure and interfaceBuilderDecodeBudgetFailure are removed; the parser-format-failure-class baseline reaches zero.

* refactor(plist): report keyed-archive allowances as typed limit failures (#10844)

keyedArchive.ts and decodeKeyedArchiveBytes throw decodeLimit parse failures (keyed-archive-bytes, -object-table-entries, -values-per-object, -reference-edges) and malformed parse failures instead of plain RangeError/TypeError. KeyedArchiveReader no longer classifies every RangeError as a limit: only a limit parse failure or a host stack/capacity limit is one. The unreachable 128-level scan nesting check is removed; projectPlistValue already refuses the archive first.

* test(plist): expect the nesting limit as its ParseFailure carrier

* fix(verify): count a filter index only on the filter's own result

* refactor(macho): select slices through one owner with typed refusals

* refactor(keyed): select FIT properties and certificate extensions through uniqueKeyed

* refactor(keyed): key each FIT node once, pin repeated FIT properties and subjectKeyIdentifiers, and type the unknown Mach-O header as unsupported

* refactor(plist): report plist and NIBArchive input faults through ParseFailure and defects through invariantViolation (#10895)

* refactor(plist): report plist and NIBArchive input faults through ParseFailure and defects through invariantViolation

* fix(plist): report a cut-short NIB header or plist length record as truncated

* fix(native): project the selected Mach-O's parse failures through the shared owner

nativeMachoSourcePrimary mapped every non-limit ParseFailureError to
AnalysisOutputError, so a malformed or truncated selected Mach-O was
reported as unreadable tool output and an unsupported one lost its
category. It now uses parseFailureAnalysisError like the ELF, Go and
other byte-format providers: malformed and truncated bytes are an
invalid path, unsupported bytes an unsupported target, and a limit
still falls through to its producer's budget.
2026-11-15 18:36:30 +08:00
morluto 54230c2e7b refactor(setup): require the platform for skill destinations instead of reading the host (#11006)
* refactor(setup): require the platform for skill destinations instead of reading the host

The setup skill owner defaulted seven entry points to process.platform. Every
production caller (setup host, doctor, uninstall, update maintenance) already
passes the platform it was composed with, so the defaults only let a caller
silently mix the host REA runs on with the one it asked about. The platform is
now a required parameter; tests name the platform they exercise, the setup
fixture passes its declared host, and the ambient-read baseline drops
SetupSkill.ts (7 to 0).

* test(setup): expect the fake host's POSIX skill paths in planning tests
2026-11-15 18:35:01 +08:00
morluto 482a98a95b perf(test): run CI change-scope cases concurrently in workspace-owned repositories (#11009)
* perf(test): run CI change-scope cases concurrently in workspace-owned repositories

* test(ci): own the change-scope seed workspace from creation so failed setup is removed
2026-11-15 18:34:39 +08:00
morluto 53f1b626e6 refactor(verify): give the Apple verifiers' fixture roots to VerifierFixture (#11024)
Seven Apple verifiers allocated their roots with mkdtemp and removed them in a
finally (or a package workspace close). They now open a VerifierFixture and
release it through withVerifierFixture or the package workspace, so a failed
removal is retained and reported instead of hidden. The verifier-fixture-root
baseline shrinks from 97 files / 102 sites to 90 / 95.
2026-11-15 18:34:17 +08:00
morluto f46cd6770c test: advance the Rizin and build computer fixed waits instead of sleeping through them (#11026) 2026-11-15 18:33:52 +08:00
morluto 5cfaac0834 refactor(process): require the platform for environment snapshots instead of reading the host (#11028) 2026-11-15 18:33:23 +08:00
morluto 4126698407 feat(mcp): project compact input schemas into the portable advertised dialect (#10632)
* test(mcp): check advertised schemas against a portable dialect in every profile

* fix(mcp): re-emit advertised patterns in a portable regex grammar

* fix(mcp): present discover_tools input through the selected profile

The hot-cold catalog registered discover_tools directly with the SDK, so
the default compact input profile still advertised its canonical schema,
including page-size defaults clients apply on their own (#5322). Route
every listed input schema, transport discovery included, through one
server-owned presentation, and pass the selected presentation to the
deferred server, which serves recorded tool schemas but not discovery.

* fix(mcp): keep the Evidence union on described borrowed inputs

borrowEvidenceInput installed its advertised projection on one schema
instance. A call site's .describe() rebuilds the schema from its
definition, so match_native_functions advertised left and right as
unconstrained values in every input profile. Install the projection in
the borrowed parser's constructor so derived schemas keep it, for both
borrowed inputs and borrowed arrays.

* fix(mcp): validate advertised dialect by selected profile

* fix(mcp): preserve schema widening through references

* feat(mcp): project compact input schemas into the portable advertised dialect

One projection (advertisedDialectProjection.ts) owns the compact input
advertisement: one type per node, items on every array, string literals
as enums, oneOf widened to anyOf, allOf merged, tuples as element unions,
exclusive bounds made inclusive, and every keyword outside the dialect
dropped. Each rewrite is exact or widening; server validation is
unchanged. The root description and sentences for dropped cross-field
constraints move to the tool description, which Codex and Zed keep.

The compact baseline falls to nullable values only (29), which the
dialect cannot express without narrowing an input. The gateway shape
repairs (typeArrayItemUnions, hasGatewayIncompatibleShape), the second
dependentRequired-to-prose copy and the strict-gateway test denylist are
removed; the dialect checker gates the inline presentation instead.

