The SECS-I Protocol FSM now enforces T3 (reply timeout) and T4
(inter-block timeout) directly, instead of leaving them as
upper-layer hooks.
T3: on complete_send, if the block we just acked had W=1, record its
system_bytes in awaiting_reply_sys_ and emit ActionStartTimer{T3}.
deliver_recv cancels T3 when a block arrives whose system_bytes
match the outstanding request. EventTimeout{T3} aborts the FSM with
"T3 reply timeout".
T4: deliver_recv emits ActionStartTimer{T4} whenever the delivered
block has end_block=false. The next block's deliver_recv cancels
the timer; EventTimeout{T4} aborts with "T4 inter-block timeout".
abort() now also cancels T3/T4 and clears the tracking state.
Test changes:
- Old "T3/T4 are FSM-level no-ops" test → REPLACED by four new
tests: T3 arm+expire, T3 arm+matching-reply cancels, T4
arm+expire, T4 arm+next-block cancels.
- Two new observer accessors on Protocol (awaiting_reply,
awaiting_next_block) so the tests can assert tracking state
without poking internals.
COMPLIANCE.md §1a: T3 + T4 rows go ⬜ → ✅.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
E84 (Parallel I/O) is fundamentally per-load-port: each port has its
own ten-wire handshake with the AMHS. Earlier revisions modeled it
as a single equipment-wide FSM; this commit refactors to a per-port
store, so multi-LP tools can run independent handshakes in parallel.
Public API change in EquipmentDataModel:
E84StateMachine e84; -> removed
E84PortStore e84_ports; // create(port_id), get(port_id), ...
Convenience pass-throughs: E84PortStore::on_signal_change auto-creates
the port on first use (ergonomic for demos); applications should call
create() explicitly with their full port set.
The two existing callsites (test_gem300_scenario, test_e87_wire_scenarios)
are updated. The multi-LP test now demonstrates the actual win:
interleaved LP1 load + LP2 unload handshakes that reach their
respective Ready states without sequencing, and an ES on LP1 that
does NOT affect LP2 — exactly the failure mode the previous design
couldn't catch.
Five new dedicated tests in test_e84_ports.cpp for the store itself.
COMPLIANCE.md §4i updated: row now reflects per-port design.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Connection now supports both HSMS-SS (single session — the
constructor's behaviour, unchanged) and HSMS-GS (multi-session).
add_session(device_id) registers additional sessions; each one has
its own NotSelected/Selected state and its own message/selected
handlers. In GS mode the Select.req carries session_id=device_id;
in SS mode it stays at 0xFFFF (legacy). Linktest/Separate remain
connection-scope per spec.
Public API additions:
add_session(device_id)
set_session_message_handler(device_id, h)
set_session_selected_handler(device_id, h)
session_state(device_id) -> State
is_session_selected(device_id) -> bool
send_request(device_id, msg, cb)
send_data(device_id, msg)
Internal refactor: state_/on_message_/on_selected_ folded into a
SessionSlot map keyed by device_id; SS-style getters/setters route
through the primary session. T7 + linktest are connection-scope —
T7 fires only when no session is selected; linktest runs while at
least one is.
Five wire-level tests:
- passive: two sessions selected independently via Select.req
with their own session_id
- GS Select.req for an unregistered session id is Rejected
(EntityNotSelected)
- data routed by session_id; data on a not-selected session is
Rejected
- active: two registered sessions both end up selected via
serialized Select.req per session
- SS legacy: existing single-session API still works (session_id
0xFFFF in Select.req)
COMPLIANCE.md §1 updated: HSMS-GS row goes ⬜ → ✅.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per-substrate transition history now survives restart. Each entry's
steady_clock timestamp is written as a system_clock-millis snapshot;
on replay the steady_clock time_point is reconstructed relative to
the current (steady_now, system_now) pair, so inter-event spacing
is preserved across restarts even if the FSM is in a different
process. Absolute wall-clock accuracy degrades by any NTP step
that happened between write and read; that's a documented caveat.
Record format goes v1 → v2. v1 (history-less) records still load,
just with empty history.
Test updates:
- the old "history is NOT journaled" test is REPLACED with one
that asserts every axis + event + label round-trips.
- hand-crafted v1 record on disk still loads (proves backwards
compat).
- 15 ms-spaced events restore with their spacing intact (±slop
for scheduler jitter).
Closes the "substrate history persistence" caveat from the post-#1-13
status writeup.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Closes the v1 caveat: the optional E40-0705 trailers on S16F11 —
recipe variables (RcpVar) and process parameters (ProcessParam),
each carrying a secs2::Item value of arbitrary type — now survive
restart.
Record format bumps to v2:
v2 header = v1 header
+ [u16 rcpvar_count][repeat: u16 name_len, name, u32 enc_len,
secs2::encode(value)]
+ [u16 ppparam_count][...same shape]
v1 records are still accepted by load_record_ (no extras come back).
Two new tests:
- round-trip mixed F4 / ASCII / U4 / nested-list values through
rcpvars + prprocessparams
- hand-crafted v1 record on disk still loads cleanly, just with
empty extras (proves backwards compat)
Closes the "PJ rcpvars / prprocessparams persistence" caveat from
the post-#1-13 status writeup.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per-EXID binary record (.ex), magic + version + atomic .tmp+rename.
Records full E5 §9 lifecycle: state, EXID, EXTYPE, EXMESSAGE, and
the candidate EXRECVRA list.
Cleared exceptions are terminal — the FSM transitions through
Cleared remove the in-memory entry AND delete the journal file
(matching the existing in-memory semantics). Recovering /
RecoverFailed states survive restart: the application can decide
on replay whether to retry recovery or abort.
Five new tests cover post+replay, Recovering-survives-restart,
autonomous-clear cleanup, RecoverFailed retry post-restart, and
corrupt-record drop.
This completes #12 in the test-gap backlog (persistence for the four
in-memory stores beyond Spool).
Closes#4 in the test-gap backlog.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per-job binary record (.pj / .cj) with magic+version, atomic
.tmp+rename. PJ store additionally writes an order.idx index file
that preserves HOQ-aware queue position across restarts.
Rcpvars / prprocessparams (secs2::Item variants) are intentionally
out of scope for v1 — they're optional E40 trailers and need a body
codec round-trip; callers re-populate via set_e40_extras() after
restart.
Five new tests cover full lifecycle replay (Processing mid-run +
HOQ-reordered queue), dequeue-deletes-file, corrupt-record drop,
CJ state + PJ-list replay, and CJ remove cleanup.
