* fix(process): report the effective regex worker heap
Node gives --max-old-space-size, including NODE_OPTIONS, precedence over
the worker's 128 MiB resourceLimits request. Name that effective
old-generation heap and stop saying the request has no runtime override.
The matching deadline stays a fixed cap.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(process): keep SIGXCPU and SIGXFSZ retries uncertain
The signal does not prove the retained rlimit was exhausted, so an
unchanged retry may succeed. Project that retry as possible instead of
saying it fails the same way.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(process): leave percentage heap overrides unestablished
--max-old-space-size-percentage takes precedence over the absolute flag
in either order. Node computes that MiB from startup memory, which a
later read does not recover, so the effective heap stays unknown instead
of the 128 MiB request or an absolute flag.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(process): read launch NODE_OPTIONS, not process.env
A preload can rewrite process.env.NODE_OPTIONS after V8 has applied the
original flags, and those flags are not copied into execArgv. Use the
initial environment block when it can be read, and leave the effective
heap unestablished when that provenance cannot be recovered.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(process): leave env-file heap overrides unestablished
--env-file and --env-file-if-exists can supply NODE_OPTIONS when that
variable was absent from the initial environment. The file is not in
/proc/self/environ, so a missing entry is not the 128 MiB request.
An initial NODE_OPTIONS is still parsed, because Node does not let the
file replace it.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(process): leave --max-heap-size old generation unestablished
A positive --max-heap-size or --max_heap_size sets the total heap, not the
old generation, and can stay below the 128 MiB worker request when an
old-space flag is also present. The effective old generation stays
unestablished while that value is in effect. Zero or empty clears it.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(process): leave config-file heap options unestablished
--experimental-config-file and --experimental-default-config-file can
apply a heap percentage from a file that is not launch provenance. The
effective old generation stays unestablished when either flag is present,
including beside an absolute --max-old-space-size.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(process): read launch exec argv from the command line
process.execArgv can be cleared after V8 applies --max-old-space-size.
Linux recovers those tokens from /proc/self/cmdline and leaves the heap
unestablished when that vector cannot be separated from the program entry.
Heap probes compare pinned absolute limits instead of a percentage of host memory.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(process): leave a title-overwritten command line unestablished
process.title and --title replace /proc/self/cmdline with the title.
A buffer without a program entry is not an empty option list, so the
heap stays unknown instead of the 128 MiB request. The command-line
probe does not read /proc off Linux.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
* feat(environment): render more read tools from the state and describe responses
Keep main's environment serving and fold in coverage for non-entity reads,
response summaries, and MCP tool annotations.
Squashed rebase of #5352 onto main.
* fix(environment): keep rendered documents and quoted samples inside their budgets
A recorded body of exponential numbers can render past 256 KiB and turn a
working read into a state size error, so that document stays a recorded replay.
Description samples are measured after JSON quoting, and the replay-fallback
catalog check uses a tool count the longer response descriptions still separate.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(environment): pin the replay card this branch serves
The replay-only card test pinned main's 11,131-byte card. This branch's descriptions name each tool's recorded responses, so the card it serves is 12,818 bytes; the pin and its name follow.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Keep the 16 MiB ASAR header/index ceiling fixed and report that budget,
the lack of a runtime override, and that source selection does not bypass
header admission.
Refs #5910
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(mcp): name response-byte and startup ceilings in recovery
Name REA_MCP_MAX_RESPONSE_BYTES, its effective value, the 10485760 byte
floor, and process restart when the stdio response budget is hit, and
state that the setting does not raise Node's string ceiling. Name fixed
MCP startup phase budgets when those deadlines fire, and keep Codex's
30 second startup figure as client guidance.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(mcp): distinguish string ceilings and keep dossier recovery
Name REA's reserved string limit separately from Node's ceiling, restore
detail:summary discovery for an oversized complete evidence bundle, and
apply the doctor deadline sentence when binary_session itself times out.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(mcp): keep retained SDK errors inside the text budget
Quoted SDK validation diagnostics expand when recovery JSON is escaped
into MCP text. Shorten only that preview so the retained-evidence
reference still arrives as parseable error content.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(mcp): keep export advice when a summary exceeds the response budget
An oversized get_evidence_bundle summary no longer repeats detail: summary.
export_evidence_bundle still writes the retained bundle outside one MCP frame,
including when a larger byte budget cannot raise the string ceiling.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* style(mcp): format the SDK diagnostic preview return
oxfmt rejected the fitted-preview ternary in EvidenceMcpServer.ts, which failed Static checks and the CI required gate.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(browser): name web-bundle budgets and recover web deadlines
Name the existing heap, per-script, and total web-bundle phases, and reject a per-script deadline above the total. Promote capture, CDP command, source-map decode, and source-map fetch deadlines at today's defaults. Name DevTools timeoutMs, and state that discovery, cleanup grace, byte budgets, and Electron hook retention have no runtime override.
Refs #5910
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(browser): apply selected web deadlines to the owning phase
Keep HTTP and browser-socket discovery on the fixed discovery deadline so a
short command_timeout_ms starts at the handshake. Apply scenario
command_timeout_ms to launch startup and the primary connect session, and
keep service-worker restoration on the fixed 5000 ms grace. Omit
command_timeout_ms from recovery text when the operation does not accept it.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(browser): retain scenario commands that finish after their deadline
A timed-out scenario command stays owned so cleanup can close a late
browser or context and restore a mutation that lands after the deadline.
Child storage replies use that deadline, scenario recovery names
command_timeout_ms in the capture JSON, and the session test connection
implements beginCleanupGrace.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(browser): keep child command timeouts as the public failure
A nested command such as Page.enable that is delivered but never answered
was stored as protocol_error, hiding the command name, selected deadline,
and recovery. Store the typed timeout and assert that public error.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(browser): keep runtime cleanup on the fixed grace
Switch runtime restoration to the cleanup grace before the first send so a
short command_timeout_ms cannot fail an otherwise successful observation.
