Source hygiene: no ticket numbers, no private-repo citations

.gitignore

@@ -73,3 +73,4 @@ test_hold
 __pycache__/
 *.py[co]
 test_box_facts
+.git-push-failures.log

include/reac/reac.h

@@ -305,7 +305,7 @@ int reac_detect_rate_fd(int fd, int window_ms);
  *        not a patch, because reac_cfg.h's PUBLIC vocabulary moves: REAC_CFG_REFUSED_NONE is
  *        "none" (was ""), REAC_ROLE_PROP and REAC_CFG_ROLE_STATE_HUNTING are ADDED, and
  *        REAC_CFG_RATE_COUNT / REAC_CFG_RATE_LIST_INIT, which nothing read, are REMOVED. A
- *        consumer pinned to the old header (openmixer's TS mirror) must move with it.
+ *        consumer pinned to the old header (a mirror of it in another language) must move with it.
  *        Behaviour: reac_detect_rate_fd measures one stream; a new clock reference is
  *        LOCKING until measured; reac_ctrl_identity_reply requires both checksums;
  *        reac_decode_plain_le refuses an oversize geometry; reac_boxreg refuses an

include/reac/reac_braid.h

@@ -21,7 +21,7 @@
  *   - obs-h8819 (norihiro): convert_to_pcm24lep, developed and LISTENING-
  *     validated against a real Roland M-200i downstream at 48 kHz;
  *   - the FreeREAC rig: the S-1608/S-0808/S-4000 upstream returns decode real
- *     microphones under the braid (reac-pw task #108, upstream fixtures here);
+ *     microphones under the braid;
  *   - zoneA/zoneB goldens (a real M-5000's two REAC ports, program audio)
  *     decode at coherence 0.99 / spectral flatness 0.002 under the braid at
  *     audio offset exactly 50, and as noise under every other layout x offset

include/reac/reac_cfg.h

@@ -7,9 +7,9 @@
  *
  * ONE DECLARATION, BOTH SIDES (spec §2). reac-pw's `param_changed` reads these
  * keys off a segment's `reac-playback[.<inst>]` node and answers on them;
- * openmixer's TS mirror (`@openmixer/core`'s `reac-cfg.ts`) is conformance-
- * pinned against this file byte-for-byte, so a knob cannot exist as a string
- * literal in one repo and silently drift from the other.
+ * a console that mirrors these names is expected to pin them against this file
+ * byte-for-byte, so a knob cannot exist as a string literal in two places and
+ * silently drift.
  *
  * THE CHANNEL IS PROPS, NOT A FILE. A config file holds STARTUP values and
  * fixed site facts, read once and never watched; every LIVE reconfiguration —

include/reac/reac_clock.h

@@ -174,7 +174,7 @@ const char *reac_clock_quality_name(enum reac_clock_quality q);
 /* ELIGIBILITY (issue #77 item 5). Is this tier fit to own a REAC segment's pace?
  * Reporting-only and PURE: this module never yanks a reference out from under a
  * running segment on its own opinion — that decision belongs to whoever offers
- * the master/follower control (the Setup -> Clock panel, openmixer). What we owe
+ * the master/follower control (a console's clock panel). What we owe
  * them is an unambiguous answer, and MARGINAL answering "no" is the whole point:
  * free-running while owning a segment is a legitimate emergency configuration,
  * not something to slide into by accident. */

include/reac/reac_ctrlblk.h

@@ -242,7 +242,7 @@ int reac_ctrl_scene_build(uint8_t *body, size_t n, const uint8_t mac[6]);
  * a constant that is not settled. Sensitivity is the input level that reaches
  * full scale, so it runs the OTHER WAY from gain: the hottest setting is the
  * most negative number. Against a 0 dBu reference this library's curve is
- * equivalent to gain_dB = 10 + value, and openmixer publishes the same control
+ * equivalent to gain_dB = 10 + value, and a console may publish the same control
  * as 0..55 dB, a different zero by 10 dB.
  *
  * NEITHER IS SILENTLY CONVERTIBLE INTO THE OTHER. The loopback that measured the

