* feat(errors): type refusals at the producer with a total recovery
A resource refusal without authored text fell back to resource-kind advice:
a fixed parser ceiling reported as memory told the caller to add host
memory, an upstream IDA reply ceiling pointed at REA's MCP transport, and a
fixed Kuna text ceiling reported as capacity was marked retryable. An
ambiguous selector was reported as invalid input that pointed at the schema.
Add AnalysisRefusal with three variants built from facts producers already
own: fixed_ceiling (FixedAllowance, now with limit, unit and narrowing),
adjustable_budget (AnalysisSettingLimit) and ambiguous_selector (new
AnalysisInputError option). Each variant is attributed to the error's
operation and projected under details.refusal. Presentation is a total
switch over the variants, and authored text still takes precedence.
Migrate the first producers. GraphQL nesting, Playwright trace input, the
Kuna function text and the IDA upstream reply become fixed ceilings. jadx
overloads and Kuna ambiguous names become ambiguous selectors, replacing the
two Kuna-specific ambiguity texts. Seven occupied admissions declare
measuredExhaustion. The rea/typed-resource-refusal lint rule rejects new
untyped resource constraints, and the remaining 24 are counted in a baseline
that can only shrink.
* fix(errors): honor refusal recovery ownership
* refactor(parsers): report format failures through one ParseFailure algebra
Add ParseFailure (malformed, truncated, unsupported, limit, with format,
offset and field) and its one carrier, ParseFailureError, in the domain.
BinaryReader now names truncation and the field it read, and its native
format and resource failures are ParseFailureError subclasses.
Providers projected a format failure with their own instanceof ladder,
and some dropped the selected path from the invalid_format message.
parseFailureAnalysisError now owns that projection for the native,
Go, devicetree, V8, Windows Shortcut and Electron fuse providers.
The V8 code cache, Windows Shortcut and Electron fuse format classes
are removed in favour of the algebra.
The architecture guard counts raw Buffer and DataView reads, plain
Error throws in parser folders and plain JSON.parse of parser input
against baselines that can only shrink.
* fix(parsers): retain malformed range offsets
* fix(hopper): name the unconfirmed shutdown step in retained cleanup
When close_binary retains a Hopper document, the cleanup error lists the
retained resources but not why shutdown went unconfirmed. On Linux CI,
hopperClient.handshakeRecovery "recovers an application exit after the
first valid handshake occurs during cleanup" fails intermittently. The
vitest blob reports show only hopper-document and the runtime root
retained with authenticated_application "readable". From those facts we
cannot tell whether the shutdown request was not sent, failed, or was
sent while the application was still observed as present.
Record the shutdown request outcome (method, status, error tag and
provider state, or why it was not sent) and the final exit observation
of the authenticated application in the retained-document diagnostics.
Exit observation now reports a positively observed Linux exit in
progress (PF_EXITING before state Z) as "exiting", separate from
"present". Neither confirms exit. The flaky test asserts the projected
error, so a later failure prints these diagnostics in the CI log.
Refs #10556
* fix(hopper): retain fallback shutdown diagnostics
* test(hopper): retry cleanup after confirmed application exit
TOOL_EFFECT_AUDIT was typed Record<string, EffectAudit>, so a tool missing
from the audit or a misspelled entry was found only by a runtime throw or
the key-equality unit test.
Check the audit with satisfies Record<PublicToolName, EffectAudit>, where
PublicToolName is the union of TOOL_CONTRACTS names. A missing, extra or
misspelled entry is now a compile error. String lookups for callers that
hold an unchecked name keep their runtime refusal.
Delete the unit test that compared audit keys with contract names; the
type now owns that check.
Table-backed add-mcp agents now project their client-table row instead of
redeclaring format, display name and profile in a second table. The row
also carries add-mcp aliases and project documents. mcp add now applies
the read-scope rules setup already enforced: Gemini settings policy blocks,
the Claude Desktop MSIX write blocker and MSIX AppData redirection.
* refactor(javascript): classify implicit execution in one owner
Add javascriptSemanticExecutionPoints, a total classification of every
runtime Babel node type into the ways its evaluation can run code without
call syntax. The member-mutation escape walk, the accessor operand
collector and the catch-source reachability check consult it instead of
their own syntax lists. Kinds those collectors have never applied are now
named in exhaustive switches rather than absent.