Stop advertising command_timeout_ms on list_browser_targets, which never
bounds a post-discovery command.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(browser): prove a short command deadline cannot fail runtime cleanup
Drive restoration through the runtime fixture. A cleanup command that
outlives command_timeout_ms still succeeds, and a stalled restoration
names the fixed 5000 ms grace.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(browser): name the effective source-map decode deadline
codecBudgetLimitation reported a fixed 20-second deadline after a
caller selected decode_timeout_ms. Result and Evidence limitations
now name that effective millisecond deadline.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(browser): separate handshake and target-attach deadlines
A browser-level Target.attachToTarget no longer shares the WebSocket
handshake budget. Blocked service-worker attachment keeps discovery on
its fixed deadline and applies command_timeout_ms to the later handshake
and attach.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(browser): name the web-bundle resource each runtime exhausts
Bun ignores the V8 old-space flag. A Windows Job reports the process
commit limit, and other Bun hosts no longer suggest raising max_heap_mb.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(browser): cite the browser-socket discovery message budget
WebSocket discovery enforces the 64 MiB protocol budget. Timeout
recovery no longer advertises the 8 MiB HTTP document budget.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(browser): bound the blocked service-worker check before discovery
The first registration check has its own fixed 10s deadline. Discovery,
handshake, and Target.attachToTarget stay on their separate deadlines.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(browser): name the operand that stopped web-bundle admission
Host heap pressure no longer claims the fixed result budget is exhausted.
A Windows Bun allocation before the commit limit still names that limit.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
* feat(environment): extract the initial state a capture's reads observed
Add the environment-state analysis view. From the parents the environment
manifest accepts, it extracts the data the capture's reads observed before
the first related write: entity types with an identity field (proved by a
path parameter or inferred from an identity name), entities with their
recorded values and provenance, documents for written non-entity paths, and
a render plan per read tool that is kept only when it reproduces the tool's
latest initial recordings from the state; every other read tool is a
recorded view with the reason. The Evidence is always inferred.
Response-body decoding moves to a domain owner shared with replay, exact
JSON parsing moves to the owner of the exact-number policy, the manifest and
state views share parent admission and record grouping, and
projectAnalysisView takes its accumulators as one object.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): bound state at redacted writes; infer identity only from items
A scenario write whose URL was redacted has no request line, so it was
dropped and reads after it counted as initial state. Such a successful
write now joins the timeline with an empty path, whose root relates to
every type, so it ends every type's state; a limitation counts them.
A list whose recordings were all empty passed the "every listed item
carries it" check vacuously and published an identity field nothing
showed; it now needs at least one item, else the view stays recorded
as no-identity.
A value that differed between two request lines of one item tool, such
as a query variant, was called volatile, and the gate then substituted
each recording's own value; only repeats of one request line now show
volatility, so differing lines go through fields-differ.
List membership is compared only while a list can still be the whole
collection, and merged entities are indexed by type and collection once,
so paged captures no longer cost pages times entities (8,000 pages of
40,000 entities: 5.7 s to 0.23 s).
Docs and the tool contract say scenario state keeps the capture's
redaction markers and how redacted writes bound it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): keep first sightings, one list pointer and linear field unions
Entities took the order of the first plan observed, and item plans run
before list plans, so a complete list read before its detail reads came
out reordered and lost "all" membership; each entity now keeps its
earliest recorded sighting.
A list plan reused the first recording's pointer for later recordings,
so a later response that was itself an entity wrapping a list was read
as that list; every recording must now locate the same list.
The field union scanned its ordered list for every key, quadratic in an
object's width; a set now answers membership.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): keep post-write values and partial lists out of the initial state
A value that changed between repeats of one request line was named
volatile whenever no write under the type's own root lay between, so a
write under another root (for example POST /api/dashboards/db before a
second GET /api/search) left its effect in the initial state. Any
recorded write between the two observations now leaves every tool of the
type recorded with reason fields-differ instead.
A list was published with membership "all" even when a total or count
envelope member exceeded the number it listed. Such a counter now makes
the membership "recorded".
Also drop a stale doc comment on ObservedEntity.order.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(environment): scope item-shaped and deleted written paths out of documents
State documents are held only for list-shaped paths a POST, PUT or PATCH
tool writes. Reads of item-shaped written paths and of paths that are only
deleted stay recorded views; the guide now says so and the spec defers them
to a later increment of #4933.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): count writes in flight when a value changes between reads
The volatility check compared write start positions with the two reads,
so a write already in flight when the earlier read started was missed and
its effect was named volatile. A write now counts when its interval
overlaps the span from the earlier read's start to the later read's end.
The check also ran once per changed field and scanned every write. It now
runs once per pair of observations, answered by a binary search over the
writes in timeline order and the latest end among each prefix.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): catch cross-tool changes across writes; read only item totals
A value that changed across a write was caught only between repeats of
one tool's request line, so a list read before a write and an item read
after it put the post-write value into the initial state. Changes are
now classified per entity in timeline order after observation: across
any recorded write, a change between repeats of a line or between two
lines of the type rewrites the type (every tool recorded as
fields-differ); with no write between, a repeat's change is volatile.
The classification lives in environmentStateMerge.ts.
Any numeric member whose name contained "total" or "count" counted as
an item total, so total_pages, account_id or page_count made complete
lists partial. Item totals are now members whose name tokens include
total or count and none of page, pages or per; a string of digits counts
as a number; and a set has_more, next or next-page token member also
marks the list partial.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): keep post-write observations out of initial state
* fix(environment): retain field provenance and bound state inference work
* feat(environment): serve tool calls from the environment state
`environment-serve --state <state.json>` serves an environment from the
initial state an environment-state view extracted. Read tools whose views
are state-backed answer from a live state store: item views render the held
entity into the recorded envelope, list views render the collection or the
recorded members, and document views return the held value; a missing entity
or one lacking a required field is refused, and requests the state cannot
answer fall back to their recordings with the reason. The /agentenv data
plane implements data/reset, data/add (a rea-state/v1 document as a data
part, base64 bytes, a file:// path or an https URL, validated and replaced
atomically) and data/get, and the card advertises those operations and an
intake declaration. Stage 1 is read-only: write tools are not served in
state mode. Without --state nothing changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(environment): record verification after merging the state view's review fixes
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(environment)!: serve the environment state by default
environment-serve derives the environment state at startup by running the
environment-state view in-process on the capture with the manifest's
template rules, so an environment comes with its state and a user supplies
nothing extra. The derived Evidence is the same one inspect-analysis-view
produces by hand. --state is removed; --replay-only serves exactly the
stateless replay of main. When the state cannot be derived, or its catalog
does not fit the discovery budget, the server starts in replay mode, logs
the reason and reports it in the stop summary's `state`. Exported bundles
serve the state by default too: export derives it the same way, so the
verifier's tool count and checked answer match what is served, and the
README describes the served mode.