include/reac/reac_disco.h

@@ -13,7 +13,7 @@
  * ownership-blind, and its verdict never reaches the FSM.
  *
  * Presence is NEVER a heuristic. A sighting requires a 0x8819 frame and (for non-FILLER) a
- * verified checksum. A packet COUNT is not evidence — openmixer #161 announced any NIC over
+ * verified checksum. A packet COUNT is not evidence — a consumer that announced any NIC over
  * ~62 pkt/s (ordinary Wi-Fi) as a REAC master because an rx_packets delta was fed in as a
  * device.
  *
@@ -50,8 +50,7 @@
  * alloc, no stdio) — the pacer thread calls them, one `reac_disco_peer_lock` per segment's own
  * pacer (so its lifetime matches the segment's: a drop/reopen gets a fresh, unlocked lock).
  * The TABLE is main-thread-only: sightings cross threads on the pacer's existing lock-free
- * event ring, so there is no new cross-thread primitive here. See
- * 2026-07-16-reac-discovery-via-reac-pw (openmixer) for the seam. */
+ * event ring, so there is no new cross-thread primitive here. */
 #ifndef REAC_DISCO_H
 #define REAC_DISCO_H
 

include/reac/reac_headamp_tx.h

@@ -1,7 +1,7 @@
 // SPDX-License-Identifier: GPL-3.0-or-later
 // Copyright (C) 2026 Pau Aliagas <linuxnow@gmail.com>
 
-/* reac_headamp_tx — the MASTER-role head-amp SEND model (task #155 reac-pw half).
+/* reac_headamp_tx — the MASTER-role head-amp SEND model.
  *
  * The protocol, as the committed captures show it (HEADAMP-PROTOCOL-AUDIT-2026-07-22
  * in reac-firmware-re; pcaps m200-s1608-BIDIR-reboot, matrix-m200-s1608 and
@@ -71,7 +71,7 @@
  * the box's own panel within one cadence; that authority over 48 V belongs to the
  * operator, and the mechanism is rig-unproven — no capture in this repo shows a
  * box receiving a redundant head-amp record, because no master here ever sent
- * one. Until it is accepted and capture-gated, openmixer's scene watch
+ * one. Until it is accepted and capture-gated, a controller's scene watch
  * (REAC_SCENE_SETTLE_REPLAY_MS) remains the writer that re-applies a recorded
  * scene, and this master asserts once, as a real M-200 does.
  *
@@ -97,7 +97,7 @@ struct reac_headamp_setting {
 
 /* Pack/unpack a (ch, param, value) head-amp command into ONE 32-bit word.
  *
- * This is the atomicity primitive for the LIVE control path (task #203): a
+ * This is the atomicity primitive for the LIVE control path: a
  * controller's per-channel phantom/pad/sens change travels from the PipeWire
  * main-loop thread (the prop handler) to the RT pacer thread through a lock-free
  * SPSC queue of these words. Because the whole triple lives in a single 32-bit

include/reac/reac_hunt.h

@@ -13,9 +13,8 @@
  * down for any of it: a role a normal box needs hand-set is a defect in the defaults,
  * exactly as a hand-set rate would be.
  *
- * (openmixer's REAC master-arbitration and trunk-VLAN daemon notes describe the same
- * three outcomes from the CONSOLE's side, which is what its `auto` role projects. They
- * are a useful cross-reference and they are not this repo's law; DESIGN.md is.)
+ * (A console describes the same three outcomes from its side, which is what an `auto`
+ * role projects. DESIGN.md is this repo's law.)
  *
  * IT DECIDES NOTHING BY ITSELF THAT reac_arbitration HAS NOT ALREADY OBSERVED.
  * reac_arbitration stays passive — it publishes who drives the segment — and this

include/reac/reac_link_state.h

@@ -2,7 +2,7 @@
 // Copyright (C) 2026 Pau Aliagas <linuxnow@gmail.com>
 
 /* reac_link_state — the PURE mapping from the MASTER FSM (reac_master_state)
- * onto the small badge vocabulary a PipeWire consumer (openmixer's stagebox
+ * onto the small badge vocabulary a PipeWire consumer (a stagebox
  * card) reads off the reac-capture / reac-playback node properties, so it can
  * show "no box / probing / connected" without inferring it from wire traffic.
  *
@@ -77,10 +77,9 @@
  *
  * IT IS PUBLISHED BECAUSE reac.master.mac IS NOT IT AND CANNOT BE. That key names
  * whoever drives the segment, which in the master role is THIS machine's NIC. A
- * consumer that keys a stagebox registry by box address — openmixer keys its
- * `reac:<box mac>` patch names exactly that way — read reac.master.mac, matched
+ * consumer that keys a stagebox registry by box address — for example by
+ * `reac:<box mac>` patch names — read reac.master.mac, matched
  * our own NIC against a registry of Roland addresses, and missed on every rig
- * (openmixer 2026-09-06-rig-headamp-and-clip-findings §5).
  * The alternative was for every consumer to re-derive the peer from a discovery
  * sighting list, which is a second implementation of a fact only the master
  * actually holds. */
@@ -98,7 +97,7 @@
  * REAC_PROP_HEADAMP_BASE is the head-amp WIRE CHANNEL the box's input 1 sits at —
  * `base` in `CH = base + (input - 1)`. It is published because the alternative is
  * every consumer re-deriving it from the width against its own copy of the
- * placement table (openmixer does exactly that today), which is a second
+ * placement table (a consumer may do exactly that), which is a second
  * implementation of a policy only reac-pw can actually observe, and it is wrong
  * the moment the placement rule gains a case. Decimal; REAC_BOX_SOURCE_NONE when
  * no box is known, so a numeric parse fails rather than reading as base 0. */
@@ -109,11 +108,10 @@
 #define REAC_BOX_SOURCE_WIRE "wire"   /* the box declared itself; we matched it */
 #define REAC_BOX_SOURCE_NONE "none"   /* no box on this wire — a normal state   */
 
