feat(phase-1/m5.5): perf pass legs --perf-smoke/--perf-verify; M5 closed

Seeds 3000 notes against the real streaming UI (279 ms total, ~0.09 ms/note),
fresh observeAll 70 ms, favorite toggle 1 ms, reopen first non-empty emission
89 ms; only 4 RENDER lines during the seed (size-keyed emission coalesces).
The multi-thousand-note unknown from the Phase-0 recommendation closes:
desktop perf is a non-event at 3000 notes. §7 marked DONE + handoff
"first actions" rewritten for post-M5 sessions (four pending-MANUAL checks
listed).

Verify: :desktopApp:run --args="--data-dir /tmp/incido-perf --perf-smoke
--exit-after 150" (PERF seed 3000/279ms, rc 0) then --perf-verify same dir
(reopen 89 ms, rc 0); :desktopApp:createDistributable + image --self-test
rc=0; --keys-smoke KEYS ALL-OK; :core:data-jvm:smoke M1a PASS;
assembleDebug SUCCESSFUL.
This commit is contained in:
Incido Developer committed 2026-10-06 23:18:27 +13:00
1 parent 6f16e87ec0
commit 9f6124a501
3 files changed
+104 -8

No files matched your search

+38
View File
@@ -972,6 +972,44 @@ M5.3 suite:
**Next up: M5.5** — §9 perf pass on the real screens.
**M5.5 — perf pass on the real screens: DONE.**
New `--perf-smoke` (fresh dir) / `--perf-verify` (reopen the same dir)
legs run against the REAL screens with the window composed and the list
streaming — recomposition pressure is part of the measurement, not a
separate microbench. Numbers on this dev box (Arch, Ryzen, no
encryption — the field-cipher leg would add the M3-proved ~55 ms PBKDF2
once per unlock, not per note):
```
--perf-smoke seed notes=3000 total=279ms per-note=0ms (UI streaming throughout)
--perf-smoke fresh observeAll notes=3000 in 70ms
--perf-smoke favorite toggle write 1ms
--perf-verify reopen first non-empty emission notes=3000 in 89ms
RENDER lines during the whole seed: 4 (size-keyed emission coalesces —
no per-save recomposition storm)
exit codes 0, clean --exit-after teardown both legs
```
Verdict: a multi-thousand-note library is a non-event on desktop — the
ObjectBox queries, flow emission and LazyColumn all stay far under a
frame budget at 3000. The §8 "perf on a multi-thousand-note library"
unknown is CLOSED.
**M5 — PACKAGING AND INTEGRATION: DONE** (sub-milestones M5.1–M5.5
above, one commit each). Remaining pending-MANUAL items, all recorded
where they arose:
1. One real Esc / Ctrl+N keypress at the desk (`esc(captured)` logs —
Wayland blocks headless injection).
2. HiDPI font crispness eyeball at `uiScale=2` (geometry scaling proven).
3. First real tag run of `desktop-release.yml` (job steps verified
locally piecewise).
4. Carried from M3: the physical phone→desktop recovery-key leg.
**Next up: Phase 2 selection (§8)** — streaming backup formatVersion 2
driven by the seed numbers above, voice notes on desktop, OS-keyring
caching, JBR bundling for native Wayland, macOS/Windows GA.