BREAKING CHANGE: environment-serve answers read tools from the derived
state and does not serve write tools unless --replay-only is given; the
stop summary reports `state` instead of `state_evidence_id`.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): bound data-plane replies, keep batch order, cancel intake on stop
- One /agentenv response is held to 64 MiB: replies are measured as they
are produced and one past the budget becomes a -32000 error stating it.
- Every message of a batch enters the arrival-order queue on arrival, so
another request's reset never lands between them.
- Stopping the server, or a caller leaving, cancels the active download
and answers queued calls `cancelled` without running them.
- file:// URIs name only this host: another host or a network share is
refused before anything is opened.
- A loaded view value may not give an entity another identity.
- List answers refuse a member lacking a field every recording carried,
as item answers do.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): keep bodies too deep for the state's nesting out of the state
A recorded body at the JSON depth limit was accepted by extraction, then
nested a few levels deeper as a document or entity value, so sealing the
state Evidence threw and environment-serve and environment-export stopped
instead of falling back. stateBodyJson, the one rule extraction and
serving share, now treats a body nesting deeper than the limit minus 16
levels as unusable, so its tool stays a recorded view.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): measure data-plane replies before building them
83fcc739 bounded one /agentenv response to 64 MiB, but it still
serialized each reply whole before measuring it, so a batch of data/get
calls on a large state built every refused reply in full only to drop
it. Each reply is now measured in bounded JSON parts (the shared
jsonParts encoder), its parts kept only while it fits; a reply past the
budget is counted for its exact reply_bytes but never assembled. A test
checks that a refused read is never handed to JSON.stringify whole and
that its reply_bytes is exact.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): release batch replies as they are encoded; judge exact numbers by value
answerBatch kept every settled reply in its pending array until the
batch ended, so a batch alternating data/add and data/get held each
loaded state alive; each reply now leaves the queue as it is encoded.
The exact-number policy judged a number by its recorded spelling, so a
state loaded with 1e20 came back from data/get as
100000000000000000000, which data/add refused, breaking the promise
that data/get output loads back unchanged. A number is exact only when
both its recorded spelling and the one serialization writes are safe,
so integers from 2^53 below 1e21 are refused in any spelling; replay
keeps such bodies as recorded text. A domain test checks that every
accepted number is accepted again once written back.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): keep room for refusals and framing in the reply budget
The response budget counted accepted replies only: a budget error that
replaced a reply, and a batch's brackets and commas, were not charged,
so near 64 MiB later refusals pushed the body past the advertised
limit. Every reply still to come now keeps room for its own budget
error, and a batch keeps room for its framing, so a refusal always fits
and the body never passes the budget. Reply ids and which messages are
answered come from one place shared by the replies and the reserve.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): refuse a loaded state that data/get would write past the intake limit
data/add measured only the document it was given, while data/get
writes the state back with the members a document may omit (types, the
state Evidence id, view fields and record citations). A document just
under 32 MiB could therefore load and be reported larger than any intake
accepts, breaking the data/get then data/add round trip. replace now
measures the state as data/get writes it, part by part, and refuses it
past the limit, leaving the state unchanged. The limit moves to the
state document module, so the application store can read it; the
intake re-exports it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): keep merged export checks on the shared test graph
Main's export tests installed module mocks for the whole services
project, which reuses one module graph. A tarball read failure then
missed its importer, and a one-shot heap mock could be taken by another
file. Read those through bindings the tests replace for one case.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(environment): keep a state file whose close failed owned until it closes
A file:// data/add that could not close its descriptor dropped the handle,
and an admission failure whose cleanup close failed dropped its owner, so
nothing could retry the close or report it.
The read now runs through withOwnedRegularFile, and a failed close is
retained in a server-lifetime ArtifactResourceScope. The next file read
retries it first and answers -32000 while it is still open; inline and
https documents still load. Closing the served environment retries it again
and rejects with the unfinished close.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>
Co-authored-by: gensokyo-dev <gensokyo-dev@users.noreply.github.com>
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
Co-authored-by: morluto <76467478+morluto@users.noreply.github.com>
Fiddler writes a zero-byte raw/NN_s.txt (or _c.txt) for a session that
never got that message. The decoder passed it to the HTTP parser, which
found no header/body separator and rejected the whole archive, so no
session could be read even with --record.
Report a zero-byte request or response member as null, like a missing
member, keep it in location.members, and add a per-record limitation
that says only that the stored member is empty. Non-empty malformed
members are still rejected.
Fixes#5411
* fix(providers): name JEB, IDA, and one-shot command deadlines
Wire JEB catalog discovery to REA_JEB_MCP_TIMEOUT_MS, name the IDA
registration timeoutMs, and add command_timeout_ms for one-shot objdump
and rizin. Debugger, Frida, and Cutter caps stay fixed and say so.