* fix(mcp): refuse zero-padded repeat bounds and record host-specific path patterns

* fix(mcp): keep projected schemas permissive

* fix(mcp): keep references valid when presentation widens their targets

* fix(mcp): keep a single union branch's closure scoped to its properties

* fix(mcp): keep an allOf member's tuple tail with its own positions

* fix(mcp): relax references into negations and conditionals presentation may remove

* test(mcp): select the borrowed Evidence contract from its family

* fix(mcp): keep astral literals and dynamic references widening in every mode

* fix(mcp): omit astral literals and relax untracked references when presentation widens

* fix(mcp): read reference pointers by schema position, not segment spelling

* fix(mcp): protect references into collapsing boolean unions and keep referencing branches out of the collapse

* fix(mcp): keep reduced schemas open when their root admits names by pattern

* test(mcp): keep the reduced-schema pattern case outside the size-limited block

* fix(mcp): omit patterns whose classes name astral characters

* fix(mcp): decide union collapse and pointer protection from canonical branches

* docs(mcp): document non-u astral matching and pin BMP agreement across V8 modes

* fix(mcp): keep prototype-named properties as own properties in the dialect projection

* test(mcp): expect the advertised base64 schema to omit the unbounded pad-bit probe

* fix(mcp): protect references into unevaluated keywords lost evaluation removes

* fix(mcp): emit stack-safe classes with non-ASCII members and protect draft-07 reference siblings

* fix(mcp): protect references into literal siblings of relaxed draft-07 references

* feat(mcp): omit output schemas from tools/list by default (#10646)

* feat(mcp): omit output schemas from tools/list by default

No surveyed model API reads an output schema, MCP client validators
repeat REA's own result validation, and complete output schemas were
about 92% of the default tools/list. REA_MCP_OUTPUT_SCHEMA_PROFILE now
resolves to off through one owner (resolveMcpOutputSchemaProfile);
compact and full remain selectable. structuredContent is still
delivered, and the deferred server serves both off and full from the
recorded catalog with the matching recorded identity.

Tests and artifact verifiers that validate results against complete
advertised output contracts select full explicitly.

* fix(mcp): select schemas in verification workflows

* fix(mcp): select complete output schemas in the Electron verifier

* fix(mcp): select complete output schemas in standalone verifier probes

* fix(mcp): select complete output schemas in every verifier transport

* test(mcp): select full output schemas in packet and signal schema checks

* fix(mcp): validate the doctor probe reply against the canonical output contract (#10655)

* fix(mcp): validate the doctor probe reply against the canonical output contract

* test(mcp): replay doctor probe validation under every output profile

* test(mcp): accept typed-branch nullable unions in the advertised dialect (#10649)

* test(mcp): select full output schemas in HTTP session workflows

* fix(mcp): keep the Evidence union on described borrowed inputs (#10597)

* fix(mcp): keep the Evidence union on described borrowed inputs

borrowEvidenceInput installed its advertised projection on one schema
instance. A call site's .describe() rebuilds the schema from its
definition, so match_native_functions advertised left and right as
unconstrained values in every input profile. Install the projection in
the borrowed parser's constructor so derived schemas keep it, for both
borrowed inputs and borrowed arrays.

* test(mcp): select the borrowed Evidence contract from its family

* docs(mcp): admit baseline growth only for restored contract declarations

* fix(mcp): state reduced multi-type properties as typed branches, not type lists

The reduced fallback derived a property's types and emitted them as a type
list, which Gemini, Moonshot and Ark reject. A property whose canonical
schema is anyOf string and number therefore traded a typed union for a
rejected construct when the inline projection fell back. Emit one type, or
anyOf typed branches, so [T, null] becomes the dialect's nullable union.
2026-11-15 18:26:13 +08:00
morluto fc6a9604a1 refactor(cli): project JSON-document workflow commands from their contracts (#10668)
* refactor(cli): project JSON-document workflow commands from their contracts

* test(cli): include WASM interface and native matching documents in the JSON-input guard
2026-11-15 18:25:34 +08:00
morluto da8f366596 fix(mcp): register fx and MCPorter where they read their configuration (#10698)
* fix(mcp): register fx and MCPorter where they read their configuration

fx reads $FX_HOME/mcp.json when FX_HOME is set, and MCPorter reads
MCPORTER_CONFIG exclusively or prefers an absolute XDG_CONFIG_HOME file.
rea mcp add always wrote ~/.fx/mcp.json and ~/.mcporter/mcporter.json,
which those clients then ignored. Resolve both from the clients' own
rules, and refuse an FX_HOME fx itself refuses.

* fix(mcp): detect MCPorter by its selected file and refuse FX_HOME aliases

Address review findings: MCPORTER_CONFIG now also names the detection
marker, so default registration includes MCPorter when only that file
exists, and is resolved with path.resolve as MCPorter does. FX_HOME with
. or .. segments is refused, so the private profile mode cannot reach
the home folder through an alias. Override tests use the host's path
rules for real temporary paths.

* fix(mcp): never apply fx's private folder mode to the home folder

A linked parent in FX_HOME can name the caller's home folder after the textual root and home checks pass. The private profile writer now compares the opened folder's identity with the caller's home folder and refuses before changing modes or publishing.

* feat(setup): let doctor, update and uninstall see MCPorter registrations (#10702)

* feat(setup): let doctor, update and uninstall see MCPorter registrations

MCPorter becomes a client-table row, so a registration rea mcp add or
rea setup --client mcporter wrote is inspected, refreshed and removed like
other clients. MCPorter already imports setup-managed clients'
registrations, so the row is applicable only through REA's own
registration: setup never selects it from detection and doctor reports
it only when its file holds a rea row or the caller selects it. A
relative MCPORTER_CONFIG is refused because MCPorter resolves it per
command directory.

* fix(setup): detect MCPorter by the folder of the file it loads

With an absolute XDG_CONFIG_HOME and only a legacy ~/.mcporter file, the row returned the legacy file but the absent XDG folder as its detection evidence, so default registration and setup skipped MCPorter.

* fix(setup): keep an unresolved client unknown without blocking others' maintenance

A client whose configuration location cannot be resolved (for example a relative MCPORTER_CONFIG) is reported as unknown, never missing or absent. Update maintenance leaves each unresolved client out of its scope and names it with its own recovery, while still maintaining the other clients; uninstall still stops, because it cannot tell whether that client's file holds REA's row.