A guard test fails when the installed @babel/types defines a node type the
table does not classify, and unknown node types classify as unsupported
syntax rather than inert.
* test(hopper): retry cleanup after confirmed application exit
* fix(javascript): classify enumeration and catch binding effects
* fix(javascript): retain catch uncertainty for identifier reads
A cancelled GDB startup whose process stop could not be confirmed retained a
capability error. The next cleanup retry then reported "This analysis is
unavailable for the current target" instead of the caller's cancellation.
Move Rizin's retention rule (the #5963 fix) into one shared
retainedDebuggerCleanupFailure in src/process/ownedDebuggerSession.ts and use
it from both session managers. The GDB command-cancel path now retains the
cancellation itself, so retry reports no longer depend on Set insertion order.
Fifteen domain schema files each declared z.string().min(1), eight of them
as boundedTextSchema although it bounds nothing. They now share
nonEmptyTextSchema from src/domain/textSchemas.ts; the one trimming copy in
reconstructionCoverage.ts is renamed trimmedTextSchema.
Three private UTF-8 prefix truncators (gdb recent MI records, rizin
diagnostics, environment replay) and src/process/boundedUtf8Text.ts are
replaced by utf8HeadExcerpt in src/domain/utf8HeadExcerpt.ts, which encodes
only the retained prefix and returns a slice of the source text.
The fake application now leaves a marker when its children never report
ready, and the tests' session requests throw on it, so capture releases the
children and the test fails at once instead of waiting for its deadline.
Refs #10366
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
Doctor reads transport, falling back to type, and reports a remote http/sse or
url-only entry as unknown rather than invalid; only a non-text tag is
invalid.
Refs #7650
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
README.md's packet-summary row says tshark/capinfos support is
Linux-verified, but sixteen translations omitted that caveat.
Fixes#7655
Supersedes #7972
The translated phrases are from #7972, reapplied to the moved pcap
guide links; simplified Chinese, which #7972 did not cover, follows the
traditional Chinese wording.
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
Use createNodeDiscoveryTempDirectory for the target-location and
runtime-observation fixtures too, so a TMPDIR containing `_` cannot make
them fail for the same reason as the cancellation test.
Refs #9996
The sibling sites are the ones #4484 switched.
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
The readiness wait from the previous commit continued after its bound
even when a child had not installed its TERM trap, so a slow runner
would reintroduce the race silently. It now exits with an error after
five seconds, and the state line keeps its original escaping.
Refs #10366
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
The changed-prefix regression rewrote its fixture at the same size within
milliseconds of creating it. Linux stamps mtime/ctime from a coarse clock,
so both writes can share every stat field the target guard compares, and the
classifier parsed the rewritten header without seeing a change. Age the
fixture's timestamps before the mutation so the rewrite is observable on any
filesystem while the test still requires a changed-input refusal.
Fixes#8685
Supersedes #8689
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
The client configuration reference had no entries for three clients that
the installation guide lists as supported and sends readers here for.
Add a section for each naming the configuration file setup writes, the
mcpServers.rea entry key, the entry fields, and the shared skill root, as
defined in src/application/SupportedClients.ts and
src/application/ClientConfigurationDocument.ts.
Fixes#7650
Supersedes #7750
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
The parser's work budget was a fixed 64 Mi units, so a real 26 MB .NET 8
NativeAOT image (bflat 8.0.2) stopped frozen-string recovery at 4,433 of
6,685 strings and reported partial. Give each image a base budget plus 8
units per admitted source byte, bounded by the source-byte ceiling, and
record the scaling in the analysis profile (contract revision v11). With it,
that image completes with all 6,685 frozen strings using 71.6 M units.
Only the budget scaling of #4559 is taken: main already resolves the
base-type closure in linear time.
Fixes#4445
Supersedes #4559
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
RFC 6891 makes a second OPT record a format error. Reading every OPT record
let a later one with extended RCODE 0 clear an earlier nonzero one, so an
error response could attribute its addresses.