-/* reac.discovery.* — passive discovery (task #178): what is ON THE SEGMENT, as opposed
+/* reac.discovery.* — passive discovery: what is ON THE SEGMENT, as opposed
  * to the reac.link-state/box-* keys above, which describe only the peer THIS master
  * joined. Same seam, same node, same 200 ms poll; see reac_disco.h for the rules that
- * keep the device list honest, and openmixer's
- * 2026-07-16-reac-discovery-via-reac-pw for the contract.
+ * keep the device list honest.
  *
  * A reader that finds reac-playback WITHOUT these keys is talking to a reac-pw that
  * predates discovery — which is "could not scan", never "scanned and found nothing". */
@@ -155,9 +153,8 @@
  * live. Adding a socket, a file or an HTTP endpoint would be a second ledger for
  * the same facts, with neither door announcing the other. One store, one writer.
  *
- * The console side is NOT built here (another lane owns that repo). openmixer's
- * telemetry is a settled design — its own SSE, latest-wins, skip when late, never
- * accumulate — and these rows are shaped to drop straight into it: every one is a
+ * The console side is NOT built here. A console's telemetry is expected to be latest-wins,
+ * skip when late, never accumulate, and these rows are shaped to drop straight into it: every one is a
  * scalar over the window that just closed, so a consumer that misses three
  * updates has lost nothing but resolution.
  *

include/reac/reac_master.h

@@ -10,7 +10,7 @@
  * the real cdea/cfea control block into the 32-byte block [18:50] of the
  * downstream broadcast frame that libreac's reac_downstream_build builds.
  *
- * EVENT-DRIVEN (task #130): a real M-5000 never advances the establishment on a
+ * EVENT-DRIVEN: a real M-5000 never advances the establishment on a
  * timer — it PROBES until the box's cold-connect (cdea 04 03) arrives, ECHOES
  * that block back as the grant burst, and settles the instant the box switches
  * to unicast. The earlier cut auto-advanced PROBING->GRANTING after ~1 s blind
@@ -247,8 +247,8 @@ const struct reac_mixer_profile *reac_mixer_profile_by_name(const char *name);
 const struct reac_mixer_profile *reac_mixer_profile_at(int i);
 
 /* Resolve the master's EMISSION rate for a requested --rate against the
- * impersonated mixer profile (task #156, "96k is not anything different, same
- * state diagram, doubled frequency" — parameterizing the existing 48k path by
+ * impersonated mixer profile (96k uses the same
+ * state diagram at a doubled frequency: parameterizing the existing 48k path by
  * mixer profile rather than re-engineering the FSM).
  *
  * THE RATE IS INDEPENDENT OF WHICH DESK WE IMPERSONATE (operator, 2026-08-21).

include/reac/reac_role.h

@@ -5,7 +5,7 @@
  * parse + validation contract is unit-testable without PipeWire/sockets.
  *
  * REAC has no fixed master: any box can be the master and the rest slave to it
- * (REAC-PROTOCOL-AND-TESTS.md §2/§4). openmixer must fit either role:
+ * (REAC-PROTOCOL-AND-TESTS.md §2/§4). a console must fit either role:
  *   - master (default): WE drive the cdea/cfea establishment and WE GENERATE THE
  *     PACE (the SCHED_FIFO pacer's cadence); a stagebox slaves to us.
  *   - slave: an EXTERNAL master drives; we RESPOND and lock to its cadence (frame
@@ -55,7 +55,7 @@ static inline const char *reac_role_name(enum reac_role r)
  *          slave to it; a STAGEBOX masters it, we refuse and say so. reac_hunt.h
  *          resolves it from what the wire actually says.
  *
- * A console face may speak its own words for the same two ends (openmixer's shows
+ * A console face may speak its own words for the same two ends (one may show
  * `mixer`/`recorder`); REAC_ROLE and this parser stay on the WIRE's words, and the
  * translation is the console's to do. */
 enum reac_role_intent {
@@ -66,8 +66,7 @@ enum reac_role_intent {
 	 * all: no announce, no join, no grant, no seglock, no TX socket. It is the only
 	 * intent that has no wire role even after the wire has answered, which is why it
 	 * cannot be spelled as a value of `enum reac_role` above. See
-	 * <reac/transport/reac_tap.h>, and openmixer's master-arbitration spec, eighth
-	 * amendment (2026-09-13): beside a real desk, a courting slave of ours kept that
+	 * <reac/transport/reac_tap.h>: beside a real desk, a courting slave of ours kept that
 	 * desk's own box from enrolling for 180 s and a granted one blocked it outright.
 	 * EXPLICIT ONLY — `auto` never resolves to it. */
 	REAC_ROLE_INTENT_TAP,