* docs(setup): restore MCPorter's release note and client-table row after rebase

* fix(update): keep unknown clients in every failed maintenance plan
2026-11-15 18:25:02 +08:00
morluto 78fd365bdd refactor(macho): name load-command codes through the shared LC table (#10714)
* refactor(macho): name load-command codes through the shared LC table

* refactor(macho): name version-min and string commands through the shared tables

* refactor(parsers): report XML, web artifact and Mach-O failures through ParseFailure (#10760)

* refactor(parsers): report XML, web artifact and Mach-O failures through ParseFailure

* refactor(xml): type XML parse failures as malformed or unsupported only

* refactor(parsers): catch per-format document and source failure classes (#10794)

* refactor(parsers): report apktool strings and XML plist source failures through ParseFailure

* refactor(apktool): drop the unused ParseFailureError import

* refactor(parsers): retire the Mach-O, source-map, NDJSON and managed-graph failure classes (#10818)

* refactor(parsers): retire the Mach-O, source-map, NDJSON and managed-graph failure classes

* refactor(plist): report property-list and keyed-archive failures through ParseFailure (#10822)

* refactor(plist): report property-list and keyed-archive failures through ParseFailure

* refactor(kuna): report Kuna document failures through ParseFailure (#10824)

* refactor(kuna): report Kuna document failures through ParseFailure

* test(kuna): pin the limit and malformed facts of Kuna document failures

* refactor(plist): report plist decode budgets through ParseFailure (#10842)

* refactor(plist): report plist decode budgets through ParseFailure

The plist preflight owner throws a ParseFailure limit (plist-decode-bytes) for its byte budget, the shared PlistNestingLimitError for nesting depth, and a malformed failure for a reference cycle, instead of a per-format PropertyListDecodeBudgetFailure whose meaning each caller re-mapped through a fail callback. The fail option, PropertyListDecodeBudgetFailure and interfaceBuilderDecodeBudgetFailure are removed; the parser-format-failure-class baseline reaches zero.

* refactor(plist): report keyed-archive allowances as typed limit failures (#10844)

keyedArchive.ts and decodeKeyedArchiveBytes throw decodeLimit parse failures (keyed-archive-bytes, -object-table-entries, -values-per-object, -reference-edges) and malformed parse failures instead of plain RangeError/TypeError. KeyedArchiveReader no longer classifies every RangeError as a limit: only a limit parse failure or a host stack/capacity limit is one. The unreachable 128-level scan nesting check is removed; projectPlistValue already refuses the archive first.

* test(plist): expect the nesting limit as its ParseFailure carrier

* refactor(plist): report plist and NIBArchive input faults through ParseFailure and defects through invariantViolation (#10895)

* refactor(plist): report plist and NIBArchive input faults through ParseFailure and defects through invariantViolation

* fix(plist): report a cut-short NIB header or plist length record as truncated

* fix(native): project the selected Mach-O's parse failures through the shared owner

nativeMachoSourcePrimary mapped every non-limit ParseFailureError to
AnalysisOutputError, so a malformed or truncated selected Mach-O was
reported as unreadable tool output and an unsupported one lost its
category. It now uses parseFailureAnalysisError like the ELF, Go and
other byte-format providers: malformed and truncated bytes are an
invalid path, unsupported bytes an unsupported target, and a limit
still falls through to its producer's budget.
2026-11-15 18:24:35 +08:00
morluto 422cff00ab refactor(process): give POSIX PATH search order one owner (#10891)
* refactor(process): give POSIX PATH search order one owner

Program lookup, the packet-tool locator and Jadx's Java lookup each split
PATH and resolved empty and relative entries themselves. They now take the
candidates from posixPathCandidates in src/process/executableSearch and keep
only their own probe and failure reporting. A counted architecture guard
keeps new PATH splits from returning; the remaining Windows and Ghidra
searches are baselined.

* refactor(process): move the colon-split PATH searches onto the search owner

REA's launcher-alias lookup and Ghidra's macOS Java lookup split PATH on
":" and skipped relative entries themselves; they now use
posixPathCandidates with relativeEntries: "skip", and the guard counts
colon splits of PATH as well.

* fix(process): produce PATH candidates lazily so an unused entry cannot fail an earlier match

* refactor(process): give Windows PATH/PATHEXT search order the executable search owner (#10916)

* refactor(process): give Windows PATH/PATHEXT search order the executable search owner

* fix(process): keep later PATH entries searchable after a stray quote or a removed working directory

* fix(reverse): refuse a Windows objdump or Rizin command that PATH does not find (#10953)

* fix(reverse): refuse an objdump or Rizin command Windows PATH does not find instead of launching the bare name

* fix(reverse): report a Windows lookup refusal after cancellation and to capability discovery

* fix(desktop): search a Windows host's PATH as Windows does when doctor looks up a program

programOnPath sent every host through the POSIX PATH search, so on Windows a
relative PATH entry reached absoluteCallerPath with platform linux and the
Windows working directory, which resolved against itself until the stack
overflowed. programOnPath now searches PATH and PATHEXT through
windowsPathCandidates on a Windows host, and absoluteCallerPath refuses a
non-POSIX working directory with a TypeError instead of recursing.
2026-11-15 18:24:04 +08:00
morluto 659772952d feat(javascript)!: require JavaScript application and source-map fields that were optional only for historical Evidence (#10958)
* feat(managed)!: require managed Evidence fields that were defaulted only for historical Evidence

Member inspections, comparisons, verification results and application graphs
require every field their producers write. nesting_read is boolean and the
bare-name comparison mode for historical observations is deleted;
signature_resolution loses type-names-v1. The unavailable-metadata member
result now writes assembly: null instead of relying on a default.

BREAKING CHANGE: managed records missing these fields are refused as
malformed; recreate the observation.

* docs(managed): drop the remaining historical-omission claims for member Evidence

* docs(process): describe process capture omissions by their current meaning

raw_data is omitted when it equals data and residual_unknowns when a failure
retained none; neither omission describes historical Evidence.