Closes#3 in the test-gap backlog.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Same pattern as carriers: per-substrate binary record (.sub) with
atomic .tmp+rename, replay on enable, delete on remove. Records
current state across all three E90 axes (location / processing /
ID-status), plus substid / carrierid / slot / free-form location
label. History is deliberately NOT journaled — it's an in-memory
ring buffer and rebuilding from replayed state would mislead.
Five new tests cover full-axis replay, every terminal processing
state, remove-deletes-journal, corrupt-record drop, and the
history-is-transient invariant.
Closes#2 in the test-gap backlog.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Mirrors SpoolStore: per-record file with atomic .tmp+rename, magic+
version-prefixed binary layout, replay on enable, delete on remove.
FSMs gain a restore_state() that bypasses the transition table and
handlers since a replay isn't a transition.
Six new tests cover write+restart+replay across every CIDS/CSMS/CAS
axis, remove-deletes-journal, malformed-record drop-not-poison, and
the persistence-disabled no-op path.
Closes#1 in the test-gap backlog.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
A host couldn't drive the new messages through the HostHandler class —
only the server side knew how to dispatch them. Adds six new senders
plus a unit test that walks each through a real loopback connection:
* send_legacy_remote_command -> S2F21
* send_event_report_request -> S6F15
* send_individual_report_request -> S6F19
* send_annotated_report_request -> S6F21
* send_pp_load_inquire -> S7F1
* send_delete_pp -> S7F17
Suite: 296 cases / 1571 assertions.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a docker-compose service `server-spool` that runs secs_server
with --spool-dir pointed at a named volume. Two-phase Python
harness (interop/spool_persistence_test.py):
1. Enqueue phase: force-spool one S6F11(CEID=300) via the
SPOOL_ON / START / SPOOL_OFF RCMD trio, then disconnect.
2. Driver runs `docker compose restart server-spool` between
the phases — the named volume preserves the journal files.
3. Drain phase: reconnect, send S6F23(Transmit), verify the
replayed S6F11 carries CEID 300.
Surfaces a real interop bug along the way: secsgem-py 0.3.0 encodes
RSDC (and other "single-byte status" fields) as <U1>, while SEMI E5
spells them as <B>. Our `as_binary_first` was strict on Binary; now
accepts either (the byte semantics are identical, and the leniency is
symmetric with the U-type widening from the first interop commit).
Result: enqueue → docker restart → drain returns CEID 300 cleanly.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds opt-in disk persistence to SpoolStore. `enable_persistence(dir)`
turns every enqueue into a single `<seq>.spool` file alongside the
in-memory queue; drain and clear delete the matching files; restart
replays the directory sorted by seq.
Writes are atomic: serialize the message via the SECS-II codec, write
to `.tmp`, then `std::filesystem::rename` to the final name. Malformed
records are dropped silently so a single bad file can't poison the
whole spool.
`secs_server --spool-dir <path>` enables persistence at startup.
Without the flag the behaviour is identical to before (in-memory only).
Two new tests: enqueue → restart → replay → drain restores the wire
order, and clear deletes the journal files.
Test suite: 291 cases / 1515 assertions.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Replaces the simplified <L,3 PRJOBID PPID MTRLOUTSPEC> demo body with
the full SEMI E40-0705 §10.2 shape:
<L,5 PRJOBID MF PRRECIPEMETHOD
<L,2 PPID <L,n <L,2 RCPPARNM RCPPARVAL>>>
<L,n MTRLOUTSPEC>
<L,n <L,2 PARAMNAME PARAMVAL>>>
ProcessJob now carries the extra fields (MaterialFlag, ProcessRecipeMethod,
RcpVar[], ProcessParam[]) so a tool's recipe engine can later consume
the recipe-variable overrides and per-job process parameters. Server
S16F11 dispatch populates them via the new ProcessJobStore::set_e40_extras
helper after a successful create.
MaterialFlag + ProcessRecipeMethod enums live in their own tiny header
(`e40_constants.hpp`) so process_jobs.hpp (the store) can use them
without dragging in messages_helpers.hpp (which would create a circular
include via data_model.hpp).
The simplified 3-arg HostHandler::send_create_process_job convenience
remains; it constructs a sensible-default PRJobCreateRequest internally.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds the SECS-II messages secsgem-py 0.3.0 ships but our C++ catalog
didn't have, plus the alternative wafer-map formats from E5 §13.
None of these were strictly required for GEM core compliance, but
they're the messages a host might send to a conformant equipment.
* S7F1/F2 — Process Program Load Inquire / Grant. Equipment-side
space-and-policy check before a host commits to S7F3.
* S7F17/F18 — Delete Process Program. Empty list = delete-all.
* S6F5/F6 — Multi-block Data Send Inquire / Grant (with MultiBlockGrant
enum: Ok/Busy/NoSpace/DuplicateMsg/BadMsg).
* S6F7/F8 — Data Transfer Request / Send. Host pulls a DATAID;
equipment replies with the nested DS/DV structure.
* S6F15/F16 — Event Report Request (host-initiated). Reply mirrors
the unsolicited S6F11.
* S6F19/F20 — Individual Report Request (RPTID -> values).
* S6F21/F22 — Annotated Individual Report Request (RPTID -> (VID, value)).
* S2F21/F22 — Legacy Remote Command (no parameter list). Delegates
to the same HostCommandRegistry as S2F41.
* S12F9/F10 — Map Data Send (array format, MAPFT=1).
* S12F11/F12 — Map Data Send (coordinate format, MAPFT=2).
* S12F13/F14, F15/F16, F17/F18 — Map Data Request variants for the
row, array, and coordinate formats.
11 new round-trip tests; suite at 289 cases / 1495 assertions.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a Docker-based interop harness that drives the C++ server with
secsgem-py 0.3.0 as the active host and probes a secsgem-py-passive
equipment from a minimal C++ active client. Surfaces and fixes four
interoperability bugs uncovered by cross-testing:
* SEMI E5 identifier formatcodes are a U1|U2|U4|U8 wildcard;
secsgem-py picks the narrowest fitting width while our parsers
only accepted U4. `as_uN_scalar` / `as_iN_scalar` now accept
any unsigned/signed width and range-check the downcast.
* PPBODY (S7F3/F6) is "ASCII | Binary | List" per the spec;
secsgem-py defaults to ASCII. Added BINARY_OR_ASCII codegen
item type with `as_text_or_binary` accessor.
* S1F23/F24 Collection Event Namelist was unimplemented; added
schema + `vids_for(ceid)` accessor on EventReportSubscriptions
plus the dispatch handler.
