inspect_native_instruction reports `not-in-listing` when an address lies
in Ghidra's memory but the listing has no instruction or defined data
there. No decode is attempted, so the answer does not show the bytes are
data, yet the result named no way forward. #10623 added the declared-seed
recovery for procedure_not_found and not_in_procedure; this is the
instruction-level sibling, where a `code` seed applies.
A not-in-listing answer now carries ghidraNotInListingLimitation(source).
It names a `code` seed (or a `function` seed for an entry point) in
REA_GHIDRA_SEED_FILE, says which setting to unset first for a selected
project or checkpoint, and does not assert that the address holds code.
decoded, data and other statuses are unchanged. The installation guide's
seed section lists the new carrier.
Validation: src/ghidra/GhidraProvider.behavior.test.ts fails 1/298 with
the caveat removed and passes 298/298 with it; check:fast passes. No
real Ghidra run.
Refs morluto/rea#3767
A search_bytes cursor resumes the same target, profile and query, and
the limitation said only that. In a real session, renaming a function
between pages changed the procedure name on the next page from the same
cursor, while the page Evidence kept the same target and profile digest
and no limitation explained the difference.
The search_bytes limitation now states that the cursor does not pin
match metadata: procedure names and other match metadata are read live
on each page, so a rename in the session between pages appears on later
pages. This takes the issue's "identify that match metadata is live"
option; binding or qualifying pages to a metadata version is left open.
No prose-asserting test was added, following the direction in #10508.
Validation: src/ghidra/GhidraProvider.behavior.test.ts 296/296;
check:fast passes.
Refs morluto/rea#8365
The npm package does not ship docs/, so a repository-relative guide path
in doctor's remediation could not be opened by npm users. Link the guide
on GitHub instead.
Refs #7018
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
The Ghidra requirements said official archives may require building the
Linux arm64 decompiler but gave no steps, so doctor's new pointer to that
section led nowhere. Ghidra's own guide lists Linux ARM64 among the
platforms that need user-built natives; link its Building Native
Components section and state it plainly in the remediation too.
Refs #7018
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
Official Ghidra release archives ship no macOS decompiler and may ship
none for Linux arm64, but doctor's native_decompiler remediation only
named the expected path. On those hosts it now says the archive's
decompiler source builds it and points to the installation guide's build
steps for that platform.
Fixes#7018
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
An empty or whitespace-only --client or --skill-client value normalizes to an empty key, which is a prefix of every id, so suggestion ranking named an arbitrary client. Skip the suggestion for an empty key and report the supported list.
Setup and doctor resolve --client and --skill-client values by trimming,
lowercasing and collapsing runs of '-', '_' and whitespace, then matching a
canonical id or an alias the client row already declares for 'rea mcp add'.
An unknown value now reports 'Unknown client' with the closest id from
suggestCommand, or the supported list when nothing is close. The error code
and field path are unchanged.
A SIGXCPU at the requested budget is the decoder's own ceiling. Raising RLIMIT_CPU does not extend it. A lower recorded ceiling remains an inherited limit. A failed close of the CPU record stays on the decoder scope and does not skip that classification.
Static checks reject a second AnalysisResourceConstraintError in CaptureFailures.ts. SIGXCPU still reports the recorded ceiling and tells the caller to raise RLIMIT_CPU. The CPU record is read through the stable artifact owner, so a failed close stays on the decoder scope.
A tighter inherited RLIMIT_CPU is kept, and that death had no reply. The decoder records the installed ceiling, and SIGXCPU becomes a CPU resource constraint that names it.
A tighter RLIMIT_FSIZE was installed, but the reply check still used the
requested budget. A reply under that budget could exceed the inherited
limit and truncate. The effective cap now bounds the write, and a reply
that does not fit returns a fixed file-size limit. Tests also keep a
higher hard limit distinct from the new soft limit.
setrlimit ran before the shutdown guard and asked for hard limits the
process may not raise. A tighter inherited hard limit then left mitmdump
running until the capture timeout.
A failed resolveTarget returned after a boolean assertion, so the setup
failure showed as a mismatch instead of the typed error. Format the long
AnalysisOutputError line that static checks rejected.
A failed first open whose client cleanup also failed cleared a staged
import even though no target had been retired. Clearing now waits until
this open has retired a previous target. A later open still retries that
cleanup before another profile can start.
* fix(desktop): claim agentenv capture exclusively and survive FIFO write errors
stop() used makedirs(exist_ok) before the control FIFO existed, so a
start already in that window kept running after stop reported the
capture absent. A broken or full FIFO also aborted stop and mark
instead of reporting that the capture was not reading.