* feat(javascript)!: require JavaScript application and source-map fields that were optional only for historical Evidence

skipped_javascript_files, protocol_registrations and the original-source and
module-edge source_index/declaration_index links are always written by their
producers and are now required.

BREAKING CHANGE: results missing these fields are refused as malformed;
recreate the observation.

* test(browser): refuse original-source rows without their source or declaration index
2026-11-15 18:23:32 +08:00
morluto 536c54ca1f feat(process): name whether a process input file failed to read or to parse (#11008) 2026-11-15 18:23:05 +08:00
morluto e0249db65c feat(evidence): mark a snapshot supplied to import_evidence_bundle with detected_kind (#11010) 2026-11-15 18:22:33 +08:00
morluto e793fd6808 feat(mcp): name the next operation when a completed result overflows the response budget (#11015)
* feat(mcp): name the next operation when a completed result overflows the response budget

* test(mcp): assert next_operation for every overflow recovery route
2026-11-15 18:22:08 +08:00
morluto b572dbaf25 feat(setup): mark a shared enable:false registration as a manual conflict in doctor status (#11019) 2026-11-15 18:21:40 +08:00
morluto e9f9d8b880 fix(rizin): abort in-flight debug session starts when the manager closes (#11020)
OwnedDebuggerSessionRegistry now owns in-flight starts: it relays the
caller's abort, aborts each start as owner-closed on close and waits for it,
and reports the caller's reason when the caller aborted first. GDB's private
start map and Rizin's counter-and-waiters drain are removed; Rizin's closeAll
previously waited for a start without aborting it.
2026-11-15 18:21:09 +08:00
morluto 4aefe2daec fix(errors): name rea doctor only for providers it can check (#10565)
* fix(errors): name rea doctor only for providers it can check

An unclassified ProviderAdapterError, an unreadable result and the final
presentation fallback all told the caller to run rea doctor, although doctor
checks only hopper, ghidra, ida, jadx, kuna and desktop_build. The doctor
scope list moves into the domain so presentation can consult it: covered
providers get `rea doctor --provider ID`, others are told plainly that the
failure is unclassified and where its reason is reported.

Producers whose actionable reason lived only in details.diagnostics now
author it: Ghidra's committed-profile conflicts share one owner that says to
reopen the target, closed analysis clients share one owner, and a jadx-reported
request failure no longer points at provider readiness. The deep-analysis
profile rejection names doctor only for doctor-checkable providers. Dead
presentation branches for tags handled earlier are removed.

* fix(errors): make provider recovery actionable

* test(android): remove brittle recovery wording check

* test: wait for observed child exit before cancellation

* test(errors): check that doctor is not offered instead of matching recovery prose
2026-11-15 18:09:40 +08:00
morluto 428e006f66 fix(filesystem): close file handles only through the owned-handle owners (#10578)
* fix(filesystem): close file handles only through the owned-handle owners

Three sites still closed a fs/promises FileHandle in a finally block, where a
failed close replaced the operation's own failure or result: the agentenv
runner head read, the Linux cgroup kernel-file read and the JavaScript
analysis transfer writer. They now use withOwnedFileHandle, which retains the
operation outcome alongside a failed close.

A lint rule, rea/no-raw-file-handle-close, refuses close() on a handle opened
from fs/promises or declared as FileHandle, and fs close/closeSync on a
descriptor, in src outside src/filesystem/Owned*.

* fix(lint): distinguish owned handle class fields

* fix(filesystem): retain diagnostic file cleanup owners

* test(hopper): retry after confirmed application exit

* fix(filesystem): settle doctor peers, sample cgroup outside admission, and bind lint this

Desktop doctor waits for every client inspection before reporting a retained close failure. Ghidra memory cgroup sampling no longer passes the resource scope's admission gate, so a pending owner cannot mask the provider failure; failed closes are still retained for session cleanup. The raw FileHandle close rule resolves this through arrow functions only, treats static blocks as static, and keeps #name distinct from constructor parameter properties.

* fix(ghidra): join in-flight cgroup samples before session cleanup

ArtifactResourceScope.track runs optional work outside the admission gate while cleanup drains still join it, so a cancelled request's final retry cannot finish before a sample retains a failed close.

* fix(filesystem): join work tracked during a cleanup drain and report doctor file cleanup as typed errors

A cleanup retry now repeats until no tracked work is active and every retained owner has had one attempt, so diagnostic work admitted while an earlier owner is releasing is joined. Doctor reports a retained runner or configuration file close as a ProviderCleanupError bound to its lease instead of rethrowing a raw file failure, so the error document keeps the retained file.

* fix(desktop): read the agentenv runner head through the artifact file owner

The runner read retained a failed close under one lease while the reported failure carried another, so a successful cleanup retry left the doctor error reporting an already-closed file. withArtifactFileHandle binds one lease to both.
2026-11-15 18:08:56 +08:00
morluto babfc87ec0 fix(mcp): present discover_tools input through the selected profile (#10592)
* test(mcp): check advertised schemas against a portable dialect in every profile

* fix(mcp): re-emit advertised patterns in a portable regex grammar

* fix(mcp): present discover_tools input through the selected profile

The hot-cold catalog registered discover_tools directly with the SDK, so
the default compact input profile still advertised its canonical schema,
including page-size defaults clients apply on their own (#5322). Route
every listed input schema, transport discovery included, through one
server-owned presentation, and pass the selected presentation to the
deferred server, which serves recorded tool schemas but not discovery.