Refs #5910
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(providers): pass remaining deadlines and name the cap that fired
Catalog discovery now sends the JEB budget still remaining to the SDK on
connect and every tools/list page. Objdump and rizin preflight share one
command_timeout_ms across stat and digest; the command still receives that
budget again. Architecture Evidence keeps a selected timeout, and Frida and
Rizin name the capture limit that actually stopped output.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(frida): preserve non-timeout device lookup failures
getDeviceById is a lookup deadline only when Frida rejects with
"Device not found" after the wait. A device-manager or backend failure
stays a capability error. Align the assertions that still expected the
previous capacity and capture sentences.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(engines): keep format ceilings fixed and recover the EVMole deadline
Promote the owned EVMole worker wall clock to timeout_ms and name it in
recovery. Fixed CPU, byte, and address-space refusals in this theme say
the adjustable deadline does not raise them.
Refs #5910
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(engines): recover the allowance that refused the request
Model capacity recovery follows a structured limit instead of message text,
and the shared header ceiling no longer names Safetensors. Firmware signal
errors keep captured tool output, oversized reports do not suggest an
intermittent rerun, and NativeAOT hydration budget exhaustion explains the
fixed parser allowance on the partial report.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(native): recover macOS command deadlines
Promote native_command_timeout_ms (default 60000), native_ui_deadline_ms
(default 180000), and pwntools_timeout_ms (default 30000) as validated
per-call settings. Keep cpuSeconds, byte allowances, LLDB grace, and plist
probes fixed, and say those caps have no runtime override.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(native): report the deadline that actually expired
Pass the selected native command timeout through xcrun resolution, thread
the pwntools deadline into pwndbg failures, and state the selected UI
deadline apart from the fixed wait-step cap. SIGXFSZ and EFBIG recovery
names the file-size allowance.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(native): honor the selected deadline on helper stages
The UI helper compile and capture subprocesses used fixed 60s and 30s
exec timeouts, so a longer native_ui_deadline_ms still failed as a generic
provider error. Preparation timeouts reported the 60000 ms default instead
of the selected native_command_timeout_ms.
Memory resource-constraint recovery names memory again, and the Darwin
plist CLI expectation matches the byte-counted input allowance.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
Select completed source-to-bundle mappings and candidates through the shared view owner, preserve reference-file provenance, and retain full Evidence. Reuse bounded JSON reading, heap admission, and cancellation before session writes.
Verified public Redux Toolkit source/map facts with an independent oracle, equivalent CLI/MCP workflows, focused negative controls, and complete CI/review. Historical paired SDK context estimates fall about 64%; timing/resource tradeoffs and final verification receipts are recorded without a general speedup or actual model-token claim.
Fixes#6112.
* fix(javascript): name heap and semantic ceilings
Refs #5910. Heap exhaustion names max_heap_mb with its effective value,
range, and next-analysis worker restart. Semantic ceiling coverage names
the four max_semantic_* arguments. Wakaru and transfer ceilings say this
release has no runtime override.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(javascript): distinguish adjustable ceilings from fixed limits
The effective transfer file ceiling and the parent result budget follow
the CLI/MCP process heap, so recovery names NODE_OPTIONS instead of
denying an override. A provenance report over 8 MiB keeps the fixed
allowance guidance, and the Wakaru deadline says an unchanged retry may
hit it again.
Refs #5910
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(javascript): keep Bun OOM and checkpoint heap controls distinct
Bun heap-exhaustion exits no longer claim the ignored max_heap_mb V8
old-space limit. Checkpoint decode names the parent heap and Node
NODE_OPTIONS, the same control as the parent result budget.
Refs #5910
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(javascript): name semantic ceilings without claiming both were reached
Truncated semantic retention names the total and per-file controls as
effective ceilings. Callers only know that retention was truncated, so
the limitation no longer says both ceilings were reached.
Refs #5910
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(javascript): name only the semantic ceiling that stopped retention
Semantic ceiling limitations claimed both the application total and the
per-file control whenever either one truncated the graph. Record which
offered budget stopped node and unknown retention, and name only those
controls. Coverage limits still list both effective values.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(javascript): require a value in the NODE_OPTIONS heap example
A bare --max-old-space-size is an illegal V8 flag and prevents Node from
starting. Parent-heap and transfer-ceiling recovery now share
--max-old-space-size=<MiB>, and the Kuna heap remediation uses the same
complete form.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(javascript): limit the old-space placeholder to NODE_OPTIONS recovery
The review covers the CLI/MCP NODE_OPTIONS example and the transfer-ceiling
copy. The Kuna heap-headroom sentence stays unchanged.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(android): name the JADX deadline and recover apktool decode
JADX operation timeouts now name REA_JADX_OPERATION_TIMEOUT_MS, the
effective milliseconds, the 1..2147483647 range, and that a change applies
on the next process start. Apktool decode accepts decode_timeout_ms,
omitted means 600000, and a timeout names that key. ADB deadlines and
apktool and Java version probes stay fixed and say there is no runtime
override.
Refs #5910
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(android): derive capture recovery from the fixed allowance
Main now stores a fixed allowance instead of a prebuilt remediation
string. Apktool and adb capture refusals keep the no-runtime-override
sentence in that allowance name, including that decode_timeout_ms does
not raise the apktool output cap.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(apktool): observe a real decode deadline and process cleanup
A hanging interpreter launcher and its pipe-holding child stay alive
until decode_timeout_ms elapses. The provider then reports
provider_timeout with completed cleanup, and both processes are gone.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
Use compact MCP input advertisements by default while preserving explicit full, canonical validation, output contracts and complete Evidence. Require an explicit selected catalog for runtime identity and honor both profiles in CLI status.
Real ripgrep/Zellij SDK tasks, independent ELF oracle, actual client trials, profile parity workflows and CI are recorded in experiments/analysis-efficiency/mcp-input-catalog.md. SDK token estimates improve; actual Codex token savings and general latency/memory gains are not established.
Refs #5861
BREAKING CHANGE: MCP input advertisements default to compact. Set REA_MCP_INPUT_SCHEMA_PROFILE=full for complete advertised guidance and nested constraints.
Name REA_SWIFTC_COMPILE_TIMEOUT_MS, its effective milliseconds, the
1..2147483647 range, and the omitted 120000 ms default when a Darwin
ownership-helper compile times out. A change applies after REA restarts,
on the next ownership-helper compile.