* S10F1 was registered as a host->equipment handler, but per
SEMI E5 §12 S10F1 is equipment->host; S10F3 is the actual
host->equipment Terminal Display Single. Added an S10F3
handler alongside (we keep S10F1 too for backward compat).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The biggest single gap I called out in the GEM300 audit — closed.
E84 is the digital handshake between AMHS (Automated Material
Handling System) and the equipment for carrier load/unload. Unlike
the rest of GEM300, this isn't SECS messaging; it's a fixed set of
ten parallel boolean wires that follow a strict sequencing protocol
(E84-0710 §6.3).
Adds:
E84Signal enum CS_0/CS_1/VALID/TR_REQ/BUSY/COMPT/L_REQ/U_REQ/
READY/ES
E84SignalSet 10-bit bitmap with bool get/set
E84State Idle / CarrierPresent / ValidAsserted /
LoadReady / UnloadReady / Transferring /
Complete / EmergencyStop
E84StateMachine re-evaluates state on every signal change,
observable via set_state_change_handler
Joins EquipmentDataModel as `e84` (top-level — there's one per tool,
not per port). ES (emergency stop) dominates regardless of other
signals; COMPT and BUSY override the VALID-handshake states. Same
FSM drives real opto-isolated I/O lines (when wired through an
asio digital input adapter) and the back-to-back test simulation.
Six test cases cover the full load handshake trace (six transitions,
including the transient LoadReady-after-BUSY-drops state), the
unload variant via U_REQ, ES dominance + recovery, reset(), and
no-op suppression for idempotent signal writes.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Extends the existing Clock with the metrics a host needs to gate
time-sensitive data against the equipment's sync state (E148 §6.3):
offset_seconds() current applied offset vs system clock
last_drift_seconds() signed drift observed at the most recent sync
sync_count() how many successful syncs have happened
sync_quality() Synchronized (|drift|<=1s) /
Drifting (<=60s) / Unsynchronized (>60s or
never synced)
The thresholds are tuneable per call; the defaults match typical fab
practice but the application can pass tighter bounds for tracelog-
sensitive flows. set_time_string() now snapshots the apparent delta
between the previously-applied offset and the new one as
last_drift_seconds_ at the moment of resync; no background timer.
Three new test cases cover the initial Unsynchronized state, a large
forward drift registering as Unsynchronized, and a same-value resync
landing as Synchronized.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Closes the slot-map verification gap I called out:
S3F19 / F20 host -> equip: verify expected slot map against what
the equipment has scanned. Equipment compares element-
wise; on match drives CSMS NotRead -> Read and replies
SVACK=Accept; on mismatch drives CSMS -> Mismatched and
replies SVACK=Mismatch.
S3F21 / F22 equip -> host: equipment-initiated slot map report
(typically pushed after CARRIERID is confirmed).
New SVACK enum: SlotMapVerifyAck { Accept, Mismatch, CarrierUnknown,
Error }. Server dispatch on S3F19 wires the actual CSMS transition
through the CarrierStore from D3.
Two round-trip tests cover both pairs; the FSM-driving behaviour is
exercised through the in-process tests because secs_server.cpp is
the dispatch entry point (no separate integration test needed beyond
the wire round-trip).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds the substrate-ID verification FSM that E90 §6.4.6 calls for:
NotConfirmed initial; equipment hasn't read the ID yet
WaitingForHost ID has been read; awaiting host accept/reject
Confirmed host confirmed (or force-bound)
Mismatched host rejected — recoverable via Bind
Events:
Read NotConfirmed -> WaitingForHost
Confirm WaitingForHost -> Confirmed
Mismatch WaitingForHost -> Mismatched
Bind any -> Confirmed (force-bind)
Reset any -> NotConfirmed
Wire-byte values pinned via static_assert. The third axis is now
exposed on SubstrateStateMachine alongside location_state() and
processing_state(); set_id_handler() observes transitions. Existing
two-axis API is unchanged.
4 new test cases cover the happy path, Mismatch+Bind recovery, Reset
from any state, and same-state event handler suppression.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Each Substrate now retains an append-only history of state transitions
(both location and processing axes), the triggering event captured as
a std::variant<SubstrateEvent, SubstrateProcessingEvent>, the location
label at the time, and a steady_clock timestamp.
E90 §6.6 requires the equipment to be able to report a wafer's
processing history — typically queried via S6F11 batched reports or
SVID reads. This commit lays the runtime substrate; wire query
plumbing is the natural follow-up.
set_history_limit(n) caps per-substrate retention (default 256, 0 =
unbounded). Oldest entries are dropped when the cap is reached;
vector-erase is fine at this scale (typical wafer lifecycle is a few
dozen transitions).
Two new test cases cover the recording invariants (every fire results
in one history entry on the right axis) and history_limit eviction.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
EptStateMachine now retains per-state cumulative dwell time so the
host can read it as SVIDs. The implementation is zero-overhead while
the FSM is idle (no timers, no background work) — on every transition
we add the prior state's dwell to its bucket and reset the entered_
timestamp. Live dwell in the current state is included in
accumulated() via a now-vs-entered_ delta at read time.
New public API:
accumulated(EptState) per-state cumulative ms (incl. live dwell)
total_elapsed() denominator for utilization ratios
reset_history() S2F43-style history clear
This closes the gap I called out: previously we emitted CEIDs on
transition but didn't accumulate the bucket the host actually queries
for utilization metrics. Wiring these into specific SVIDs is the
application's job (equipment.yaml declares SVIDs against any read
callable); the runtime data is now there.
4 new test cases cover accumulation, live-dwell inclusion, and reset.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The S9F3/F5 fallback was previously inlined in apps/secs_server.cpp;
this commit lifts it onto Router as a template helper and adds two
focused tests asserting the wire behaviour against a real back-to-
back HSMS Connection pair.
template <typename EmitFn, typename HeaderProvider>
std::optional<Message> dispatch_with_s9(emit, header, msg);
The helper does the has_handler / has_handler_for_stream check and
calls the supplied emit function with S9F3 (unknown stream) or S9F5
(unknown function in known stream). The header_provider returns the
optional MHEAD bytes — keeping the helper free of any direct
Connection coupling.
Tests:
- SUT registered only for S1F1; peer sends S1F5 -> SUT replies
S9F5 to the peer.
- SUT registered only for S1F1; peer sends S7F19 -> SUT replies
S9F3 to the peer.
Closes Tranche I — SML parser and the auto-S9F* fallback closeout
both verified end-to-end.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds parse_sml(text) -> Item / try_parse_sml(text) -> optional<Item>
in secs2/sml.hpp. Round-trips with the existing to_sml() emitter for
every Item shape the codec produces: lists with nesting, ASCII / JIS8,
Binary (decimal and 0xHH literals), Boolean (T/F or 1/0, scalar and
multi-element), U1-U8 / I1-I8 / F4 / F8 vectors, and the optional
`[n]` count syntax (accepted but not enforced).