* fix(mcp): validate advertised dialect by selected profile

* fix(mcp): preserve schema widening through references

* fix(mcp): refuse zero-padded repeat bounds and record host-specific path patterns

* fix(mcp): keep references valid when presentation widens their targets

* fix(mcp): relax references into negations and conditionals presentation may remove

* fix(mcp): keep astral literals and dynamic references widening in every mode

* fix(mcp): omit astral literals and relax untracked references when presentation widens

* fix(mcp): read reference pointers by schema position, not segment spelling

* fix(mcp): protect references into collapsing boolean unions and keep referencing branches out of the collapse

* fix(mcp): omit patterns whose classes name astral characters

* fix(mcp): decide union collapse and pointer protection from canonical branches

* docs(mcp): document non-u astral matching and pin BMP agreement across V8 modes

* test(mcp): expect the advertised base64 schema to omit the unbounded pad-bit probe

* fix(mcp): protect references into unevaluated keywords lost evaluation removes

* fix(mcp): emit stack-safe classes with non-ASCII members and protect draft-07 reference siblings

* fix(mcp): protect references into literal siblings of relaxed draft-07 references

* fix(mcp): keep the Evidence union on described borrowed inputs (#10597)

* fix(mcp): keep the Evidence union on described borrowed inputs

borrowEvidenceInput installed its advertised projection on one schema
instance. A call site's .describe() rebuilds the schema from its
definition, so match_native_functions advertised left and right as
unconstrained values in every input profile. Install the projection in
the borrowed parser's constructor so derived schemas keep it, for both
borrowed inputs and borrowed arrays.

* test(mcp): select the borrowed Evidence contract from its family

* docs(mcp): admit baseline growth only for restored contract declarations
2026-11-15 18:08:12 +08:00
morluto 57f75bb723 refactor(native): build debugger and native cancellations with a required reason (#10599)
* refactor(errors): require a reason for every analysis cancellation

Add cancelledBy, the one constructor for AnalysisCancelledError. Its
reason argument is required: an aborted signal, an earlier cancellation,
a reason carried from an inner failure, or an OwnerCancellation when REA
closed the work's owner itself.

Closed adb, apktool, firmware, sqlite, gdb, web-bundle analyzer and
binary-session record owners now report OwnerCancellation.ownerClosed
instead of an absent reason. Firmware's predecessor wait returns its
cancellation as a Result, removing an unreachable reason-less fallback.

An architecture guard rejects new direct constructions, with a counted
per-file baseline that can only shrink.

* refactor(errors): share one aborted-signal predicate with cancelledBy

* refactor(process): return a departing caller's reason from a shared startup wait

* refactor(providers): build decompiler-provider cancellations with a required reason (#10587)

Migrate every AnalysisCancelledError construction under src/ghidra,
src/ida, src/cutter, src/jeb and src/kuna to cancelledBy and lower the
guard baseline from 671 to 572. Closed Kuna target commands report
OwnerCancellation.ownerClosed instead of an absent reason, and Kuna and
JEB close abort their lifetimes with that typed reason.

* fix(errors): read the reason of any signal-shaped cancellation source

* refactor(native): build debugger and native cancellations with a required reason

Migrate every AnalysisCancelledError construction under src/native,
src/gdb and src/rizin to cancelledBy and delete Rizin's local
rizinCancellation constructor; its reason choice becomes a
CancellationReason. GDB close aborts active startups with
OwnerCancellation.ownerClosed.

* test(ci): cover the counted-baseline comparison behind the architecture guard

* refactor(native): drop an unused cancellation import from native UI observation
2026-11-15 18:07:27 +08:00
morluto f6752bbe2c refactor(process): let debugger session cleanup errors read a confirmed stop (#10805)
* refactor(errors): read cleanup completion from dependencies instead of pushing it to dependents

AnalysisError.cleanupIncomplete becomes a getter over the error's own unresolved state and its bound dependencies. The WeakRef dependents graph and resolveCleanup's push propagation are removed, and subclasses mark their initial state through markCleanupIncomplete instead of overriding a public mutable field.

* fix(errors): settle bound dependencies before binding more

A composite whose dependencies had released but whose status was never read
stayed incomplete when a later dependency was bound.

* refactor(errors): let errors read a lease group's release instead of resolving them (#10759)

AnalysisCleanupLeaseGroup no longer calls resolveCleanup() on its error.
The error holds the group as a release claim and reads cleanupReleased when
asked, so completion follows the lease owner without a push.

* chore(verify): count resolveCleanup and bindCleanupDependencies calls so they only shrink (#10763)

* refactor(errors): let errors read a lease group's release instead of resolving them

AnalysisCleanupLeaseGroup no longer calls resolveCleanup() on its error.
The error holds the group as a release claim and reads cleanupReleased when
asked, so completion follows the lease owner without a push.

* chore(verify): count resolveCleanup and bindCleanupDependencies calls so they only shrink

* refactor(process): let worker cleanup errors read their root epoch's release

OwnedWorkerRuntime's root epoch is now an AnalysisCleanupReleaseClaim. Errors bound to it read its release when asked instead of the epoch pushing resolveCleanup() through weak references, and a runtime without a retained root binds an already-released owner. Removes the OwnedWorkerCleanupFailure structural type and four resolveCleanup() pushes; the counted baseline shrinks accordingly.

* refactor(pcap): let pcap cleanup errors read command and snapshot release (#10792)

Retained pcap commands, retained snapshot failures and snapshot completion descriptors now own their release and errors read it: commands and snapshot failures implement AnalysisCleanupReleaseClaim, and a failed completion-descriptor close is retried through an AnalysisCleanupLease bound to its error. Removes four resolveCleanup() pushes; the counted baseline shrinks accordingly.

* refactor(process): let Darwin helper cleanup errors read their owner's release

Darwin helper cache and run-token reader root cleanup now record an AnalysisCleanupRelease that every failure reports; a successful retry confirms it. Replaces the first-failure anchor that later failures depended on and the resolveCleanup() push. Removes two resolveCleanup() and two bindCleanupDependencies() calls; both counted baselines shrink.

* refactor(ghidra): let Ghidra cleanup errors read the client's and checkpoint's release

GhidraClient and GhidraCheckpoint bind each cleanup failure to an AnalysisCleanupRelease and confirm it when a later close or retry releases everything, instead of keeping a set of failures and pushing resolveCleanup() to each. Removes two resolveCleanup() calls; the counted baseline shrinks.