Producer SIGXCPU and SIGXFSZ recovery says REA retains the tighter
inherited limit. Agent-visible fixed caps in this theme say this release
has no runtime override. Numbers are unchanged.
Refs #5910
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
Bind artifact lifecycle and analysis calls to their advertised named SDK
output schemas, sharing discovery per connection and rebinding on reconnect.
Reject schema overrides and ambiguous catalogs, preserve request options,
and retain primary failures while attempting all owned cleanup.
Add real-stdio boundary regressions for malformed output, discovery reuse,
reconnection, schema ambiguity, override rejection and timeout preservation.
Assisted-by: Reimu (gensokyo-skills:reimu-incident-triage)
Assisted-by: Kogasa (gensokyo-skills:kogasa-surprise-testing)
An Electron scenario stopped at the first failed or cancelled action and
returned no record for the remaining declared actions, so
normalized_result.actions could not account for the whole scenario.
Record every remaining action as cancelled with a not-executed error, the
representation browser scenarios already use. Both runners now share one
helper for that error, which also distinguishes a stop caused by request
cancellation from one caused by an action failure.
Fixes#5943
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(browser)!: keep malformed console argument inventories unknown
An omitted, null, or non-array Runtime.consoleAPICalled args inventory now
reports argument_types as null, text capture as unavailable, and
console_events as invalid_protocol_value instead of a complete empty
inventory. Malformed argument elements keep their CDP position as
"unknown", so retained text values keep their original argument_index.
Call-site attribution now uses the frame at position zero: a malformed top
stack frame is no longer replaced by its caller. Console events are
excluded and request initiator locations are null, both reported as
invalid_protocol_value.
BREAKING CHANGE: console event argument_types may now be null, and
console text_capture.status may now be "unavailable", when the CDP
producer omits or malforms the argument inventory.
Fixes#5934
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(browser): report malformed WebMCP registration stacks as incomplete
topCallFrame now owns the omitted-versus-malformed decision for the whole
stack value: an omitted stack has no call site, while a non-object stack,
non-array callFrames, or a malformed top frame is malformed. WebMCP
discovery keeps the tool with a null registration_source but records
webmcp_tools invalid_protocol_value, so capture comparisons stay unknown
instead of confirming a changed source. Console and initiator stacks use
the same rule.
Refs #5934
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(capture): keep excluded MIME protobuf applicability partial
Automatic protobuf framing treated an explicitly excluded, conflicting or
unavailable Content-Type the same as a known-absent one, so a selected body
was skipped and coverage still upgraded to complete. Body metadata now
reports whether MIME applicability is established; an unresolved non-empty
body adds a record limitation and keeps protobuf_coverage partial. HAR uses
its required mimeType, falling back only to a visible header value.
Fixes#5926
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(capture): compare repeated Content-Type declarations by essence
Repeated Content-Type headers that differ only by case, whitespace or
parameters declare the same media type, yet the raw-string comparison
collapsed them to an unknown value and the new applicability check
reported partial coverage. Body metadata now compares media-type
essences and owns automatic framing selection, so the consumer no longer
re-parses the raw value; framing lookup ignores inherited properties.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
fix(adb): derive push digest source from the parsed device digest
A zero-exit sha256sum whose output held no recognizable digest produced
device_sha256: null with device_digest_source: "device_sha256sum" and
dropped the not-device-side-verified limitation. The push receipt now
takes one tagged device digest, so the source can only name sha256sum
when a digest was parsed; unrecognized output reports "unavailable",
the unverified limitation, and the raw command output. The sha256sum
parser moves to AdbShellOutput and accepts GNU's escaped-name prefix.
Fixes#5924
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(firmware): keep a missing producer report distinct after exit 0
A firmware engine that exits 0 without writing its requested report now
fails with the observed report state (missing, empty, invalid_json,
oversized, not_regular_file), the command, arguments, exit status and
captured output, plus report-specific recovery instead of generic
`rea doctor` advice. REA still does not retry or substitute an empty
result.
Fixes#5876
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(firmware): keep missing-report recovery from replaying removed paths
The recorded command arguments name REA's private workspace, which is
removed after the attempt. Recovery now presents them as provenance and
tells callers to replace every private path when reproducing the run.
Cover the oversized and not_regular_file report states.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
fix(errors): derive fixed-allowance recovery from a typed refusal field
inspect_plist refused a binary plist above its fixed 64 MiB decode
estimate, or nested deeper than 128 levels, with resource "memory" and no
recovery, so the shared presentation told the caller to add worker or
host memory. Neither raises that allowance. The 64 MiB source-size refusal
likewise got the worker file-size advice for evidence replies.
AnalysisResourceConstraintError now carries a typed fixedAllowance
{ name, kind }. fixedAllowanceOptions sets it instead of a prebuilt
string, and the presentation derives the recovery from it, so every
existing fixed-allowance producer moves to the same field. Operation
re-attribution, IDA startup re-attribution and explicit redaction copy the
field through one resourceConstraintRecovery helper; redaction still
applies to the allowance name.
The plist decode, nesting and input refusals use the field. A nesting
refusal now reports maximum_nesting_depth instead of the decode byte
limit it did not exceed.
Fixes#5847
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Observe private-runtime ACL ownership, trustees and rule count from one SID
rule collection. Guard the account-name presentation member through the owning
PowerShell type and verify that guard is installed. Keep the original15s probe
deadline and every ACL/readonly assertion.
Record bounded phase diagnostics and document the observed failure boundary.
The repeated historical timeout cause is unconfirmed; no retry or deadline
increase is presented as a fix.
Validation: complete current-head CI passed40 jobs with one source-only skip;
real Windows typecheck, curated/native controls, package and Node24/22 checks
passed. PowerShell7.5.2 API/parser experiment and278 routing regressions passed.
Full evidence: https://github.com/morluto/rea/actions/runs/39599008194
Assisted-by: Reimu (gensokyo-skills:reimu-incident-triage)
summary.by_evidence_authority bucketed each obligation's source_state, so a
shipped-artifact and a controlled-replay ledger returned the same summary and
the promised authority breakdown was unanswerable.
