igneum-network/igneum
1
0
Fork 0
Code Issues Pull requests Releases Activity

Lock: three run slots for functional runs; a measurement waits for all of them

Browse source 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
igneum-labs 2026-10-05 01:10:17 +00:00
parent 00baf7501d
commit be99cb3821
1 changed files with 17 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

21
tools/lock/with-lock.sh
View file

@ -5,7 +5,8 @@
# three at once use 12 of the 16 cores); waits when all are busy
# tools/lock/with-lock.sh measure <command...> exclusive: waits for every build slot and blocks new builds
# tools/lock/with-lock.sh run <command...> a functional run (test network, attack harness, simulator whose
# outputs are counts, locks, forks, seconds): one at a time, builds continue
# outputs are counts, locks, forks, seconds): one of THREE run slots,
# builds continue; a measurement waits for all of them
# tools/lock/with-lock.sh status who holds what, and for how long
# Rule (4 October 2026): code and documents run in parallel; builds share three slots; measurements are exclusive.
# Evening of 4 October: one build slot and a 3-hour network run under `measure` left twelve agents queued; hence the
@ -15,7 +16,7 @@ set -euo pipefail
kind="${1:-}"; shift || true
dir=/tmp/igneum-locks; mkdir -p "$dir"
if [ "$kind" = status ]; then
for f in measure build build-0 build-1 build-2; do
for f in measure build build-0 build-1 build-2 run-0 run-1 run-2; do
[ -e "$dir/$f" ] || continue
holder=$(python3 - "$dir/$f" <<'PY'
import fcntl, sys, os
@ -70,10 +71,22 @@ if mode == 'build':
if got[0] != k: os.close(fd)
held = [(f"build-{got[0]}", got[1])]
elif mode == 'measure':
for name in ['measure', 'build', 'build-0', 'build-1', 'build-2']:
for name in ['measure', 'build', 'build-0', 'build-1', 'build-2', 'run-0', 'run-1', 'run-2']:
fd = opened(name); wait_lock(fd, name); held.append((name, fd))
else:
fd = opened('measure'); wait_lock(fd, 'measure'); held.append(('measure', fd))
# run: one of THREE run slots (functional runs on their own port ranges can overlap), none while a measurement
# holds every slot; a measurement waits for every run slot in turn
slots = [opened(f"run-{k}") for k in range(3)]
got = None
while got is None:
for k, fd in enumerate(slots):
if try_lock(fd): got = (k, fd); break
if got is None:
if time.time() - t0 > 7200: print("with-lock: gave up waiting for a run slot after 2 h", file=sys.stderr); sys.exit(1)
time.sleep(5)
for k, fd in enumerate(slots):
if got[0] != k: os.close(fd)
held = [(f"run-{got[0]}", got[1])]
for name, fd in held:
os.ftruncate(fd, 0); os.lseek(fd, 0, 0)
os.write(fd, f"pid {os.getpid()} since {time.strftime('%H:%M:%S', time.gmtime())}Z waited {int(time.time()-t0)} s: {' '.join(cmd)[:120]}\n".encode())