#!/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 = ${WAKE_AGENT:-default} (per-agent queue namespace) # STATE_DIR = / # --------------------------------------------------------------------------- wake_state_dir() { local base agent base="${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). # _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="$1" dir tmp dir="$(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" }