Derive the buckets from authority_references[].authority using the Evidence
authority vocabulary owned by the Evidence schema. An obligation counts once
under each distinct authority it cites, so counts can exceed total. Keep the
source-state histogram under its own name, by_source_state. The readiness
example now derives its summary from the same owner function instead of
hand-writing buckets.
BREAKING CHANGE: summary.by_evidence_authority keys are now Evidence
authorities (shipped-artifact, controlled-replay, historical-reference,
external-service, analyst-inference) instead of source states. The previous
state counts move to the new required summary.by_source_state. Ledger closure
digests change accordingly.
Fixes#5838
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): count one IPC observation per site-resolved channel
A send with a local constant channel was recorded twice: once unresolved
by the syntax pass and once resolved by wrapper substitution, because
duplicates were keyed on the channel. A channel the call site determines
now resolves that observation in place. A channel supplied through a
wrapper argument stays a candidate beside the unresolved observation.
Invocation identity now follows the call that supplies the wrapper's
literal arguments, so one wrapper call reached through several entries
is recorded once.
Fixes#5710
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): keep wrapper invocations distinct while removing repeats
Taking the nearest literal call as invocation identity collapsed separate
outer calls that pass different handlers through one wrapper call. Entry
identity is restored, and a finding is a repeat only when an identical
finding was reached through the same trailing local calls. A registration
replaces its unresolved source observation only when the site also
determines its handler, so a parameter handler keeps the generic source
registration beside per-invocation candidates.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): compare site-determined IPC handlers by value
In-place resolution compared handler representations, so a constant
channel with a named local handler (recorded as an identifier by the
syntax pass and as its function by substitution) stayed counted twice.
The handler is now site-determined when it evaluates to the same function
in the site's lexical environment. A replaced observation also removes
the syntax pass's unknown-channel contribution, so a fully resolved file
is no longer reported as partial.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): keep closures over invocation values as distinct IPC handlers
A handler node created in different wrapper activations may capture
different values, so the same function node alone does not prove the call
site determines the registration.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(javascript): guard optional call paths in closure handler test
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): decide IPC handler identity from activation captures
A closure is one site-determined handler only when it reads no binding of a
function this trace activated, such as a wrapper parameter or factory local.
Unused bindings in the activation no longer split a known listener, and
closures over outer-call values stay distinct when a shorter entry traced
the same wrapper call.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): treat implicit captures and bundle modules precisely in IPC handler identity
An arrow handler reading its enclosing function's arguments or this captures
that activation. A bundle module factory runs once per module load, so its
bindings are shared by every trace rather than per-invocation captures.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): attribute an arrow handler's inherited context to its owner
An arrow inherits this and arguments from the nearest non-arrow function, so
an activated arrow inside that owner does not make the handler depend on
the traced call.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): skip suffix scans for per-invocation IPC handlers
Check activation capture before scanning recorded call chains, keep the
chains of capture-dependent handlers out of the index, append chains in
place, and treat an arrow's inherited super and new.target as captures.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(javascript): supersede nested-entry IPC records and share module-init factories
A shorter entry traced alone no longer swallows the outer calls that invoke
the same wrapper sequence: the first fuller path takes over its record and
separate outer calls stay distinct invocations. A factory activation called
at module scope runs once per load, so its closure is shared unless the
trace itself makes that call. The resolver now owns bundle module-factory
recognition instead of an environment marker.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
resolve_adb_packages marked coverage partial when pm list packages printed
unrecognized rows, but dropped the rows themselves and could still emit a
limitation describing an unparsed_lines field the result never carried.
ADB parsers now own unrecognized-line facts as a bounded sample (first ten)
plus the exact total, and a single provider projection derives
unparsed_lines, coverage, and the truncation limitation from those facts for
every parsed listing and the package search. The resolver result schema now
carries unparsed_lines, matching list_adb_packages through CLI and MCP.
Fixes#5843
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
fix(artifacts): keep unknown inventory traversal out of complete inspection coverage
Artifact inspection derived its coverage status only from always-zero
omission counts, so an inventory substep with unknown container-entry
traversal (for example, an omitted FIFO) yielded
complete-within-substeps. Analysis-view summaries also exposed the
inventory's typed source_coverage only for PCK and PyInstaller roots.
Derive the inspection status from the inventory substep's coverage
(unknown traversal is partial), and carry the producer's source_coverage
plus the traversal-unknown caveat in every artifact summary. Hash-count
fields keep their existing format gating.
Fixes#5837
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
inspect_signal_decoders turned sigrok's explicit missing-decoder refusal
into execution_failure with rea doctor advice, while decode_signal_recording
already reported the same refusal as invalid_request on decoders.
The sigrok selector classifier now owns the unknown-ID match for both
selections: help names the decoder field and points to catalog discovery,
decoding keeps decoders/annotations. The provider passes each producer
command's caller selection to that one owner instead of per-call closures.
Installed decoders that fail to load and refusals naming other IDs remain
provider failures. The real-sigrok verifier now asserts the typed refusal
for both operations.
Fixes#5849
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Use pinned tsrs on supported development hosts while retaining stable official
TypeScript on Windows and unsupported platforms. Partition compile/typecheck
caches by actual compiler and qualify fallback hosts for compiler inputs and
lifecycle hooks.
Reduce PE acceptance startup through the existing bounded compiled CLI. Retain
upstream named MCP definitions, every fixture/assertion and the original 30s
deadline. Verify that browser transport loss preserves the observed target.
Validation: full qualification and real host lanes in
https://github.com/morluto/rea/actions/runs/39590450889 ; 450 focused regressions
passed with two documented conditional skips. All 1724 production emit files
match SHA-256 and size across both compilers. Current-head Codex/Greptile review
found no actionable issues. Measurements and limitations are in the PR.
Assisted-by: Reimu (gensokyo-skills:reimu-incident-triage)
Expose per-call firmware analysis budgets with exact recovery guidance, shared
CLI/MCP validation and retained actual settings. Keep publication and fixed
resource policy independent; preserve typed interruption and one captured-output
owner through command and one-shot cleanup.
