Commit graph

16 commits

Author SHA1 Message Date
alt-glitch
dc3c7dc405 opentui(v6): restore scrollback cap 1000 → 3000 under windowing (#27 payoff)
With transcript windowing (S1+S2) the mounted set no longer scales with
the store (peak 31 rows over a 1500-row burst), so the handle-table
clamp that forced 1000 rows is unnecessary when windowing is on. The
ceiling is now windowing-aware: 3000 rows (the originally-shipped
default, regression documented in opentui-fixes-audit.md §2) with
windowing, 1000 with HERMES_TUI_WINDOWING=0 (every row mounts again).

Measured at the restored cap (full 3000-msg store): mem3000 360MB peak
styled end-to-end (pre-campaign: ~870MB + unstyled past ~1,400 rows;
before that: crash). scroll3000 p50=2 p90=3 p99=8 max=17ms (Ink same
workload: p90=35 p99=96). Gate digest unchanged.
2026-06-12 08:29:24 +05:30
alt-glitch
fd956d3189 bench: S2 controller verification — mem2000 307/373MB, scroll p99 6ms, digest unchanged 2026-06-12 08:29:24 +05:30
alt-glitch
f7381800f7 bench: windowing A/B knobs + S1 results — windowing −170MB, GC laziness −90MB at 2k msgs
composeEnv passes HERMES_TUI_WINDOWING through (clean env stripped it);
--heap N overrides the mem-cell V8 cap. Same-build A/B at 2000 msgs:
OFF 686MB peak / ON 518MB / ON+512MB-heap 427MB; scroll p99 16ms vs
17ms baseline (no jank regression), determinism digest byte-identical.
Residual slope ~108MB/1k (Ink ~37) — burst-time mounted peak is the S2
target.
2026-06-12 08:29:24 +05:30
alt-glitch
6c76908fde bench: emulator-leg memory verification — tmux server flat ~5MB for both UIs
Pipeline cell re-run with external VmRSS/VmHWM sampling on the dedicated
tmux servers: ink 5.07MB peak, otui 4.94MB peak, zero growth across the
800-msg stream (alt-screen = fixed grid, no scrollback accrual). The
emulator leg is a tie on memory as well as CPU.
2026-06-11 10:30:41 +05:30
alt-glitch
7776aeb064 bench: render refresh + controller gate-replay results
Re-rendered after committing the verification gate replays so the
report's result-file count matches the results directory.
2026-06-11 09:36:18 +05:30
alt-glitch
ad16ec9c53 bench: report rewrite — plain-language verdicts up top, real-workload memory framing, chaos/pipeline/echo sections 2026-06-11 09:32:41 +05:30
alt-glitch
de446a26a5 bench: chaos + pipeline + echo results — both UIs auto-heal gateway death; total-pipeline CPU is parity
Chaos (5 scenarios x 2 UIs): every gateway-death/hang scenario fully
recovers on BOTH UIs — Ink respawns immediately (~80ms, no backoff),
OpenTUI after its 1s backoff; transcripts converge byte-identical to
the never-killed digest; zero orphans. PTY EOF: both exit and reap the
gateway in ~100ms (Ink takes 4.1s to die vs OpenTUI 0.2s).

Pipeline (800 msgs @30ev/s inside a dedicated tmux server): UI CPU
82.4s (ink) vs 79.1s (otui); tmux-server leg ~0.4s BOTH — the
'Ink costs more in the emulator' hypothesis is not supported at this
workload. Frame pacing: otui 22.3fps vs ink 15.8fps, interframe p95
103ms vs 209ms. Echo latency: both excellent (p50 1-2ms); submit to
first-token-paint 44ms (ink) vs 107ms (otui).
2026-06-11 09:19:07 +05:30
alt-glitch
af1e4bb9ab bench: total-pipeline CPU (tmux leg) + frame pacing + input-echo latency 2026-06-11 09:16:08 +05:30
alt-glitch
22792d2791 bench: chaos/stability cells — gateway death, hang, resize storm, PTY EOF 2026-06-11 09:15:04 +05:30
alt-glitch
cbe703cf48 bench: forensics.sh — merged gateway/TUI/OOM/sessions/worktree timeline for a time window
Reconstructs what killed gateways and TUI sessions: merges
~/.hermes/logs, journalctl/dmesg OOM kills, sessions-DB abnormal ends,
and git worktree state into one timestamp-sorted timeline plus a
summary (OOM victims, tui_gateway exit-reason histogram, orphaned
sessions). Findings from the 2026-06-04..11 window: gateway deaths were
kernel OOM kills selecting hermes via OOMScoreAdjust=200 under pressure
from OTHER tools' multi-GB leaks; TUI deaths were top-down SIGHUP/EIO
cascades from unit teardown; tui_gateway itself crashed in 0/152 exits.
2026-06-11 08:18:55 +05:30
alt-glitch
5c6438fd28 bench: T1 real-workload memory cells — session-DB distribution + mem100/300/2000
Real sessions DB (~/.hermes/state.db, 444 tui+cli sessions): p50=20,
p75=53, p90=182, p95=340, p99=1941 msgs. The assumed 200-300 'realistic
band' is actually the p90-p95 region; the typical session is ~20 msgs
and the p99 tail reaches ~1940 (real ~2k-msg sessions exist).

New cells at those anchors (2 reps, ink vs otui-capped, 2G scope), VmHWM
medians: 100 msgs 163 vs 222MB; 300 msgs 180 vs 268MB; 2000 msgs 234 vs
671MB (2.9x). session-distribution.mjs regenerates the JSON; run.mjs
gains --configs to skip redundant configs (otui-uncapped == capped below
the 3000-row cap).
2026-06-11 08:13:00 +05:30
alt-glitch
805e08081f bench: document build/run parity audit (expose-gc inert; pinned-Node caveat) 2026-06-11 08:01:39 +05:30
alt-glitch
fe50861c2e bench: live-attach kit — sample/profile a running TUI session (Ink or OpenTUI) 2026-06-11 07:34:08 +05:30
alt-glitch
e3973050df bench: post-fix otui results — crash eliminated
Re-ran the cells that crashed, against the fixed binary (a939c9a):

- mem3000 otui-capped:   crashed_after_stream exit 7 @ ~900MB
                       → completed exit 0, vmhwm 859MB
- mem3000 otui-uncapped: crashed_after_stream exit 7
                       → completed exit 0, vmhwm 834MB
- slope10000 otui-uncapped: died exit 7 at 3,000 msgs (197d499 result)
                       → completed ALL 10,000 msgs, exit 0, vmhwm 1.57GB —
                         under the 2GB cgroup cap, no cap-hit, no crash
- fresh ink baselines for both cells (mem3000 257MB / slope10k 328MB).

No "Failed to create SyntaxStyle" anywhere; the store cap now binds at the
handle-safe ceiling (1000 rows) long before the native 65,534-slot handle
table exhausts. report.html + report-assets regenerated via render.mjs
(results are append-only; pre-fix runs remain as the baseline).
2026-06-11 04:09:27 +05:30
alt-glitch
e35d953a45 bench: E1/E3 results + report render 2026-06-11 03:22:27 +05:30
alt-glitch
14ee1a52c0 bench: fake-gateway + PTY harness + matrix runner (methodology in docs/plans) 2026-06-11 02:29:37 +05:30