Refs #10211
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
A retroactive protocol conformance sets flag bit 0x40. The static
witness check admitted only the type-reference kind bits (0x38), so a
retroactive conformance lost its static witness table. The admitted
mask is now 0x78; conditional requirements, resilient witnesses and
generic witness tables stay unadmitted.
Fixes#7445
Only the mask change and its tests are taken from #7489.
Supersedes #7489
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
An XML plist whose DOCTYPE ends in an empty internal subset ("[]",
optionally with whitespace) declares nothing that could expand, but the
preflight refused it as internal entity expansion. Interface Builder
bundles then omitted the nib and blamed aggregate decode budget
exhaustion. Skip whitespace-only subsets; any declaration still refuses.
Fixes#8407
Supersedes #8438
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
The nesting guard ran on a child visit of GgufHeader#value, so an empty
innermost array never produced one and a 33rd frame parsed as success
while the reported limit stayed maximum_array_frames: 32. Charge the
frame when entering the ARRAY branch instead, so the bound counts the
advertised unit and no longer depends on a scalar child that is not a
frame.
Fixes#7312
Supersedes #7698
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
inspect-native-exports treated any nonzero byte in .dynsym entry 0 as a
malformed table. That is right when the reserved slot holds a real symbol,
which was #7389, but some Android arm64 libraries set only st_name. Those
files were rejected before any export could be read.
Skip that entry, keep the inventory complete, and record the nonzero
st_name in limitations. Any other nonzero field still fails the table.
Fixes#9700
Supersedes #9752
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
Charge the manifest split facts against the native inventory's retained JSON
allowance and refuse more than 64 split types or 256-character names as an
unknown, so a crafted manifest cannot overflow the response. Reports saved
before split facts existed still parse: the report schema treats absent
facts as optional while the live bridge contract requires them. The JADX
regression now decodes a generated binary manifest through
nativeDeclarations, with named and resource-ID-only attributes.
Refs #9756
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
A base APK whose manifest requires ABI or density splits keeps its native
libraries in those splits, yet inspect_android_jni_bindings reported a
complete library inventory with zero libraries. The bridge's native
declaration inventory now also decodes the root binary manifest without a
session and reports its requiredSplitTypes and isSplitRequired facts, or why
they are unknown (bridge identity 11). When splits are required or unknown,
the library inventory is partial with an unknown that says so, and a binding
without a candidate is unknown instead of unmatched.
The binary manifest fixture from #9769 was used to check the bridge with the
real jadx-headless-mcp 0.7.1 JAR, including stripped attribute names that
resolve through framework resource IDs.
Fixes#9756
Co-Authored-By: 2SinceJuly <202121119+2SinceJuly@users.noreply.github.com>
Co-Authored-By: Phon <twilyt1@proton.me>
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
Setup's result is agent-only, so incur hides it when stdout is a terminal and
no format was chosen. Only an interactive setup or a fully interactive TTY dry
run rendered a summary, so `rea setup --yes` printed nothing. Render the dry
run plan or the existing result summary under incur's own hidden-output
condition instead.
Fixes#10531
Supersedes #2185
Co-Authored-By: balkan16 <57933796+balkan16@users.noreply.github.com>
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
JSON.parse keeps only the last value for a repeated name, so an ASAR header
that declared a path twice listed one member, hid the other and still
reported complete coverage.
Add findDuplicateJsonMember, an iterative scan of validated JSON text that
does not recurse, so the deeply nested headers REA already admits stay
readable. readAsarHeader now refuses a repeated member name as a malformed
index, the same way the listing refuses unsafe path components.
trace_native_values ended at a provider-classified thunk whose model
holds only a synthetic return (an external import) or borrows its
target's operations, without saying the target was never examined, so
a stub's small model read as the end of the call chain. Thunk targets
are now reconciled after traversal: one another path traced, attempted
or queued is left to that path, and the rest are reported in unknowns
with the target in a new optional related_procedure field.
Fixes#8476
Supersedes #8678
Based on #8678 and rebased onto main's separate call-binding module.
related_procedure is exactOptional and set only on the thunk unknown,
so other unknowns and retained traces keep their shape.