The parser is whitespace-insensitive outside quoted strings and uses
a small Cursor type for read_word / read_quoted / skip_ws. Numeric
literals go through strtoul/strtoll/strtod so SML can carry hex,
octal, and decimal interchangeably (the emitter writes hex for Binary
and decimal everywhere else).
11 test cases cover the full round-trip surface, the whitespace
invariant, unknown-tag rejection, the try_parse error-swallowing
variant, and the optional `[n]` count.
secsgem-py has secs/sml.py for the same purpose; this brings the C++
port to parity on the tooling side.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds the canonical E157 collection-event identifiers in the 1570+ block:
1570 ModuleProcessStateChange (generic; fired on any transition)
1571 ModuleNotExecuting
1572 ModuleGeneralExecuting
1573 ModuleStepExecuting
1574 ModuleStepCompleted
Server installs a state-change handler that fires both the generic
CEID and the state-specific one for each transition. Hosts that
prefer "wake me on any module change and I'll fan it out myself" can
subscribe to only the generic CEID; hosts that want narrower
notifications subscribe to specific states.
Closes Tranche H — E157 Module Process Tracking end-to-end (FSM +
Store + CEIDs + server emission).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per-module process-tracking state machine. An E157 instance models a
single recipe step at a single module, with the canonical lifecycle:
NotExecuting -> GeneralExecuting (StartGeneral)
-> StepExecuting (StartStep)
-> StepCompleted (CompleteStep)
Plus universal escape hatches: Reset returns any state to
NotExecuting; Abort terminates from any state to StepCompleted.
ModuleStore wraps the FSM with the now-standard pattern:
- non-movable (this-capture lambdas)
- per-module bind() carries current_substid + recipe_step
- fire(module_id, event) delegates to the FSM
- set_state_change_handler observes every transition with module_id
Joins EquipmentDataModel. 5 test cases cover happy path, Reset from
each interior state, Abort, store-level create dedup + bind, and the
multi-module change handler keying.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Hierarchical object tree for equipment self-description. Each object
carries a CemObjectType (Equipment / Subsystem / IODevice / Module /
MaterialLocation / Other), an optional parent_objid, and a flat
attribute map keyed by name (the wire shape S14F1 / F3 returns).
Operations covered:
add(CemObject) - dedup'd, validates parent exists
get / has - lookup by objid
get_attr / set_attr - E14 GetAttr / SetAttr semantics
children(parent) - tree traversal; empty parent = roots
The flat-map representation matches how E14 ObjectService traffic
addresses nodes (by OBJSPEC string). Wiring S14F1/F2 GetAttr and
S14F3/F4 SetAttr to this store is a downstream commit; the data model
is what was missing.
Joins EquipmentDataModel alongside the other top-level stores. Three
test cases cover hierarchical add+dedup, children() traversal, and
get/set/missing attribute semantics.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
EquipmentDataModel now carries an EptStateMachine as a value member
alongside the other top-level state machines. Server installs a
state-change handler that maps every EPT transition to a CEID emission
through the existing emit_event path:
1100 NonScheduledTime 1103 Engineering
1101 ScheduledDowntime 1104 Standby
1102 UnscheduledDowntime 1105 Productive
CEIDs land in the 1100+ block to keep clear of the demo equipment.yaml
(100s/200s/400s) and E90 (900s). Log lines include the dwell time of
the previous state so trace-level diagnostics show utilization without
extra tooling.
Application code drives transitions by calling model->ept.on_event(...);
the existing event-report machinery (subscription state, S6F11
batching, spool) gates wire emission so EPT events spool on offline
hosts like every other CEID.
Closes Tranche F — E116 Equipment Performance Tracking end-to-end.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds the six E116-0712 §6.2 buckets for classifying equipment time:
NonScheduledTime (0) not scheduled to operate
ScheduledDowntime (1) planned maintenance window
UnscheduledDowntime (2) faults / unplanned stoppage
Engineering (3) engineering / qualification time
Standby (4) idle but available
Productive (5) actively producing
Wire-byte values pinned via static_assert to E116 §10.3.
The FSM is a classifier rather than a strict lifecycle — every
(state, event) pair is legal — but it remains data-driven through the
shared CarrierTransitionTable template so the default cross-product is
expressible declaratively.
The state-change handler also surfaces dwell time (how long the
previous state was held) computed off std::chrono::steady_clock, so
accounting code can compute MTBF / availability / utilization from a
single source without maintaining a parallel timestamp log.
4 test cases cover the initial state, every event firing, dwell-time
reporting, and the no-op same-state event (no handler call).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a typed substrate-event callback to HostHandler that decodes the
canonical E90 CEIDs from incoming S6F11 messages into the matching
SubstrateState / SubstrateProcessingState enum values. Host
applications now get strongly-typed substrate observability without
having to maintain their own CEID-to-state lookup.
using SubstrateEventHandler =
std::function<void(uint32_t ceid, SubstrateState location,
SubstrateProcessingState processing)>;
void set_substrate_event_handler(SubstrateEventHandler);
Axes not addressed by a given CEID stay at NoState — the handler
distinguishes "this CEID updates the location axis" from "this CEID
updates the processing axis" so the host can keep its own per-
substrate FSM in sync.
Closes Tranche E — E90 Substrate Tracking end-to-end (FSM + Store +
CEIDs + server emission + host observer).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds the canonical E90-0716 §6.5 collection-event identifiers as a
single header of inline constants (gem/e90_constants.hpp) keyed off
SubstrateState / SubstrateProcessingState transitions:
901 AtSource 910 NeedsProcessing
902 AtWork 911 InProcess
903 AtDestination 912 Processed
913 Aborted
914 Stopped
915 Rejected
916 Lost
917 Skipped
Values use the 901+ block to avoid collision with the demo CEIDs in
data/equipment.yaml (100s/200s/400s).
Server installs location + processing change handlers on
model->substrates that map every transition to emit_event() with the
matching CEID. The existing event-report machinery (subscription
state, S6F11 batching, spool) gates the actual wire emission, so this
plays nicely with hosts that subscribe to only a subset of substrate
events.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds the Carrier Management wire surface that drives the FSMs from D1:
S3F17 / S3F18 CarrierAction (host issues ProceedWithCarrier /
CancelCarrier / BindCarrierID / etc.; CommandParameter
list is reused from S2F41).
S3F23 / S3F24 PortGroupChangeReport (equipment notifies host of port
group composition changes).