* fix(desktop): retry agentenv stop until the reader accepts it
stop() treated the first failed FIFO write as final and then started a
second full timeout. A start() that has created the FIFO but not opened
it kept running, and an interrupted startup could take twice the
caller's timeout.
* fix(desktop): remember an absent agentenv stop for a retry
stop() claimed a missing directory and returned absent. A retry found
that directory and waited for a FIFO that start() will not create, so
the build-computer release could not clear the folder. The claim is a
file in the directory, and a later stop returns absent.
* test(lifecycle): name the renamed JEB and BEAM coverage tests
The full source suite still expects titles removed by the JEB catalog consolidation and the BEAM CLI consolidation. Point those entries at the tests that replaced them.
State the second pinned layout, its word table and how admission reads
the payload length for each layout. Record that the six-word roles come
from the V8 sources and that runtime compatibility with a V8 8.8 through
11.7 producer remains unverified.
No V8 8.8 through 11.7 producer runs in the test environment, so the
fixture is synthetic, as it already is for the narrow header: a real
cache without its build-constant word and its alignment word, which is
the container those releases define.
Cover the admitted layout through the parser and the public CLI, the
bit 31 module flag, a truncated six-word container, a container that
fits both layouts, and the word-table rule for a 24-byte header. The
short-header case now also covers the new 24-byte minimum.
The CLI case is the recorded lifecycle coverage of the provider, so its
new title is updated there as well.
The reader pinned the seven-word header that V8 11.8 and later write and
took the payload length from the word at 0x14. V8 8.8 through 11.7,
which Node 16 through 20 embed, define six words: there is no read-only
snapshot checksum word, the payload length is the word at 0x10 and the
header is 24 bytes on both pointer widths. For such a container the word
at 0x14 is the payload checksum, so a complete cache was refused as
truncated or corrupt.
Admit the six-word layout when the word at 0x10 reaches the end of the
file from a 24-byte header, and report its six words. A container that
fits both layouts is refused as ambiguous. The six-word roles come from
the V8 sources of those releases, so only the magic and the
payload-length relation are reported as confirmed, and the result states
that the layout was not measured against one of their producers.
A container that fits neither layout is no longer described as certainly
truncated or corrupt, since an unadmitted header has the same effect.
`file -b` reports `data` for an XML plist containing a control character
or NUL byte, which CoreFoundation still reads. plistSourceFormat only
probed the source for markup when `file` reported text, so these sources
were labelled `unknown` and skipped REA's XML reader. When that reader
would have refused them, the result also lacked the limitation that says
the value came from plutil's conversion. Two plists shipped with macOS
27.2.1 are affected.
Probe the source prefix for any description that does not name binary,
XML or JSON. Such XML now takes the same route as XML that `file`
recognizes: REA's reader decodes it directly, or #12251's plutil fallback
records REA's refusal reason. OpenStep and other non-markup sources stay
`unknown`.
dyld treats an image as restricted only when a __RESTRICT segment holds a
section named __restrict. inspect_binary_hardening reported restrict_segment
enabled for any segment named __RESTRICT, so a binary linked with
`-sectcreate __RESTRICT __other` was reported restricted although dyld still
honours DYLD_* variables for it.
The segment reader that already walks every section header now also returns
the first __restrict section. The row is enabled only for that section and
points at its name; a __RESTRICT segment without it is disabled with a basis
that names the missing section and points at the segment name.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
OpenPGP packet headers read their one-, two- and four-byte lengths with
raw Buffer reads after the header-width checks. Read them through an
openpgp BinaryReader. This was the last counted raw numeric read, so the
raw-buffer-read baseline is removed and the guard now refuses any new
copy outright. The CI change-scope test names another baseline file as
its static-only path.
Refs #11865
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
BMPString and UniversalString decoding read code units with raw Buffer
reads after their length-multiple checks. Read them through the
pe-certificate-table BinaryReader instead.
Refs #11865
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
The fixed-string custom-attribute prolog and the embedded resource length
were raw Buffer reads after their length and directory checks. Read them
with the managed BinaryReader formats the other metadata readers use.
Refs #11865
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
The Objective-C method and Swift relative-pointer decoders read 32-bit
displacements straight from the image bytes, beside the checked u32 and
pointer readers. The shared dispatch readers now provide i32 through the
image BinaryReader and both facets use it, so they no longer take the raw
image bytes.
Refs #11865
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
The build-version platform and tool count were read through a local
DataView after the command-size check. Read them with machoWordReader,
as the other load-command decoders do, and drop the now-unused viewOf.
Refs #11865
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>
readGnuPropertyWords decoded property type, size and feature words with
endian-selected raw Buffer reads. Every read already sits behind the
record length checks, so the shared reader returns the same values.
Refs #11865
Co-Authored-By: gensokyo-dev <342580529+gensokyo-dev@users.noreply.github.com>