* refactor(process): let debugger session cleanup errors read a confirmed stop

OwnedDebuggerSessionRegistry binds each retained gdb or rizin cleanup failure to the session's AnalysisCleanupRelease and confirms it when a stop is confirmed, keeping the latest failure for retry reporting, instead of pushing resolveCleanup() to a set of failures. Removes one resolveCleanup() call; the counted baseline shrinks.

* refactor(process): drop the debugger registry's unused closeOperation option

Returning the latest retained failure removed the last reader of closeOperation; remove the option and the values gdb and rizin supplied.

* refactor(server): let environment shutdown errors read a successful close (#10807)

serveEnvironment binds each shutdown failure to an AnalysisCleanupRelease that a later successful close confirms, instead of pushing resolveCleanup() to a set of failures. The state-file retry test now also requires the earlier report to read the release. Removes one resolveCleanup() call; the counted baseline shrinks.
2026-11-15 18:05:34 +08:00
morluto 06e57d40a4 docs(process): describe process capture omissions by their current meaning (#10956)
* feat(managed)!: require managed Evidence fields that were defaulted only for historical Evidence

Member inspections, comparisons, verification results and application graphs
require every field their producers write. nesting_read is boolean and the
bare-name comparison mode for historical observations is deleted;
signature_resolution loses type-names-v1. The unavailable-metadata member
result now writes assembly: null instead of relying on a default.

BREAKING CHANGE: managed records missing these fields are refused as
malformed; recreate the observation.

* docs(managed): drop the remaining historical-omission claims for member Evidence

* docs(process): describe process capture omissions by their current meaning

raw_data is omitted when it equals data and residual_unknowns when a failure
retained none; neither omission describes historical Evidence.
2026-11-15 18:04:34 +08:00
morluto bf074b8c06 test(process): assert regex deadline and launcher grace outcomes instead of elapsed-time bounds (#10957)
* test(process): assert regex deadline and launcher grace outcomes instead of elapsed-time bounds

* test(process): hold the launcher grace open to prove stop waits for it

* test(process): release and join the held launcher grace before asserting
2026-11-15 18:03:52 +08:00
morluto 80755a9d16 fix(doctor): refuse Ghidra on Windows when REA_GHIDRA_PROJECT_PATH selects an existing project (#10981)
* fix(doctor): refuse Ghidra on Windows when REA_GHIDRA_PROJECT_PATH selects an existing project

Runtime refused an existing project on Windows before admission, but doctor
did not read REA_GHIDRA_PROJECT_PATH, so it reported Ghidra ready for a
registration the provider then refused. The existing-project host check is
now part of Ghidra's admission owner; doctor reads the setting from the
client's registration and reports a failed existing_project check, and the
provider projects the same refusal. An unresolved project path makes doctor's
Ghidra answer unknown only on Windows, the one host it affects.

* test(ghidra): read the existing-project setting through doctor's system inspection
2026-11-15 18:03:20 +08:00
morluto add15d65b3 refactor(verify): give verifier fixture roots one lease on src/process (#10988)
* refactor(verify): give verifier fixture roots one lease on src/process

* fix(verify): keep fixture registration closed, carry the lease on the caught failure, share cleanups once
2026-11-15 18:02:28 +08:00
morluto 83b3bc530a test: replace pcap, web-worker and Rizin sleeps with ordering barriers (#10994)
* test: replace pcap, web-worker and Rizin sleeps with ordering barriers

* test(pcap): join the held command before asserting its deadline ordering
2026-11-15 18:01:43 +08:00
morluto 27b732e957 refactor(kuna): decide Kuna readiness once for doctor and runtime (#10997)
Runtime availability admitted Kuna from its settings and a runnable command,
while doctor also required a passing version probe and cache inspection, so
doctor could report Kuna unavailable while runtime admitted it.

admitKunaCommand is now the one readiness decision: provider availability,
the installation inspection and doctor's installation check all project it.
Doctor reports the version probe and the cache inspection as separate health
checks (version, cache) beside admission; they no longer decide availability,
and runtime still does not pay for them. Setup continues to list Kuna as a
ready provider only when every check passes.
2026-11-15 18:01:05 +08:00
morluto acbbbaa346 refactor(browser): bind a capture decoder's reported cleanup to its process lease (#10998) 2026-11-15 18:00:09 +08:00
morluto 5775c8124f test(desktop): inject the build computer's drawn-screenshot wait so its MCP test runs in 2 s (#10999) 2026-11-15 17:59:26 +08:00
morluto 05bbb73046 test(filesystem): assert process input recovery and target constraints instead of guidance prose (#11000) 2026-11-15 17:58:39 +08:00
morluto 4a13082efc test(mcp): assert snapshot-import and unknown-receipt facts instead of guidance prose (#11004) 2026-11-15 17:58:12 +08:00
morluto b3cdbb92ee refactor(browser): build browser cancellations with a required reason (#10581)
* refactor(errors): require a reason for every analysis cancellation

Add cancelledBy, the one constructor for AnalysisCancelledError. Its
reason argument is required: an aborted signal, an earlier cancellation,
a reason carried from an inner failure, or an OwnerCancellation when REA
closed the work's owner itself.

Closed adb, apktool, firmware, sqlite, gdb, web-bundle analyzer and
binary-session record owners now report OwnerCancellation.ownerClosed
instead of an absent reason. Firmware's predecessor wait returns its
cancellation as a Result, removing an unreachable reason-less fallback.

An architecture guard rejects new direct constructions, with a counted
per-file baseline that can only shrink.

* refactor(errors): share one aborted-signal predicate with cancelledBy

* refactor(process): return a departing caller's reason from a shared startup wait

* refactor(browser): build browser cancellations with a required reason

Migrate every AnalysisCancelledError construction under src/browser,
src/javascript, src/electron, src/inspector and src/captures to
cancelledBy and lower the guard baseline from 671 to 572. Service-worker
setup aborts with OwnerCancellation.ownerClosed on close.