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
When the macOS process-ownership helper's swiftc compile hit
REA_SWIFTC_COMPILE_TIMEOUT_MS, inspect-macho and other native commands
reported a non-retryable provider failure, while process capture
already treated the same fact as retryable. The helper's error now owns
that interpretation (compilerTimedOut and compileTimeout, with its
deadline, message and recovery). Native commands return a retryable
AnalysisTimeoutError naming REA_SWIFTC_COMPILE_TIMEOUT_MS, and process
capture reads the same fact.
Fixes#8936
Supersedes #8938
Based on #8938; the interpretation moves to the error's owner so both
consumers share it.
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
A worker heap too small for REA's bridge stops the JVM before the protocol
handshake, so the launcher's retained stderr is the only evidence, and the
failure was reported as capability_unavailable with configuration advice.
When that stderr carries HotSpot's java.lang.OutOfMemoryError heap reasons,
report resource_constraint for memory with observed_max_heap_bytes null and
the same REA_JADX_HEAP_MIB remediation the bridge-reported heap failure uses,
now shared and stating -Xmx precedence. Other launcher stops are unchanged.
Refs #3517
Supersedes #3573
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
planOutputTree admits the destination before writeOutputTree creates any
parents, so the parent-creation defect fixed for environment bundles does not
reach desktop exports. Keep that sibling comparison as a test.
Refs #7688
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
Desktop capture used XDG_* values without checking that they are
absolute, so a relative XDG_CONFIG_HOME (or data, state, cache) moved
state-directory, desktop-entry and build-secret lookups under the
working directory. The XDG Base Directory Specification says to ignore
a relative value; one xdgBase helper now applies that rule and falls
back to the home default.
Fixes#9702
Squashed from #9705 and rebased onto main. Setup's client registration
paths are left unchanged: each client's own handling of a relative
XDG_CONFIG_HOME has to be verified before REA writes elsewhere.
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
The witness seed accepted its folder with a readdir check that follows
symlinks, treats a missing path as usable, and reports a regular-file
parent as "not empty". The record is later saved through planOutputTree,
so a seed could accept a folder the save then refused. The seed now
admits the folder with planOutputTree too, and returns its refusal.
Fixes#9291
The MCP boundary test is from #9292 by viniciusdsandrade; its parallel
admission helper is not taken, since planOutputTree already owns the
rule and export already uses it.
Co-Authored-By: Vinícius DS Andrade <95101505+viniciusdsandrade@users.noreply.github.com>
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
environment-build/export ran an unclassified recursive mkdir of the
destination's parent before SafeOutputTree validated anything. A
trailing `..` therefore created directories before it was refused, and
a regular file in the parent path surfaced as unreadable_output instead
of refusing the selection.
SafeOutputTree.create gains a createParents option that admits the
root's own syntax first, then creates the parents, mapping EEXIST and
ENOTDIR to a destination refusal. The environment bundle writer uses it
instead of its own mkdir.
Fixes#7688Fixes#7571
The non-directory mapping comes from #7734 by HeadmasterEggy, and the
syntax-first check from #7588 by harshitethic; their tests are kept.
Co-Authored-By: Avocado <headmastereggy@gmail.com>
Co-Authored-By: Harshit Sharma <66710144+harshitethic@users.noreply.github.com>
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
The block branch of the symbolic return flow already appended without a
call spread, but the adjacent if branch and both continuation-condition
joins still spread, so a generated wrapper with a wide tree of unknown
conditions could still overflow the argument limit.
Refs #10445
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
`target.push(...items)` passes every item as a call argument and
overflows the stack once a target-sized array grows past the engine's
argument limit. This class was fixed one site at a time in #10304,
#10343, #10362 and #8632. Add one shared appendAll helper in
src/domain and use it at the remaining sites whose arrays grow with the
input: XML member counting, app-task row suggestions, native call
routes and unresolved calls, call-path caveats, environment-manifest
merges, reference-source import entries, dyld trace edges and JS
symbolic flow values. The web bundle merge from the previous commit now
uses the shared helper too.
Bounded spreads (fixed literals, page-sized reads) are left as they are.
Refs #10445
Refs #10342
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
morluto