S3F25 / S3F26 CarrierTransfer (host instructs source -> target port
transfer).
S3F27 / S3F28 CancelCarrier (host cancels an outstanding carrier op).
Two new ack enums in messages_helpers.hpp:
CarrierActionAck — CAACK byte; covers the common error responses
(CarrierIDUnknown, Inaccessible, ActionInProgress).
PortGroupAck — PGACK byte (Accept / Error).
Round-trip tests for all four pairs. Server-side dispatch + the
CarrierStore + LoadPortStore that the FSMs key into land in D3/D4.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per-carrier triple FSM: CIDS (id verification), CSMS (slot-map), CAS
(access). Per-port triple FSM: LPTS (transfer), LRS (reservation), LAS
(association). Wire-byte enum values pinned via static_assert to match
E87-0716 §10.3.
CarrierStateMachine combines the three carrier-side FSMs because they
are independent but always observed together; same for LoadPortState-
Machine. Generic CarrierTransitionTable<State, Event> template is
reused across all six tables — same row shape as the PJ/CJ/Exception
tables that already exist.
Default tables cover the spec's documented transitions:
CIDS: NotConfirmed <-> Confirmed/Mismatched/Unknown, Cancel returns
to NotConfirmed from any state, Bind force-confirms.
CSMS: NotRead -> Read -> {Mismatched, Reset}.
CAS: NotAccessed -> InAccess -> Complete (terminal).
LPTS: OutOfService <-> InService <-> Loading/Unloading.
LRS / LAS: simple boolean toggle pairs.
15 test cases assert the happy-path lifecycles, cross-state cancels,
and that change handlers fire only on real transitions (Read in
NotConfirmed is a no-op, not a handler call).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
ALCD's lower 7 bits are a bitmap of category flags per E5 §10.3 and
E30 §6.13; a single alarm may carry multiple categories at once
(e.g. an irrecoverable equipment-safety fault is 0x10 | 0x02).
Adds:
enum class AlarmSeverity : uint8_t
PersonalSafety EquipmentSafety ParameterError ParameterWarning
Irrecoverable EquipmentStatus Attention
has_severity(alcd, bit), severity_bits(alcd)
Alarm::has(bit), Alarm::is_safety()
constexpr severity_mask = 0x7F
Tests cover single-category alarms, multi-category combos, and that
the bit-7 SET/CLEAR flag is correctly excluded from category bits.
Closes Tranche C (E5 alarm/exception state model).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Per-EXID exception lifecycle for E5 §9. States mirror the wire flow:
Posted equipment sent S5F9, awaiting host or autonomous clear
Recovering host's S5F13 accepted; equipment running recovery
RecoverFailed S5F15 reported a failed result; host may retry
Cleared terminal — store removes the row
Events:
Created synthetic NoState->Posted observer signal
Recover host's S5F13 (Posted/RecoverFailed -> Recovering)
RecoveryComplete equipment internal (Recovering -> Cleared)
RecoveryFailed equipment internal (Recovering -> RecoverFailed)
RecoveryAbort host's S5F17 (Recovering -> Posted)
Clear equipment internal (Posted/RecoverFailed -> Cleared)
ExceptionStore mirrors ProcessJobStore: per-EXID FSMs heap-allocated via
unique_ptr, non-movable to keep `this`-captures safe, synthetic Created
fires after the row lands so observers can decide whether to emit S5F9
out of band. on_recover validates EXRECVRA against the candidates the
post advertised.
The store joins EquipmentDataModel alongside process_jobs / control_jobs.
S5F9-F18 server-side dispatch lands in C2.
Tests (12 cases) cover FSM transitions including retry, abort, and
autonomous clear, plus store-level duplicate-rejection, EXRECVRA
validation, and Cleared-removes-the-row semantics.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Fleshes out the host-side message surface so the demo client app no
longer has to inline message construction. Senders added (each is a
one-line wrapper over the codegen builders + Connection::send_request):
Remote command: S2F41/F42, S2F49/F50
Alarm management: S5F3/F4 enable, S5F5/F6 list, S5F7/F8 list-enabled,
S5F13/F14 recover, S5F17/F18 recover-abort
Process programs: S7F3/F4 send, S7F5/F6 request, S7F19/F20 EPPD
Spool: S6F23/F24
Terminal: S10F1/F2 single, S10F5/F6 multi
E40 Process Jobs: S16F11/F12 create, S16F5/F6 command, S16F13/F14 dequeue
E94 Control Jobs: S14F9/F10 create, S14F11/F12 delete, S16F27/F28 command
CommandParameter is reused from store/host_commands.hpp rather than
inventing a parallel ParamPair — host and equipment talk in the same
struct now.
Closes the outbound side of the host-mode menu. The remaining piece
is an integration test that drives this against the equipment server
end-to-end (B4).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds the five GEM event-subscription primitives the host needs to drive
the equipment's data-collection lifecycle (E30 §6.11):
S1F3/F4 selected status data
S1F11/F12 status variable namelist
S2F33/F34 define reports
S2F35/F36 link event reports
S2F37/F38 enable/disable events
Each is a one-line wrapper over the codegen builders + Connection's
send_request, surfacing the codegen-generated DefineReportEntry /
LinkEventEntry structs to callers behind a {id, [vids]} pair API.
This is the minimum surface a host needs to walk a fresh equipment
through "define report -> link CEID -> enable" and start receiving
S6F11 event reports — the same pattern the existing demo client does
inline. B3 lands the RCMD / recipe / job / terminal senders that
build on top.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
GEM host-side counterpart to the existing equipment server: wraps an
HSMS Connection (Active mode), installs an inbound dispatch table that
auto-acks the messages a host is expected to passively accept, and
exposes the GEM workflow primitives.
Inbound dispatch:
S5F1 Alarm Report observe (alarm handler) + S5F2 Accept
S6F11 Event Report observe (event handler) + S6F12 Accept
S6F25 Spool Data Ready S6F26 Accept (host policy: pull on demand)
S10F1 Terminal Display observe + S10F2 Accepted
S9F* Equipment errors observe (s9 handler); no ack (one-way)
Workflow shortcuts:
establish_communication() S1F13 -> S1F14
go_remote() S1F17 -> S1F18
go_offline() S1F15 -> S1F16
Plus a low-level send_request() escape hatch so the senders coming in
B2/B3 don't have to friend the connection internals.
