The handle-recycle loop disposed the device client with no gate. It could null a client a
concurrent EnsureConnectedAsync had just installed (orphaning a live two-socket session),
or dispose one a poll was mid-read on (ObjectDisposedException -> BadCommunicationError +
Degraded on every recycle interval).
The recycle now takes the device's ConnectGate, detaches the client, and disposes it via
SynchronizedFocasClient.DisposeWhenIdleAsync, which waits (bounded by the call timeout plus
slack) for ops that already hold the client. The gate stays held through the drain, so no
fresh session opens beside the old one.
SynchronizedFocasClient no longer disposes its call gate. Releasing a disposed semaphore
threw out of reads that had otherwise completed. Ops that reach a disposed client now get
ObjectDisposedException and are never run on the dead socket.
Refs #740
Claude-Session: https://claude.ai/code/session_01Ub9hGmYLQvn9XETzw7CMsT
Servo (0x0056) and spindle (0x0040) meter records are 8 bytes:
[i32 value][u16 base][u16 exponent], the same self-describing envelope as the
position blocks.
Servo: offset 4 was read as a decimal count. It is the BASE (a steady 10), so
every load was divided by 10^10 and published ~0 -- the sister ignitionoee panel
read servo 0% on every axis of a Makino cutting at 20,000 rpm. Loads are now
value / base^exponent, a direct percentage.
Spindle: the raw meter is a 1/16384 fraction of full scale. Settled on the real
Z-34184 by regressing the type-1 meter against cnc_rddynamic2 spindle speed over
596..20,000 rpm (rpm = 1.220749 x meter; 16384 x 1.220749 = 20000.75). So:
- Spindle/{name}/Load publishes % of rated load (Float64; was the raw Int32
count -- a reading of 9155 meant 55.9%).
- Spindle/{name}/MaxRpm is RENAMED Spindle/{name}/MotorSpeedPct (Float64, % of
maximum speed) and is polled with Load. The old node was actively wrong: the
type-1 meter is the live motor speed, and it was read once at bootstrap and
served forever as if it were a max-RPM constant. Client-visible: FIXED: tags
on .../MaxRpm must be re-pointed.
- The spindle read no longer stops at the first zero after slot 1 (which dropped
an idle second spindle and kept non-zero residue); every slot is returned.
A record whose envelope is undescribable (base not 2/10, exponent > 18 -- e.g.
residue past the configured axis/spindle count) publishes BadNoData, never a Good 0.
Ported from the sister ignitionoee FOCAS driver v0.17.2/v0.17.3 (Focas.Wire-019);
live captures live-spmeter-{load,maxrpm}-0x0040.bin added as fixtures.
Every 0x0026 position block is self-describing: [i32 raw][u16 base][u16 exponent],
value = raw / base^exponent. The live 31i-B capture (ref-dynamic2-axis1.bin)
reports base 10 / exponent 4 on every block, so axis 1 is 280.1574 mm, not
2801574. The driver ignored the envelope and instead sent 'cnc_getfigure'
(0x00D3) -- a command that does not exist in the FOCAS wire protocol, served only
by the in-tree mock; the real control refuses it with rc=1, so on hardware every
position fell back to PositionDecimalPlaces (default 0) and published raw counts.
- ParsePositionBlock decodes the envelope into FocasPosition; the driver applies
raw / base^exponent at publish. PositionDecimalPlaces is now only the fallback
for a scale-less (4-byte) block.
- A block describing no position (0xFFFF vacant marker, base not 2/10, exponent
> 18 -- raw / 10^400 would read as a Good 0.0) publishes BadNoData on that node
alone; siblings are untouched.
- ReadPositionFiguresAsync / IFocasClient.GetPositionFiguresAsync and the
per-axis figure cache are deleted.
Ported from the sister ignitionoee FOCAS driver v0.17.0 (+ Focas.Wire-014);
live axis-scoped captures live-dynamic2-axis{1,2}-0x0026.bin (Z-34184 X 0.0374,
Y 370.177) added as fixtures.
Adds <summary>/<param>/<returns>/<inheritdoc> where missing and removes
project bookkeeping IDs (task/tracking refs) from shipped code comments,
so the docs read cleanly and CommentChecker is quiet except for known
false positives (PLC/protocol terms, event/IEqualityComparer inheritdoc).
Doc/comment-only; no logic changed; solution builds clean.
Equipment tags were stuck at Bad_WaitingForInitialData on the deployed driver: the equipment poll, fixed-tree loop, probe and recycle shared one FOCAS/2 socket with no serialization, and the steady-state read had no timeout — concurrent reads collided and a stalled read hung forever, never overwriting the node's initial-data seed.
- SynchronizedFocasClient: per-device SemaphoreSlim gate + per-call timeout around every wire op (Connect/Probe gated, not double-bounded); wired in EnsureConnectedAsync. ReadAsync/WriteAsync map a per-call timeout to BadCommunicationError instead of rethrowing.
- FlexibleStringConverter on FOCAS config Series: the AdminUI persists the enum as a number ("series":6); accept number-or-string instead of throwing -> stub.
- FocasHostAddress.TryParse tolerates a scheme-less {ip}[:{port}] (AdminUI hostAddress form); canonical focas:// unchanged, malformed schemes still rejected.
247 FOCAS tests green; each fix has a regression test. Live-validated on wonder-app-vd03 (tags read Good).
Joseph Doherty