stack/packages/mosaic/framework/tools/wake/_wake-common.sh

#!/usr/bin/env bash
# _wake-common.sh — shared state resolution + atomic-write primitive for the
# wake/heartbeat durable queue (W2 of the wake canon, EPIC #892).
#
# CONTRACT ANCHORS (docs/scratchpads/heartbeat-planning/CONVERGED-DESIGN.md):
#   §1.2  three-cursor durable queue; ALL state XDG, atomic write-tmp+rename.
#   §2.3  durability is NEVER bypassed.
#
# This file is sourced by store.sh and ack.sh. It defines NO top-level actions;
# sourcing it is side-effect-free except for setting readonly path vars.
#
# Operator-agnostic (framework firewall): state location is derived purely from
# XDG / env. No operator paths, names, or secrets appear here.

# ---------------------------------------------------------------------------
# State layout (XDG). §1.2: "all state XDG".
#   base   = ${WAKE_STATE_HOME:-${XDG_STATE_HOME:-$HOME/.local/state}/mosaic/wake}
#   agent  = ${WAKE_AGENT:-default}   (per-agent queue namespace)
#   STATE_DIR = <base>/<agent>
# ---------------------------------------------------------------------------
wake_state_dir() {
  local base agent
  base="${WAKE_STATE_HOME:-${XDG_STATE_HOME:-$HOME/.local/state}/mosaic/wake}"
  agent="${WAKE_AGENT:-default}"
  printf '%s/%s' "$base" "$agent"
}

# File names within STATE_DIR.
#   observed_seq   — highest observed_seq ever recorded (max monotonic int).
#   consumed_seq   — consumer cursor: top of the contiguous consumed prefix.
#   observed.set   — observed seqs in the live window (> consumed_seq), one int
#                    per line; the gap-detector's source of truth. Immune to
#                    coalescing (coalescing removes a pending ENTRY, never the
#                    fact that its seq was observed).
#   pending.jsonl  — the durable pending-inbox: one entry JSON object per line,
#                    {observed_seq, locators, class, emit_ts, hmac}.
#   ack-ledger.jsonl — local-write-only ack ledger (RECEIVED / CONSUMED).
#   ack-sync.state   — background-sync bookkeeping (last shipped / outage flag).
#   consumed-hashes.jsonl — #932 store-owned last-consumed record: one object per
#                    (kind,id), {kind,id,observed_hash,observed_seq}, written at
#                    consume-truncation. The reconciler's THIRD accounting source
#                    (a consumed state matches neither the truncated inbox nor the
#                    reconciler's own seen-ledger). ADDITIVE / lazily created;
#                    older code ignores it (on-disk read-compat preserved).

# _wake_tmp_glob DIR — the glob used for atomic-write temp files, so readers can
# ignore in-flight/crashed writes. A crash leaves one of these; it is NEVER the
# live file (only rename promotes content), so it can never corrupt a read.
_wake_tmp_prefix='.wake.tmp.'

# _atomic_write TARGET  (content on stdin)
# §1.2 / task: EVERY state mutation is atomic write-tmp+rename. Write a temp
# file in the SAME directory (so mv is a same-filesystem atomic rename), fsync
# is best-effort, then rename over the target. A crash before the rename leaves
# the old target fully intact and a stale .wake.tmp.* that readers ignore.
_atomic_write() {
  local target="$1" dir tmp
  # --- TEST-ONLY FAULT SEAM (issue #934) — PROD-INERT. ------------------------
  # Forces the ALREADY-EXISTING atomic-write failure PATH (the fail-loud +
  # rollback handling #908/#917 built) to be taken for ONE named write target, so
  # the seq-integrity failure assertions (T9 arrow-1 no-burn, T11 cursor-write
  # gate) RUN UNPRIVILEGED in the real non-privileged CI runner instead of being
  # skipped behind an unshare+bind-mount injection. It is honored ONLY when the
  # test-only env var WAKE_TEST_FAULT is explicitly set to name a write point; it
  # writes nothing, adds no new behavior, and changes no on-disk format. No
  # production input (CLI args, locators JSON, on-disk state, watch-list) can set
  # a process env var, so with WAKE_TEST_FAULT unset this is a no-op and the write
  # proceeds exactly as before. Map: pending->pending.jsonl, cursor->observed_seq.
  if [ -n "${WAKE_TEST_FAULT:-}" ]; then
    case "${WAKE_TEST_FAULT}:$(basename -- "$target")" in
    pending:pending.jsonl | cursor:observed_seq)
      cat >/dev/null 2>&1 || true # drain the producer, then report the commit as failed
      return 1
      ;;
    esac
  fi
  dir="$(dirname "$target")"
  [ -d "$dir" ] || mkdir -p "$dir"
  tmp="$(mktemp "$dir/${_wake_tmp_prefix}XXXXXX")" || return 1
  if ! cat >"$tmp"; then
    rm -f "$tmp"
    return 1
  fi
  # Best-effort durability of the temp file before the rename. Not fatal if the
  # platform lacks it — the rename atomicity is the load-bearing guarantee.
  sync "$tmp" 2>/dev/null || true
  if ! mv -f "$tmp" "$target"; then
    rm -f "$tmp"
    return 1
  fi
}