Drive-by: event_reports.hpp was missing `<optional>` (worked transitively
through the equipment-side include chain but not when included from the
host-side standalone).
secsgem-py has `gem/hosthandler.py`; this mirrors its surface for the
inbound-ack and lifecycle parts. Outbound senders land in B2/B3.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Wires the SECS-I Protocol FSM behind an asio TCP socket so the block
protocol can run over loopback without serial hardware. Mirrors
secsgem-py's `secsitcp/` adapter — useful for back-to-back simulators
and CI without a serial device.
Adds:
include/secsgem/secsi/tcp_transport.hpp
src/secsi/tcp_transport.cpp
tests/test_secsi_tcp.cpp
The transport:
- Splits outgoing SECS-II messages into blocks (transparent multi-block).
- Accumulates incoming blocks until end_block=true, then assembles and
delivers as a single SECS-II message — same surface as the HSMS
Connection's MessageHandler.
- Drives T1 / T2 timers from asio steady_timer; T3/T4 stay upper-layer
per the FSM contract.
- Auto-allocates monotonic system bytes per send.
Tests cover single-block delivery, multi-block reassembly (700-byte
ASCII body spanning multiple SECS-I blocks), and bidirectional exchange.
This closes Tranche A (catch-up to secsgem-py wire/transport surface).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds a complete IO-free SECS-I implementation:
include/secsgem/secsi/header.hpp 10-byte block header (R/W/E bits)
include/secsgem/secsi/block.hpp length + header + body + checksum
include/secsgem/secsi/protocol.hpp half-duplex FSM (ENQ/EOT/ACK/NAK)
src/secsi/* implementations
tests/test_secsi.cpp header, block, multi-block split,
back-to-back FSM drive, RTY,
contention, T2 timeout
The protocol is event-driven (`Event` → `Action` queue), so wiring it
to an asio serial_port is a thin adapter — that lands in the next
commit so this one stays reviewable.
Key design points:
- Master/slave contention: slave yields on simultaneous ENQ (E4 §7.1.4).
- RTY exhaustion raises ActionRaiseError, clears the send queue, resets
to Idle (no zombie state).
- Multi-block assembler validates contiguous 1..N numbering and exclusive
E-bit-on-last invariants — rejects malformed sequences with nullopt.
- Block::checksum is exposed publicly for the receive path's verification.
Tests cover the happy path (back-to-back delivery), error paths
(checksum mismatch, short input, oversize body), retries (NAK chain to
exhaustion), and protocol corner cases (contention, T2 timeout).
secsgem-py implements SECS-I block framing but lacks the explicit RTY
state machine; this commit puts the C++ port ahead on transport
correctness.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds the core map-management messages: setup send/request (F1-F4),
transmit inquire/grant (F5/F6), data send in row format (F7/F8), and
the one-way error report (F19). Reference points (REFP) are an
external struct shared across F1 and F4.
The alternative data encodings — array format (F9/F10), coordinate
format (F11/F12), and the corresponding request pairs (F13-F18) —
are scope-deferred. They're mechanical YAML edits once a tool needs
them; the codec already handles the underlying BINARY/list shapes.
Three new ack enums: MapSetupAck (SDACK), MapTransmitGrant (GRANT),
MapDataAck (MAPER). No state machine yet — maps are a data exchange,
not a lifecycle.
secsgem-py ships S12F0-F19 as a single block; this commit covers the
practically-used subset and matches their wire shapes where they
overlap.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
GEM300 layer: SEMI E40-0705 Process Job and E94-0705 Control Job
state machines, plus the E30 §6.1 communication-state machine that
sits between HSMS SELECT and full GEM communication. Data-driven
via data/process_job_state.yaml and data/control_job_state.yaml,
mirroring the existing control_state.yaml pattern.
Wire coverage:
S14F9/F10 CreateObject (CJ) host -> equipment
S14F11/F12 DeleteObject (CJ) host -> equipment
S16F5/F6 PRJobCommand host -> equipment
S16F9 PRJobAlert equipment -> host
S16F11/F12 PRJobCreate (simplified body) host -> equipment
S16F13/F14 PRJobDequeue host -> equipment
S16F27/F28 CJobCommand host -> equipment
Process Job FSM exposes 8 states matching PRJOBSTATE bytes (E40 §10.3.2);
HOQ is reorder-aware (move-to-head against an insertion-order vector);
Stop/Abort on a Queued PJ routes through ABORTING so the host observes
PRJOBSTATE=7 on the wire (§6.3); alert_enabled is settable per-PJ for
PRALERT control; FSM dispatches through ProcessJobStore::on_change_
dynamically so a late set_state_change_handler() reaches existing PJs.
Hardening: loader rejects NoState (sentinel) as initial/from/to and
rejects `on: created` rows; static_asserts pin enum values to wire
bytes; ProcessJobStore is non-movable to keep the per-PJ this-capture
safe.
Server simulator cascades the full CJ -> PJ lifecycle on CJSTART so
the wire trace exercises every legal state. CEIDs 400/401 fire on CJ
state changes via the existing event-report pipeline.
Tests: 60+ new assertions across test_process_jobs, test_control_jobs,
test_communication_state, test_hsms_connection, plus loader and
messages round-trip coverage.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The final additions: S10F5/F6 multi-line terminal display (closes the
last partial Additional capability — Equipment Terminal Services flips
✅), and a thorough COMPLIANCE.md / README pass that states the 100%
claim honestly.
Catalog + handlers
data/messages.yaml S10F5 / S10F6 added.
apps/secs_server.cpp router.on(10, 5) iterates the line list,
acks with S10F6.
tests/test_messages.cpp Round-trips a 3-line multi-line display.
COMPLIANCE.md (rewritten)
Every GEM Fundamental ✅. Every GEM Additional that E30 binds to a
concrete message set ✅. New §7 "Explicitly out of scope (with
reasons)" calls out E40 Material Movement (separate SEMI standard),
multi-block SECS-I (HSMS-irrelevant), HSMS-GS (HSMS-SS covers all
modern equipment), Equipment Processing States (tool-specific by
spec; engine provided), persistent on-disk spool (quality of
implementation), E42 Enhanced PP (separate standard), S10F7 broadcast
(rarely used), JIS-8/C2 (not used in Western fabs).
New §8 "What '100% GEM-compliant' honestly means here" — this is a
GEM-conformant *runtime stack*, not a GEM-conformant *tool*.
Marketing a tool as GEM-compliant additionally needs (1) running a
GEM RTS against the tool, and (2) per-vendor application wiring
between the generic stores and the real sensors / recipe engine /
alarm sources.
README.md (rewritten)
Architecture diagram updated to reflect the actual store list (nine
stores). "Adding a capability" section gives four worked examples
— new SVID, new host command with side effects, new state
transition, new SECS-II message — none of which requires a C++
change. Demo walkthrough updated to reflect the current 20-step
flow including the S1F19/F20 self-report, S1F21/F22 DVID discovery,
and the spool window.