include/reac/reac_upstream.h

@@ -15,7 +15,7 @@
  * reac_decode(): it uses the channel-pair byte BRAID — the REAC wire layout,
  * see <reac/reac_braid.h> for the byte map and the full evidence trail.
  *
- * Resolved 2026-07-10 (reac-pw task #108) against the rig captures in
+ * Resolved 2026-07-10 against the rig captures in
  * reac-captures: under the braid the loud wired capture's music channel decodes
  * at lag1 autocorrelation +0.998 and every idle channel collapses to the mic
  * noise floor; under plain LE every channel reads sign-extension garbage (the

include/reac/transport/reac_conf.h

@@ -17,7 +17,7 @@
  *
  *   1. THE COMMAND LINE            --rate, --live, ...
  *      Handled by main.c, not here: an explicit argument is not a lookup. It is
- *      the layer openmixer uses — the console owns the desk's configuration and
+ *      the layer a console uses — the console owns the desk's configuration and
  *      hands it over when it launches us, which is what "omx needs to override"
  *      means in practice.
  *
@@ -31,8 +31,8 @@
  *      boxes, different NICs, potentially different rates. A per-segment fact
  *      is the key SUFFIXED with the segment's name — REAC_ROLE_enp131s0 — in
  *      any of the layers below, and it outranks the bare key in every one of
- *      them. Segments are discovered, not declared (openmixer's
- *      2026-08-23-reac-trunk-vlan-daemon, amendment 2026-09-02), so there
+ *      them. Segments are discovered, not declared (every linked wired
+ *      interface is listened on), so there
  *      is no per-segment FILE to create: a segment's name is its interface's,
  *      and the console generates the key into reac-pw.env.
  *
@@ -41,7 +41,7 @@
  *
  *   5. ~/.config/openmixer/reac.env        LAST RESORT
  *      The bottom layer. It is what makes a STANDALONE install work with no
- *      console present, and openmixer overrides it and surfaces the value to the
+ *      console present, and a console overrides it and surfaces the value to the
  *      user. It is NOT a stray duplicate and it is NOT to be deleted; it is the
  *      floor of the stack, and a floor is the thing you land on when nothing
  *      above answered.

include/reac/transport/reac_ifname.h

@@ -2,7 +2,7 @@
 // Copyright (C) 2026 Pau Aliagas <linuxnow@gmail.com>
 
 /* reac_ifname — a segment's STABLE, PHYSICAL-ADDRESS-DERIVED name: "pci<N>" / "usb<N>",
- * ported from openmixer's classifyBus (packages/server/src/net-interfaces.ts) with one
+ * derived from the sysfs bus of the interface, with one
  * correction forced by measurement (see reac_ifname.c's header comment): a raw kernel
  * interface NAME (enp128s20f0u2, eth0, ...) is assigned by udev from bus topology or plain
  * enumeration order, and BOTH can change under the operator with the physical NIC untouched

include/reac/transport/reac_ifscan.h

@@ -3,8 +3,7 @@
  *
  * reac_ifscan — WHICH interfaces to listen on, and which of them are SEGMENTS. The host's
  * netdev table, watched over rtnetlink, folded into one decision per interface
- * (openmixer's 2026-08-23-reac-trunk-vlan-daemon §7-§9, amendment
- * 2026-09-02: every linked WIRED interface is listened on; nothing is declared per NIC).
+ * (every linked WIRED interface is listened on; nothing is declared per NIC).
  *
  * LINK IS THE GATE TO LISTEN; HEARING IS THE GATE TO SERVE. An Ethernet interface that carries
  * link (IFF_LOWER_UP) is SNIFFED — a passive 0x8819 socket that transmits nothing, so an office