Fixes#5851. Executed verification, real-provider limits, known baseline issues
and authorized readiness override are recorded in #5863.
Assisted-by: Kogasa (gensokyo-skills:kogasa-surprise-testing)
* fix(sqlite): attribute generated-column evaluation failures to the selection
A selected virtual generated column whose expression needs a function the
read-only connection lacks (or raises an SQL error for a sampled row) was
classified by the worker's catch-all as a malformed database and projected as
an invalid_format issue on `path`.
Row reads now attribute SQLITE_ERROR from selected cells by probing each
selected virtual generated column over the sampled rows, and report a
typed capability_unavailable refusal naming the table, each failing column
with SQLite's message, and the remaining columns to select. Other causes,
including corruption, TOOBIG and NOMEM, keep their classification.
Fixes#5719
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(sqlite): recommend only probed-readable columns after evaluation failures
Record every per-column probe failure with its own reason, discover stored
columns outside an inventory-scope selection without decoding unrelated
malformed names, and present recovery as the structured columns selection
instead of shell-quoted command text built from database-supplied names.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* docs(specs): plan a template rule that shares templated positions
Record the outcome, requirements, non-goals, plan and tasks for
prefix-variance, an opt-in environment-manifest rule that templates a
literal segment with the parameter another tool already uses at that
position, so a session's writes to one entity share the read's template.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(analysis-view): share templated positions across manifest tools
Add the opt-in prefix-variance template rule. It runs after the other
opt-in rules and templates a literal at a position that another tool,
of any method or segment count, templates by one of those rules after
the same preceding parts, when the literal is one of that parameter's
observed values or carries a digit and is spelled like one. The
parameter takes the shared parameter's name, tools that now match merge,
and a merged position keeps a tool's own rule when one templated it.
A write to the one entity a session edited, such as
PATCH /api/v1/collections/0RZ576QZQA5KG, becomes
patch_collections_by_collection_id on the read's template instead of a
literal single-entity tool.
Tool naming moves to its own module, and the three comparing passes
share the 32-segment split. Manifests that do not list the rule are
unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* refactor(analysis-view): drop the unused ManifestRule alias
Parameters now type their rule as a format rule, a shared opt-in rule
or prefix-variance, so nothing reads the union alias.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(specs): record the prefix-variance checks and corpus results
Record the gates, and the 50 showcase captures with and without the
rule: identical output without it, 25 write and 34 read tools templated
with it, and the write tools the write-plan branch serves on top.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(analysis-view): share positions with digit-free generated identifiers
prefix-variance accepted an unobserved literal only when it carried a
digit, so a generated identifier that happened to hold none stayed a
literal tool: an n8n workflow id such as FkMwxmXNGaGVsXFG, whose reads
then never joined the item view. A literal now also qualifies when every
value observed at the position has one length of eight or more characters
and the literal has that length and uses only their character classes.
Across the 50 showcase captures this templates 10 more read tools in
Memos and n8n and no words.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(specs): record five environments rebuilt end to end with the rule
Five randomly drawn environments rebuilt on VMs with the rule and the
write-plan branch: smoke, retrieval and write tasks for two models, and
the same calls against the live applications. Record the results, the
misses and their causes, and the corpus numbers after the fixed-length
clause.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(analysis-view): share a position with the parameter that observed the literal
A shared position folded every donor's values into the earliest donor's
parameter. With an opaque-token read before sibling-variance reads at
the same position, a PATCH of /api/items/b2 qualified on the sibling's
evidence but became /api/items/{uid} instead of joining the item read's
/api/items/{item_id}. Each position now keeps one donor per naming rule
with its own values, and a literal takes the donor that observed it,
else the earliest one it is spelled like.
Each donor's spelling (length and character classes of its values, and
its fixed-length family) is summarised once per position, so a literal
is checked in time independent of how many values were observed, rather
than copying and rescanning them for every candidate.
A family now forms from one observed value, as the requirement says, so
a digit-free id of the same length shares a position that only one
opaque-token or after-uid-literal tool supplied. The 50 showcase
captures give the same tools as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(specs): record the prefix-variance review round
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(analysis-view): bound the spelling summary of shared positions
spellingSummary classified every observed value with three full-size
replaceAll copies and a set per distinct character, so a donor holding
a multi-megabyte segment cost unaccounted memory and time in finish().
Characters are now classified in one pass, and a value longer than 128
characters takes no part in spelling comparisons: it shares a position
only as an exactly observed value. The 50 showcase captures give the
same tools.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Expose a per-call SQLite worker deadline with exact CLI/MCP recovery controls,
parent-owned execution metadata and historical unknown preservation. Keep fixed
resource limits and cleanup admission intact. Share typed cleanup composition
with one captured-output owner and preserve the timeout after cleanup resolves.
Fixes#5812. Follow-up tracking: #5808. Existing CLI test-oracle failure: #5845.
Verified 155 SQLite tests (3 skipped), static/docs/discovery checks, six real
CLI/MCP pairs and 13 actual Codex sessions including four corrected deadline
recoveries. Three pre-dispatch discovery failures remain explicit, not passes.
Review findings include real retained-cleanup and fixture ownership regressions.
Assisted-by: Kogasa (gensokyo-skills:kogasa-surprise-testing)
Expose the existing WABT command deadline as an optional validated per-call CLI/MCP budget, preserving its 30000 ms default and fixed safety allowances.
Retain selected execution profiles in success and rejected-module Evidence. Keep typed timeout recovery through cleanup failures without duplicating command output or making incomplete cleanup retryable.
Verified focused tests, real WABT CLI/MCP workflows and actual Codex recovery/blocked controls; detailed measurements and qualification limits are recorded in issue #5797. Merge uses the explicitly authorized maintainer override with pending CI and unchecked human-review attestations disclosed on PR #5798.