Code clarity
include/secsgem/gem/data_model.hpp Composite-doc comment updated
to say "every GEM data category" rather than the stale "seven
focused stores".
Verified
- Tests: 84 cases / 487 assertions pass.
- Demo: 198 server/host log lines; exits 0 end-to-end.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
New TraceStore keyed by TRID; each entry is a TraceConfig with
DSPER + TOTSMP + REPGSZ + SVID list. S2F23 validates that every SVID
exists (TIAACK=4 otherwise) and registers the trace.
S6F1's body is L,4 of {TRID U4, SMPLN U4, STIME ASCII, list_of <Item>}
— the application chooses whether each value Item is a scalar SVID
value or a packed batch.
The periodic sampling timer that turns an active TraceConfig into
S6F1 emissions is intentionally left to the application (E5 doesn't
mandate a specific scheduler and vendors typically already have one).
Four new SxFy in the catalog.
COMPLIANCE.md: Trace Data Collection Additional capability flips ✅.
Tests: 82 cases / 477 assertions.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
New LimitMonitorStore keyed by VID; each entry is a vector of
LimitDefinition (LIMITID + upper/lower deadband as arbitrary Items).
S2F45/F46 set, S2F47/F48 read. VLAACK validates each VID exists.
Four new SxFy in the catalog; codegen handles the nested
list-of-(VID, list-of-LimitDefinition) shape. LimitDefinition is
defined in store/limits.hpp and referenced as external_struct so the
data model and the message codecs share one type.
The actual "value crossed limit" detection + CEID emission is left to
the application's set_value path (E30 §6.21 leaves *how* the equipment
detects crossings up to the implementer).
COMPLIANCE.md: Limits Monitoring Additional capability flips ✅.
Tests: 80 cases / 465 assertions.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Closes the S9 stream. Every documented protocol-error condition is now
auto-emitted by Connection (with the assist of one Router predicate),
without involving the application.
Router (include/secsgem/gem/router.hpp)
Adds two predicates: has_handler(stream, function) and
has_handler_for_stream(stream). Lets the wrapping message handler
decide whether an unhandled message is "unknown stream" (S9F3) or
"unknown function in a known stream" (S9F5).
Connection (include/secsgem/hsms/{connection.hpp, connection.cpp})
- emit_s9() goes public so the message_handler can call it.
- New current_header() accessor returns the HSMS header of the
primary currently being dispatched. Non-null only inside the
on_message_ call; cleared on the way out.
- handle_data sets current_header_ before invoking on_message_.
- on_length on oversized frame: synthesizes a 10-byte MHEAD whose
first 4 bytes are the offending length prefix, emits S9F11, and
sets close_after_flush so the S9F11 goes out before the socket
closes.
Server (apps/secs_server.cpp)
The conn->set_message_handler lambda now wraps router.dispatch. For
any inbound primary without a registered handler, it captures the
MHEAD via current_header() and emits either S9F3 (stream unknown) or
S9F5 (function unknown). The wrapper still returns the Router's
reply (SxF0 for primaries with W) so transactional semantics are
preserved.
COMPLIANCE.md
Error Messages row flips from 🟡 to ✅. S9F3/F5/F11 rows in the
coverage matrix flip from 🟡 to ✅. Each row in the matrix now
states its trigger condition explicitly. Drops the
"Finish S9 wiring" bullet from the "what would 100% take" list.
Verified
- Tests: 78 cases / 454 assertions still pass (no behavioural change
on the happy path; new emission paths fire only on protocol errors
that the demo doesn't induce).
- Build clean.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds the GEM "Documentation" Fundamental capability: the equipment now
self-reports which GEM capabilities it supports, and the host can
discover the DVID namelist with the same shape used for SVIDs.
Catalog (data/messages.yaml -> generated messages.hpp)
S1F19 W header-only Get GEM Compliance Request
S1F20 <L,3 <A SOFTREV>
<A EQPTYP>
<L,a <L,2 <U1 CCODE> <A CDESC>>>>
Get GEM Compliance Data
S1F21 W <L,n <U4 VID>> DVID Namelist Request (n=0 = all)
S1F22 <L,n <L,3 <U4 VID>
<A VNAME>
<A UNITS>>> DVID Namelist Data
Codegen emits CapabilityEntry and GemCompliance structs. S1F22 reuses
S1F12's StatusName struct (same wire shape; dedup avoids redefinition).
Equipment data dictionary (data/equipment.yaml)
device: Adds `equipment_type: "EQUIPMENT"`
for the S1F20 EQPTYP field.
capabilities: New section. List of
- {code, name} (CCODE, CDESC) pairs honestly
reflecting what the codebase
implements: 1, 2, 3, 5, 6, 7, 8,
9, 11, 12, 14 (partial), 15.
dvids: New section, same schema as
svids:. Demo populates two:
- WaferCounter (U4, units wafer)
- ChamberPressure (F4, units Torr)
Loader (src/config/loader.cpp + include/secsgem/config/loader.hpp)
EquipmentDescriptor gains equipment_type and capabilities (vector of
(uint8_t, string) pairs). load_equipment now reads `capabilities:`
into the descriptor and `dvids:` into model.dvids.
Server (apps/secs_server.cpp)
router.on(1, 19) returns S1F20 with desc.software_rev,
desc.equipment_type, and desc.capabilities converted to
vector<CapabilityEntry>.
router.on(1, 21) returns S1F22 built from model.dvids.all().
Client (apps/secs_client.cpp)
Two new demo steps after Request Online and before SVID discovery:
S1F19 -> S1F20: logs SOFTREV, EQPTYP, and every (CCODE, CDESC)
the equipment claims.
S1F21 -> S1F22: logs each DVID with units.
Tests
tests/test_messages.cpp Round-trip S1F19/F20 with a 3-entry
capability list; round-trip S1F22 with two
DVIDs.
tests/test_loader.cpp Asserts equipment_type, the capabilities
list contains CCODE 14 (Spooling), and the
two DVIDs land in model.dvids.
COMPLIANCE.md
"Documentation" Fundamental moves from ⬜ to ✅.
S1F19/F20 + S1F21/F22 rows in the coverage matrix flip to ✅.
The "what would it take" list drops the documentation-messages bullet.
Verified
- Tests: 77 cases / 444 assertions pass.
- Demo: client logs the full capability list received from the
equipment, including CCODE 14 "Spooling (partial; S2F43/F44 +
S6F23/F24)" — the equipment honestly reports its partial
implementation rather than overclaiming.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds S9F1, F3, F5, F7, F9, F11, F13 to the message catalog and wires
the two emission paths that the Connection layer can drive without help
from the Router or the application: S9F7 on a body-decode failure and
S9F9 on a T3 transaction-timer timeout.