* refactor(providers): build decompiler-provider cancellations with a required reason (#10587)

Migrate every AnalysisCancelledError construction under src/ghidra,
src/ida, src/cutter, src/jeb and src/kuna to cancelledBy and lower the
guard baseline from 671 to 572. Closed Kuna target commands report
OwnerCancellation.ownerClosed instead of an absent reason, and Kuna and
JEB close abort their lifetimes with that typed reason.

* fix(errors): read the reason of any signal-shaped cancellation source

* test(ci): cover the counted-baseline comparison behind the architecture guard

* refactor(electron): drop an unused error import from the fuse failure mapper
2026-11-15 17:46:20 +08:00
morluto a761efda70 refactor(cli): project WASM inspection and firmware commands from their contracts (#10648)
* refactor(cli): project WASM inspection and firmware commands from their contracts

The contract command description gains lists of caller paths and optional object fields flattened into one option per member. inspect-wasm-artifact, inspect-firmware-regions and extract-firmware are derived from their contracts. extract-firmware now records the caller's output_directory spelling in Evidence, like every other caller-path field.

* refactor(cli): project Apple bundle commands onto a declared session target (#10656)

Session-scoped operations take their target from the MCP session, so the CLI's target positional is not a contract field. The contract command description declares a session target: the CLI takes it as the first positional, opens it through runProviderAnalysis as the hand-written commands did, and passes the rest of the contract input. Input issues reported after argv parsing name renamed fields by their CLI key. inspect-keyed-archive, trace-dylib-resolution, inspect-asset-catalog and decode-interface-builder are the pilot; trace-dylib-resolution --root now admits only the roots its contract admits.
2026-11-15 17:45:33 +08:00
morluto 26520bd290 refactor(providers): finalize apktool and shared artifact resources through runOwned (#10653)
* refactor(providers): finalize apktool and shared artifact resources through runOwned

runOwned moves to the provider layer so provider sites can use it without depending on application workflows. It gains a completed-value projection (partial) for values that are not themselves partial observations, an optional signal, and cleanupLeases that bind a close failure to the exact leases whose confirmed release resolves it.

ArtifactAnalysisResources.finish is removed; TShark and sigrok finalize through runOwned with the shared release as close. Apktool's hand-written release, cancellation and lease-group composition is replaced by runOwned. A failed primary now keeps a cancellation observed during final cleanup as a secondary failure, as the documented precedence requires.

A counted architecture guard refuses new hand-written completed-result abort reclassification outside runOwned.

* refactor(adb): finalize runtime-root work through runOwned (#10660)

* refactor(adb): finalize runtime-root work through runOwned

The ADB runtime-root finalizer re-derived the precedence by hand and kept its own cancellation-with-cleanup combiner. It now runs through runOwned: the work body maps snapshot failures, close releases the root lease and binds its own cleanup error, and partial projects the completed execution. A cancellation whose cleanup fails is now the shared cleanup composite with the cancellation as its primary, as runOwned produces for every other owner.

* refactor(providers): keep completed values as diagnostics in runOwned and migrate four finalizers (#10691)

* refactor(providers): keep completed values as diagnostics in runOwned and migrate signature, Ghidra, bundle and LLDB finalization

* test(providers): run the lifecycle scenarios over every runOwned adopter shape (#10707)

One table of testing.md lifecycle rows runs over partial, diagnostics, publishing and lease adopters, replacing the hand-written precedence cases it covers.

* chore(lint): move runOwned's classifier baseline entry with the file
2026-11-15 17:44:13 +08:00
morluto b87b318ae7 refactor(wasm): leave WABT cancellation to the shared owners (#10796)
* fix(pcap): report a refused tool launch as an unavailable tool

When a capture read's tool launch was refused with EPERM, pcap reported it
as unreadable tool output (AnalysisProtocolError) instead of an unavailable
tool. Pcap also never told "not executable" apart from "missing".

Add classifyLaunchFailure in src/process as the one owner of which spawn
errnos mean missing (ENOENT, ENOTDIR) and which mean present but not
executable (EACCES, EPERM, ENOEXEC, EISDIR). Pcap, apktool and gdb used three
different errno lists; all three now use it.

* refactor(process): classify owned command outcomes through one owner

Add ProviderOutcome, read once from the supervisor's facts, and classifyProviderOutcome, the one mapping from an outcome to a typed error. Migrate pcap and the shared artifact producer command failure (TShark, sigrok). Add the rea/provider-failure-classifier lint rule with a shrink-only baseline of per-site (cause: unknown) => AnalysisError functions.

* refactor(process): classify Java and adb launch refusals with the shared owner (#10598)

The Ghidra and JADX Java probes and adb's version probe now take
missing-versus-not-executable from classifyLaunchFailure instead of their
own errno checks. ELOOP and ENAMETOOLONG join the owner's missing set, so
adb keeps its existing classification and both Java probes now treat a
non-directory path component as a missing runtime.

* fix(pcap): preserve command launch attribution

* test(pcap): expect an unknown failing stage when a limited capture tool exits unsuccessfully

* refactor(process): classify owned command outcomes through one owner

Add ProviderOutcome, read once from the supervisor's facts, and classifyProviderOutcome, the one mapping from an outcome to a typed error. Migrate pcap and the shared artifact producer command failure (TShark, sigrok). Add the rea/provider-failure-classifier lint rule with a shrink-only baseline of per-site (cause: unknown) => AnalysisError functions.

* test(lint): cover baselined files in the provider-failure classifier rule

* fix(apktool): report the decoder JVM's heap exhaustion as memory, not a target refusal

The supervisor now watches for a managed runtime's out-of-memory report while output streams (JVM java.lang.OutOfMemoryError, V8 'JavaScript heap out of memory'), so a retention budget cannot hide it. ProviderOutcome gains a runtime-exhausted kind and classifyProviderOutcome maps it once. Apktool decode opts in and moves to the shared classifier; Ghidra's single-purpose output token becomes the JVM runtime option and its area matching reuses the shared table.

