The tool table probed the bare name, and coreutils' hard-link utility owns that name on
every host that has it -- including the Git Bash and MSYS installs on Windows, where the
row would be trusted most. The probe therefore reported the MSVC header toolchain present
on a host with no MSVC at all, and the present-tool count was one too high.
The row now probes editbin, whose name has no Unix collision, and its note says that
link.exe is not probed by name. strings stays in the table: binutils strings reads PE
images, so that row is true on a Unix host, and the test carries the reason the
distinction is per-name rather than per-existence.
tests/unit/test_hwprobe_tool_table.py pins three things: no row probes a name from the
ambiguity list; editbin is the row that carries header edits; and, as the general form,
every probed tool that resolves on this host is run with --version and must not identify
as GNU coreutils. FINDINGS F17 records the defect and how it surfaced -- by reading the
probe's own output while collecting host facts for a protected-sample run, where a row
claiming the MSVC linker sat beside gdb, a MinGW gcc and no cl.
A captured image often has no import data: the descriptor is gone, the IAT directory is
gone, and nobody took a write watch. The table is still findable, because the rewrite left
a relation behind -- the slot holds the address the call targets -- so the call sites
become the search key.
discover_iat scans the non-executable sections for pointer-aligned entries equal to the
target of a relative call that leaves the image, then clusters the matches by proximity.
One cluster is used and its span becomes the inventory. Two are printed and refused: a
duplicated table, a copied table and a table split by a section boundary all look like
this from the outside, and picking one produces a file that loads and calls the wrong API
for one import. A single match is called weak rather than used.
Discovery runs when the described slots explained none of the calls, not only when the
inventory came back empty -- a protector that leaves a plausible-looking but useless
descriptor is the case the narrower rule missed. --iat suppresses it because that is the
reader asserting the range, and --no-discover is the control that keeps "the search
worked" from being a statement about the fixture.
walk_sites and pair_sites replace the single scan-and-pair pass, because discovery needs
the targets before any slot exists and one function could not express that order.
decide_result gains ambiguous-iat, which exits 1 like the other two verdicts that mean
this run did not finish the job.
Tests: the same image with its slots outside every directory recovers the two pairs the
explicit range produced; the control with --no-discover still reports needs-trace; a
second copy of the table in another section is refused with its own token; and a search
that matched nothing says so instead of reading as a clean image.
The reference gains the search, its two refusals, and the pair of failures that look
identical from the report -- slots holding tags have no value relation to search, and a
wrong base finds nothing either, so the base is checked first.
An empty pair list had two causes and one verdict. When the run never found the import
table at all, the report said needs-trace, which sends the reader off to write a trace of
the protector's resolver -- the right move for a table full of handles and the wrong move
for a table this run simply never located. The misdiagnosis is the expensive kind: it
costs a debugger session and it looks like progress.
decide_result now returns no-slots when the inventory is empty, and the printed hint says
which of the two it is (supply --iat, or reconstruct the table). no-slots exits 1 with
needs-trace, because the run did not do the job it was asked to do; the docstring's exit
code table says so.
The reference gains the matching sentence in the inventory step: an inventory that came
back empty is a finding about the run, not a table that holds no addresses.
Test: test_no_slot_values_is_its_own_verdict pins both sides -- no slots plus unpaired
calls is no-slots, some slots plus unpaired calls is still needs-trace -- and the existing
token test now carries a slot count so it cannot pass for the wrong reason.
references/iat-virtualization-and-call-site-rewriting.md records the two edits one
protector family makes to a native image: the import slot stops holding an address and
the indirect call that read it becomes a relative call to a stub. It carries the pairing
rule -- a site's target equals a slot's value -- the three placements of the filler byte
in a six-byte rewrite and the refusal when there is none, the divergence between builds
that modify the code section and builds that only redirect the table, the small set of
imports redirected twice, the choice between repairing, extracting a payload, observing
and reporting, and the limits an automatic unpacker inherits from its own dump.
scripts/iat_callsite_restore.py implements the mechanical half offline: inventory the
slots from three sources, pair by value, decide the six-byte window per site, refuse an
ambiguous slot and a site whose sixth byte is the next instruction, rewrite, then
re-decode every written site from the file it wrote. It is report-only unless --out is
given, and it does not unpack: a virtualised entry has no site list to restore.
Registered in both symptom tables, the routing, reference and script indexes, the
capability registry as call_site_restore, and the coverage, route-inventory and evidence
tables. dynamic-debugging.md gains the instrument rule the second write-up demonstrates
(a target that runs at full speed and dies when stopped has a perturbation detector),
memory-and-dumps.md gains the call-site step of a rebuild, environment.md gains the
compatibility-shim trap, and protectors-and-packers.md gains one recognition row plus a
pointer rather than the detail, because that file makes no product-level claims.
The target-behaviour claims stay unverified: no protected sample was analysed here. What
is observed is the tool's behaviour on a generated image -- both filler positions, both
refusals, both displacement forms -- and what the source material states about itself.
FINDINGS F16 records the defect this pass produced, a slot's magnitude read as its
meaning, with the test that fails without the fix. .gitignore gains ref/, which was not
covered by the existing rule because that rule carries a leading dot.
- check_budget.py escalated a hard index-row violation to 'warn' on both
branches, so INDEX_ROW_HARD could never fail anything. It fails now, and
tests/unit/test_budget_measurement.py pins both thresholds.
- The corpus note counted scripts/__pycache__/*.pyc, so the same commit
reported 27 scripts after the suite had run and 16 on a clean checkout.
- TOKENS_PER_CHAR is re-derived from a measurement (26,795 chars -> 6,065
tokens = 4.418); the comment's old figure and its arithmetic were both
wrong for this file.
- check_repo.py's yaml sentinel carries the type ignore, so mypy is clean
with PyYAML installed as well as without it.
- The CPUID test now skips with the probe's own note on a host without
CPUID instead of failing, and pytest.ini points at tests/README.md, which
exists, instead of an EXTENSION file that never did.
Tooling: .flake8 (a measured 120-column limit), .gitattributes, .gitignore
and .github/workflows/ci.yml (the five gates, the suite on 3.11-3.14 and a
lint job), which is what CONTRIBUTING.md already claimed CI runs. The
pull-request template it refers to now exists.
Content: pitfalls P26-P34 -- the analysis-level failures the catalogue was
missing (sampling read as observation, a self-built parser trusted, a
verification that could not have failed, a shell-mangled command line) --
registered in the symptom index and its routing mirror. FINDINGS F11-F15
record the defects above.
Tests: the coverage guard derives its subject from scripts/ rather than a
hand-written list, pe_triage.py and pe_diff.py gain their first tests, and
plan.py / preflight.py are declared uncovered with reasons.
root