Closes#5797
Assisted-by: Kogasa (gensokyo-skills:kogasa-surprise-testing)
* perf(ci): pack each test tier from its own timing history
Fast shards were still ordered with the latest full-suite timing artifact,
which stayed stale across compiled-CLI session reuse. Publish and download
test-timings-fast and test-timings-full separately, and keep the legacy
artifact as a fallback until a tier has its own history.
* perf(ci): drop the shared timing-history fallback
Shard lookup now reads only the artifact for the current tier. A missing
artifact leaves history unset instead of packing from the old test-timings
file.
* test(ci): look up fast history when only full timings exist
The empty-history case now requests the fast tier against a full-tier
artifact and checks that the fast lookup ran and returned nothing. A
missing tier stays a separate test that does not query.
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Decode direct Objective-C class names separately from nominal descriptors, retain indirect class bindings as partial facts, and preserve readable protocol and witness evidence when related bytes are unavailable.
Validated with 98 focused tests, check:ci, docs:check, and named-schema CLI/MCP replays of compiler-produced imported-class and struct fixtures plus an inert kind-3 control. The fixtures were inspected statically, never executed.
Fixes#5738
Distinguish CDP session detachment, observed crashes, and transport loss. Preserve the observed window alongside the original typed coverage-stop failure and unconfirmed cleanup.
Validated with 40 focused boundary tests, check:ci, docs:check, and the real Chrome public CLI/MCP verifier on macOS. Browser timing comparison remains incomplete because unchanged-main fixture readiness was intermittent.
Fixes#5756
Decode native binary plists from bounded owned source bytes before conversion can erase scalar values. Share scalar preflight and isolate overlapping decoder records while retaining exact integers, date uncertainty, and original bytes. Refuse numeric payloads that read structural table bytes, and avoid retaining projection metadata when callers only need structural validation.
Closes#5701Closes#5703Closes#5705Closes#5782Closes#5783Closes#5792
Preserve unsupported GGUF version bytes and both UINT32 interpretations without inventing byte order at the shared parser. Keep existing layout support, typed errors, evidence identity, scoped reads and resource limits.
Validation: 78 focused tests, full static/architecture and documentation checks, real CLI/MCP version controls and an independently checked public model. Latest-head Codex found no issues; Greptile is 5/5 and all review threads are resolved.
Some CI remained pending at merge qualification. Merge under the human owner's explicit standing bypass authorization does not claim the full CI suite passed.
Fixes#5749.
* fix(environment): report later bundle write failures as host failures
Only destination admission now refuses the selected directory as invalid
input. A path failure on a nested bundle file after admission, such as a
host pathname-length limit, is reported as an output failure that names the
host cause and the absolute failed path, with recovery guidance for the
partial folder it leaves in place.
Fixes#5522
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): name the host-refused path in export permission failures
A permission failure after earlier bundle files were written was reported
as access denied for the bundle root, so the documented failed path was
lost. The refusal now names the absolute path the host reported, falling
back to the root, and keeps the root as `output_directory`.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): attribute pathless bundle write failures to their file
Descriptor operations such as `write` and `fsync` fail without a path, so
a disk-full failure after earlier files named only the bundle root. The
bundle writer now attributes such failures to the file it was writing.
Host-write recovery is applied only to actual host failures; an integrity
refusal keeps its own message without disk-space advice.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(environment): attribute package tarball read failures to the tarball
The destination-write attribution also caught failures from reading the
admitted package tarball during the copy, so a source `EIO` named the
destination file and advised freeing disk space. Host failures are now
attributed by direction: tarball reads name the tarball with read
recovery, bundle writes name the file being written.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(environment): state which export failures name a path
The export documentation and host-write message promised a failed path
for every later write failure, but a directory sync at commit time fails
without one. They now promise the failed bundle file or package tarball
when the failure is on one, and only the cause for a folder sync.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(apple): refuse invalid binary dates and unknown sub-millisecond instants
Reachable binary plist dates must use marker 0x33. Other type-3 markers are
malformed and are no longer projected as observed timestamps. A valid float64
date stays known only when it is exactly the canonical millisecond instant;
other finite instants remain typed unknown dates with a counted limitation.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(apple): share one decoder copy and bound binary date payloads
Wide-integer rewrites and unrepresentable date payloads now share the
single parser-input copy already reserved by the decode budget. Reachable
date markers and payloads must end at the offset table even when other
decoder-bounds checks read trailing bytes.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* test(apple): refuse out-of-region dates through the keyed reader
Decoder-bounds date truncation is now asserted on decodeKeyedArchiveBytes,
the path that selects those bounds.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
Go, IL2CPP, dylib resolution, and application-workflow tests were still
starting a compiled CLI for every command. They now reuse one process
per file. Warm focused runs dropped from 40s, 46s, 35s, and 21s to
14s, 21s, 20s, and 2s.
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
* perf(test): reuse one compiled CLI for export-binding parity
The slow export-binding cases each spawned the compiled CLI three times.
One warm process now serves those JSON commands from a single command
instance, so startup and retained registrations are paid once per file.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* style(test): format the warm CLI session
oxfmt --check rejected the session and its worker.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(test): report warm CLI shutdown, deadlines, and exit status
A compiled CLI session that needed a signal after stdin closed reported
success. A command that never answered left later checks waiting, and a
failure recorded only on process.exitCode was returned as status 0.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(test): close the warm CLI worker when a status check fails
The failure-then-success case only closed the compiled worker after both
assertions passed. Register the close with the test so a failed check still
stops the child.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(test): retain a timed-out compiled CLI worker until it exits
A request deadline cleared the child before SIGKILL was observed. The next
command could start a replacement, and close could miss the original process.
Retired workers stay owned until exit, and a failed stop remains for a later
close to retry.
Co-authored-by: morluto <morluto@users.noreply.github.com>
* fix(test): reject a compiled CLI launch that close interrupts
Closing during the ready handshake detached the child before the launch
promise settled. The queued command then waited on the startup timer.
Close now rejects that startup before the child is detached.
Co-authored-by: morluto <morluto@users.noreply.github.com>
---------
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: morluto <morluto@users.noreply.github.com>
morluto