* fix(process): keep the runtime's out-of-memory report on the memory error it supports

* fix(pcap): attribute a limited capture's exit through the prlimit wrapper

* fix(process): keep a shortened out-of-memory line marked as shortened

* fix(process): identify the refused executable for every launcher shape

Name the executable on errnos Node throws synchronously from spawn
(ENOEXEC, ENOTDIR, ...) and read the Windows native launcher's
executable-launch constraint, so pcap attributes those refusals to the
selected tool or limiter instead of an unknown stage.

* test(pcap): prove synchronous launch naming with ENOTDIR on every POSIX host

* fix(apktool): advise the JVM limit that governs the reported memory area

* fix(ghidra): read the supervisor's parsed OOM area instead of reparsing lines

* fix(sqlite,capture): read V8 heap exhaustion from streamed output, not retained stderr (#10724)

The SQLite worker and the Node capture decoders now declare runtime: "v8", so the supervisor observes V8's heap report as output streams. Both classifiers read the runtime-exhausted outcome instead of matching retained stderr, so a report beyond the retention budget, or one whose process then overflowed its output budget, is still reported as memory exhaustion.

* refactor(evm): classify EVMole worker outcomes through the shared provider classifier (#10716)

* refactor(evm): classify EVMole worker outcomes through the shared provider classifier

The EVMole worker's cancellation, deadline, output budget and CPU/file-size signals now go through classifyProviderOutcome; only the prlimit launch refusal and the reserved EFBIG exit remain in its dialect. The worker is opted into V8 exhaustion scanning so a heap report names its fixed 256 MiB heap. providerOutcomeOf now reads a launch refusal only from a spawn errno, so a private-storage or carrier read errno is no longer reported as a limiter launch failure.

* fix(evm): keep a stream failure's exit status and update Linux worker expectations

* refactor(process): read launch refusals through the launcher identity owner

* refactor(wasm): classify WABT command outcomes through the shared provider classifier (#10727)

* refactor(wasm): classify WABT command outcomes through the shared provider classifier

WABT's cancellation, command deadline, output budget, and inherited-limit signals now go through classifyProviderOutcome; its dialect keeps only the configuration-probe and launch-refusal unavailability and the unclassified-exit output error. A tool ended by SIGXCPU/SIGXFSZ or the memory limit is now a resource constraint instead of an output error, and a refused launch in the producer phase is reported as unavailable. The shared timeout context gains the analysis setting that selected the deadline, and the output-limit context may name its exceeded budget.

* fix(wasm): keep setup guidance when a WABT version probe's output cannot be collected

* refactor(providers): classify raw provider failures through one owner

* refactor(process): build classified cancellations with cancelledBy and the carried caller reason

* chore(lint): count runOwned's close-failure mapper from main in the classifier baseline

* fix(process): carry the runtime's exhaustion report on every memory error that claims it

SQLite and historical-capture memory errors recognised a streamed V8 heap report but dropped it, so a report outside retained output reached callers without its source line or omission flags. One projection, runtimeExhaustionReport, now owns that shape for the shared classifier and both consumers. Also lowers the EVMole and WABT cancellation-constructor baselines.

* fix(process): keep the provider's recovery for a native-thread exhaustion report

The shared classifier replaced the provider-supplied recovery with generic capacity advice whenever the JVM reported it could not create a native thread, dropping Apktool's launcher guidance and flipping retryable to true. The provider's recovery now wins; the capacity advice remains the fallback for providers that supply none.

* refactor(pwntools): classify pwntools failures through the shared provider classifier (#10782)

* refactor(pwntools): classify pwntools failures through the shared provider classifier

pwntools' per-site classifier is replaced by pwntoolsFailureDialect for classifyProviderFailure. Command outcomes go through classifyProviderOutcome, launch refusals through launcherRefusalOf instead of a 'spawn ' syscall prefix check, the input limit through the file dialect's fileSize, and the cleanup report through commandCleanup. The shared read cancellation now carries the reader's partial observation, and resource-signal reports accept provider detail.

* fix(pwntools): report a failed command's output once and expect resource_constraint for overflow

The shared outcome owner already attaches the command's output as capturedOutput, so the pwntools adapter fallback no longer copies the same text into diagnostics.stdout, diagnostics.stderr and diagnostics.captured_output; diagnostics keep the exit facts. Update the overflow expectations the dialect migration missed (layout provider cleanup and over-budget rows, pwntools cleanup) to resource_constraint.

* refactor(wasm): leave WABT cancellation to the shared owners

WabtFailures matched any Error named AbortError as a cancellation. No WABT path produces one: reads, path checks and owned commands report cancellation as ArtifactReaderFailure or OwnedCommandFailure, which classifyProviderFailure and classifyProviderOutcome already classify with the caller's reason. The provider's two signal checks now build the cancellation with cancelledBy. Lowers the AnalysisCancelledError constructor baseline for both WABT files to zero.

* fix(wasm): report a cancelled snapshot write as the caller's cancellation

The private snapshot write passed the caller's signal to Node's writeFile, whose abort envelope reached the dialect's unclassified output failure once the name-based AbortError guess was removed. The write now turns the envelope carrying its own signal's reason into cancelledBy at the producer.

* refactor(firmware): classify firmware failures through the shared provider classifier (#10810)

Firmware producers now report their own deadline expiry: the start check, the output inventory and launcher admission throw the phase timeout where they observe the abort, and a read the phase signal cancelled reports it through the dialect's interruption. The classifier no longer rereads a cancellation's reason as a timeout. A new readCleanup hook keeps firmware's cleanup report shape.
2026-11-15 17:42:49 +08:00
morluto 2caedcfb57 refactor(platform): give the host platform one owner and ratchet ambient process.platform reads (#10911)
* refactor(platform): give the host platform one owner and ratchet ambient process.platform reads

* refactor(process): pass the host platform to process sampling instead of reading it ambiently

#10755 added two process.platform reads (ps selection in process sampling and the
desktop build doctor's ps fact) after the host-platform ratchet was measured.
The desktop fact already receives the host platform from doctor; process sampling
now receives it from the capture harness, removing all five ambient reads there.
2026-11-15 17:42:23 +08:00