Catalog (data/messages.yaml -> generated messages.hpp)
All six MHEAD-carrying messages (F1/F3/F5/F7/F9/F11) use the same
shape — a single <B 10> body with the offending 10-byte HSMS header.
S9F13 (conversation timeout) carries <L,2 <A MEXP> <A EDID>>.
Connection-side emissions (src/hsms/connection.cpp)
emit_s9(function, mhead) New private helper. Builds a 9/function/W=0
data message whose body is <B 10> with the
MHEAD bytes, allocates a fresh sys_bytes,
and queues it onto the write path. No
reply is tracked.
S9F7 on body decode handle_data wraps Message::from_body in a
try/catch. Previously any decode error
closed the connection; now it emits S9F7
with the offending header and continues
reading. Reply-side decode failure also
emits S9F7 and surfaces the new
Error::IllegalData to the waiting
ReplyHandler (rather than making the
caller wait out T3).
S9F9 on T3 timeout The send_request T3 callback rebuilds the
original outgoing MHEAD from
(device_id, expected_stream,
expected_function-1, sys, W=1) and emits
S9F9 before invoking the callback with
Error::Timeout (unchanged).
What's intentionally not yet wired (logged in COMPLIANCE.md)
- S9F3 / S9F5 — "unknown stream / function". These need to live in
the Router's fallback path, which would require either the Router
knowing about a Connection-shaped sender or the Connection's
message wrapper learning which streams the Router has handlers
for. Deferred — today the fallback returns SxF0 only.
- S9F11 — "Data Too Long". Currently we close on oversized frames;
we'd need to also build a synthetic 10-byte MHEAD substitute (the
real header isn't yet available at the point of detection) and
flush it through close_after_flush.
Tests + docs
tests/test_messages.cpp Round-trip every S9F* using a representative
10-byte MHEAD literal; check S9F13 carries
MEXP + EDID. +2 cases / +37 assertions.
COMPLIANCE.md Error Messages row moved from "no S9 stream"
to a detailed status describing what's
emitted vs catalog-only. Coverage matrix
expanded per-message (F1/F7/F9/F13 ✅;
F3/F5/F11 🟡 catalog-only).
Build/demo unaffected: 75 cases / 420 assertions pass; the happy-path
demo never trips a decode error or T3, so the S9 path isn't exercised
end-to-end (but unit tests prove the wire shape).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Implements the largest functional gap from the compliance audit. The
equipment now queues events the host can't immediately receive (either
because there's no SELECTED session or because the demo's force-spool
flag is on) and transmits the queue on host request.
What's new
include/secsgem/gem/store/spool.hpp
SpoolStore: a deque queue with a configurable per-stream whitelist
(so only streams 5+6 spool by default), a max_size cap with FIFO
eviction on overflow, and a `force_spool` test flag. Enqueue
returns one of Queued / Dropped_NotSpoolable / Dropped_Full so the
caller can fall back to live delivery when appropriate. Drain
pops the entire queue in FIFO order. Two new ack enums:
ResetSpoolAck (S2F44 RSPACK) and SpoolRequestAck (S6F24 RSDA), plus
SpoolRequestCode (S6F23 RSDC, Transmit/Purge).
data/messages.yaml + auto-regenerated messages.hpp
S2F43 W <L,n <B stream>> Reset Spooling
S2F44 <L,2 <B RSPACK> <L,a ...>> Reset Spooling Ack
S6F23 W <B RSDC> Request Spooled Data
S6F24 <B RSDA> Request Spooled Data Ack
data/equipment.yaml
`spool:` section: max_size + spoolable_streams list. Two new host
commands SPOOL_ON / SPOOL_OFF that flip the force-spool flag (these
stand in for "host link down" in the demo without dropping TCP).
include/secsgem/gem/store/host_commands.hpp
Spec/Result gain an optional<bool> force_spool field. S2F41
dispatch returns the result, the server applies it after S2F42 is
queued.
src/config/loader.cpp
Reads `spool:` from equipment.yaml; reads `force_spool` from each
host_commands entry; populates SpoolStore + CommandSpec.
apps/secs_server.cpp
New `deliver_or_spool(msg, what)` helper. emit_event and
emit_alarm_set funnel through it: if force_spool is on (or there's
no active session), msg.stream is checked against the spoolable
list and the message is enqueued; otherwise it's sent live.
Two new handlers:
S2F43 parses the stream list, updates SpoolStore, replies S2F44
S6F23 RSDC=Transmit drains and re-sends each as a fresh primary
(posted on the executor so the S6F24 ack flushes first);
RSDC=Purge clears the queue and acks.
The S2F41 handler now also propagates result.force_spool into the
SpoolStore.
apps/secs_client.cpp
Demo extended with 4 new steps after the FAULT branch:
SPOOL_ON -> S2F42 Accept
START -> S2F42 Accept; CEID 300 emission spooled (no live S6F11)
SPOOL_OFF -> S2F42 Accept; queue still has the message
S6F23(Transmit) -> S6F24 Accept; spooled S6F11 arrives next
Then the existing S7F19/S7F5/S10F1/S1F15/Separate flow continues.
tests/test_data_model.cpp
Four new TEST_CASEs for SpoolStore (whitelist, FIFO eviction at
max_size, drain ordering, force flag).
tests/test_loader.cpp
Confirms equipment.yaml's `spool:` section populates the store and
`force_spool: true/false` flows through to dispatch results.
COMPLIANCE.md
Spooling moves from ⬜ to 🟡. Adds S2F43/F44 + S6F23/F24 as ✅ in
the message coverage matrix; calls out what's still missing
(S6F25/F26 notification, automatic activation on HSMS NOT-SELECTED,
persistent on-disk spool).
Verified
- Tests: 73 cases / 383 assertions pass (+4 spool cases).
- Demo (docker compose up server client) walks the full happy path
and the spool path, observed in the server log as:
spool: force_spool=true (depth=0)
spool: S6F11 CEID=300 queued (depth=1)
spool: force_spool=false (depth=1)
S6F23 transmit: draining 1 messages
and on the host side as the queued S6F11 arriving in the correct
order after S6F24.
Known limitations (logged in COMPLIANCE.md)
- Spool activation is manual via SPOOL_ON/OFF rather than
automatically triggered by HSMS NOT-SELECTED.
- No S6F25/F26 spooled-data-ready notification on re-SELECT.
- In-memory only; an equipment restart loses queued events.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
raphael