# ---------------------------------------------------------------------------
# Enqueue serialization (single store-side observed_seq allocator, #908).
#
# store.sh is the SOLE allocator of observed_seq: it reads the observed_seq
# cursor, computes next = observed_seq + 1, and writes the pending record +
# observed.set + the cursor as ONE transaction. That read-modify-write MUST be
# serialized so two concurrent enqueues cannot read the same cursor and collide
# on a seq. These helpers take an exclusive lock for the duration of the
# transaction. flock (Linux/CI) is the load-bearing mechanism; where flock is
# absent the open still succeeds so a single-threaded enqueue is unaffected (the
# concurrency guarantee then degrades — the concurrency test SKIPs without flock,
# exactly as the detector's single-instance test already does).
#
# The lock uses a FIXED fd (8) within one store.sh process. Each store.sh
# invocation is its own process (the detector/reconciler call it as a
# subprocess), so fd 8 is always free here and never clashes with the detector
# run-loop lock (fd 9, a DIFFERENT process).
# ---------------------------------------------------------------------------
_wake_lock_acquire() {
  # _wake_lock_acquire LOCKFILE — open fd 8 on LOCKFILE and take an exclusive
  # (blocking) lock. Returns non-zero if the lock cannot be taken.
  local lf="$1" dir
  dir="$(dirname "$lf")"
  [ -d "$dir" ] || mkdir -p "$dir" 2>/dev/null || true
  # Open the lock fd. If the file does not yet exist and the dir is writable it
  # is created; if the dir is read-only but the file exists, opening it O_WRONLY
  # still succeeds (write perm on the file, not the dir).
  # NB: a command-less `exec` redirection persists for the WHOLE shell, so we must
  # NOT append `2>/dev/null` here (it would permanently silence the caller's
  # stderr and swallow every later fail-loud diagnostic). Redirect only fd 8.
  exec 8>"$lf" || return 1
  if command -v flock >/dev/null 2>&1; then
    flock 8 || return 1
  fi
  return 0
}

_wake_lock_release() {
  # _wake_lock_release — drop the enqueue lock (closing fd 8 releases the flock).
  # The `2>/dev/null` is SCOPED to the brace group (suppressing a "bad fd" close
  # error) — it must NOT sit on a bare `exec`, where the redirection would persist
  # for the whole shell and silence every later fail-loud diagnostic.
  { exec 8>&-; } 2>/dev/null || true
}

# _wake_clean_stale_tmp DIR — reap ORPHANED atomic-write temp files left by a
# crash mid-write (a crash before the rename leaves a .wake.tmp.* that no reader
# ever promotes). These are never the live store.
#
# AGE-SCOPED (#927): only tmp files whose mtime is older than
# ${WAKE_TMP_STALE_MIN:-5} minutes are removed. A LIVE in-flight atomic write's
# tmp is at most milliseconds old (mktemp -> cat -> sync -> rename all complete
# well under a second), so it can NEVER match this age filter. That is what makes
# the cleanup safe even if it ever overlaps a concurrent enqueue's atomic write:
# it deletes only DEMONSTRABLY-orphaned tmps, never another process's live write.
#
# This is why #927 is fixed: an UNCONDITIONAL delete of every .wake.tmp.* (the
# old behaviour) clobbered a concurrent enqueue's in-flight tmp -> spurious
# "durable pending write FAILED". It is ALSO no longer invoked from the per-
# enqueue hot path (see _wake_init_dir); it runs only at maintenance / daemon-
# start (store.sh init) and the detector poll tick, where accumulation is bounded
# once per pass rather than raced on every enqueue.
_wake_clean_stale_tmp() {
  local dir="$1" min="${WAKE_TMP_STALE_MIN:-5}"
  [ -d "$dir" ] || return 0
  case "$min" in '' | *[!0-9]*) min=5 ;; esac
  find "$dir" -maxdepth 1 -name "${_wake_tmp_prefix}*" -type f -mmin "+$min" -delete 2>/dev/null || true
}

# _wake_read_int FILE DEFAULT — read a single integer from FILE, or DEFAULT.
_wake_read_int() {
  local file="$1" def="$2" val
  if [ -f "$file" ]; then
    val="$(tr -d '[:space:]' <"$file")"
    case "$val" in
    '' | *[!0-9]*) printf '%s' "$def" ;;
    *) printf '%s' "$val" ;;
    esac
  else
    printf '%s' "$def"
  fi
}

# _wake_init_dir STATE_DIR — ensure the state layout exists; idempotent.
#
# #927: this runs on the HOT enqueue path (cmd_enqueue calls it BEFORE taking the
# enqueue lock) as well as on consume/cursors/ack. It must therefore NEVER touch
# another process's tmp files: the old _wake_clean_stale_tmp call here deleted a
# concurrent enqueue's LIVE in-flight tmp mid-write -> spurious durable-write
# abort. Stale-tmp reaping is now an explicit maintenance action (store.sh init /
# detector poll tick), NOT a side effect of ensuring the layout. Keep this
# function limited to creating the dir + seeding the cursor files.
_wake_init_dir() {
  local dir="$1"
  mkdir -p "$dir" || return 1
  [ -f "$dir/observed_seq" ] || printf '0' | _atomic_write "$dir/observed_seq"
  [ -f "$dir/consumed_seq" ] || printf '0' | _atomic_write "$dir/consumed_seq"
  [ -f "$dir/observed.set" ] || printf '' | _atomic_write "$dir/observed.set"
  [ -f "$dir/pending.jsonl" ] || printf '' | _atomic_write "$dir/pending.jsonl"
  [ -f "$dir/ack-ledger.jsonl" ] || printf '' | _atomic_write "$dir/ack-ledger.jsonl"
}