docs(phase-1/m5.5): HiDPI correction — Xft.dpi is the real scaling mechanism

Controlled re-runs kill the M5.3 sun.java2d.uiScale verdict: the jpackage
launcher JVM picks the property up yet renders 1:1, in-process setProperty is
also inert, and 1.7.3 Window/rememberWindowState expose no scale arg (javap).
Xft.dpi IS the lever: xrdb 192 + plain bin/incido -> 1920x1016 (1090.dp
clamped), pixel RMSE 0.071 for the 2x-upscaled scale-1 crop (0.000 identical /
0.250 unscaled controls) -> glyphs render proportionally; HiDPI check CLOSED.
Also recorded: uinput virtual keyboard dropped by the compositor like XTEST
(keypress check is physical-keys-only), branch pushed to origin, and the tag
run held for user decision since v*.*.* also fires release.yml.

Verify: bin/incido under Xft.dpi=192 geometry 1920x1016, under 96 -> 1090x730
restored; magick compare RMSE 0.0707 (b6 vs 2x-a), controls 0 and 0.2497;
Main.kt --ui-scale experiment reverted (git checkout, tree was clean).
This commit is contained in:
Incido Developer committed 2026-10-06 23:52:45 +13:00
1 parent 9f6124a501
commit 765191ac8e
2 files changed
+56 -10

No files matched your search

+45 -8
View File
@@ -924,6 +924,8 @@ verdict.
geometry — 1090x730 → clamped at 1920x1016 by the screen (requested geometry — 1090x730 → clamped at 1920x1016 by the screen (requested
2180px-wide; the compositor's cap PROVES the scale applied). Font 2180px-wide; the compositor's cap PROVES the scale applied). Font
crispness at scale not machine-verified; that part is pending-manual. crispness at scale not machine-verified; that part is pending-manual.
**⚠ SUPERSEDED — see "HiDPI correction" under M5.5: this verdict did
NOT reproduce in a controlled re-run; Xft.dpi is the real mechanism.**
- **Native-Wayland verdict:** A probe compiled+run on BOTH installed - **Native-Wayland verdict:** A probe compiled+run on BOTH installed
JDKs (Arch 17.0.20.1 and 27) with `WAYLAND_DISPLAY=wayland-0` JDKs (Arch 17.0.20.1 and 27) with `WAYLAND_DISPLAY=wayland-0`
present resolves `sun.awt.X11.XToolkit` either way — this stack has no present resolves `sun.awt.X11.XToolkit` either way — this stack has no
@@ -939,6 +941,7 @@ M5.3 suite:
image bin/incido … --self-test SELFTEST id=1 → RENDER notes=1, rc=0 image bin/incido … --self-test SELFTEST id=1 → RENDER notes=1, rc=0
(pre-fix image run) NoClassDefFoundError sun/misc/Unsafe → fixed (pre-fix image run) NoClassDefFoundError sun/misc/Unsafe → fixed
uiScale 1 vs 2 window geometry 1090x730 → 1920x1016 (clamp) = scaling applied uiScale 1 vs 2 window geometry 1090x730 → 1920x1016 (clamp) = scaling applied
(⚠ NOT reproducible — see M5.5 correction)
AwtProbe on JDK17 / JDK27 (Wayland env) both sun.awt.X11.XToolkit (no native Wayland here) AwtProbe on JDK17 / JDK27 (Wayland env) both sun.awt.X11.XToolkit (no native Wayland here)
:desktopApp:run … --keys-smoke KEYS ALL-OK :desktopApp:run … --keys-smoke KEYS ALL-OK
:core:data-jvm:smoke RESULT: M1a PASS :core:data-jvm:smoke RESULT: M1a PASS
@@ -997,14 +1000,48 @@ frame budget at 3000. The §8 "perf on a multi-thousand-note library"
unknown is CLOSED. unknown is CLOSED.
**M5 — PACKAGING AND INTEGRATION: DONE** (sub-milestones M5.1–M5.5 **M5 — PACKAGING AND INTEGRATION: DONE** (sub-milestones M5.1–M5.5
above, one commit each). Remaining pending-MANUAL items, all recorded above, one commit each).
where they arose:
1. One real Esc / Ctrl+N keypress at the desk (`esc(captured)` logs — **HiDPI correction + remote-verification session (06/10, after M5
Wayland blocks headless injection). commit):** branch pushed to origin (`spike/desktop-feasibility`);
2. HiDPI font crispness eyeball at `uiScale=2` (geometry scaling proven). three pending checks attacked remotely with mixed results:
3. First real tag run of `desktop-release.yml` (job steps verified
locally piecewise). 1. **Real keypress — harder NO.** `/dev/uinput` became writable (user
4. Carried from M3: the physical phone→desktop recovery-key leg. grant) so a kernel-level virtual keyboard was built (uinput device,
libinput-visible, same mechanism as ydotool) and fired at the
focused window — plus three XTEST/XSendEvent variants earlier. All
silent on this KDE-Wayland session (`KEYS` never logged although
focus confirmed). Only a physical keyboard remains.
2. **HiDPI — CLOSED machine-verified, but the M5.3 verdict was WRONG.**
Controlled re-runs of `-Dsun.java2d.uiScale=2.0`: picked up by the
jpackage launcher's JVM ("Picked up" in log) → NO scaling
(1090x730); set in-process via System.setProperty before
application{} → NO scaling; `Window(uiScale=)`/
`rememberWindowState(uiScale=)` don't exist in 1.7.3 (javap of
ui-desktop-1.7.3.jar: no scale arg in ANY Window overload; the
earlier failed attempts surfaced as the usual overload-ambiguity
lambda error). **`Xft.dpi` IS the mechanism:** `xrdb` Xft.dpi 192 +
plain `bin/incido` (no env at all) → geometry **1920x1016** (1090.dp
→ 2180 px, clamped to screen). Pixel check: scale-1 header crop
upscaled 2× vs scale-2 crop → RMSE **0.0707** (controls: 0.000
identical, 0.250 unscaled) → glyphs/layout render proportionally at
2× — crispness at scale is machine-verified; a human only needs to
LIKE it (hinting taste-level). Linux X11 HiDPI = Xft.dpi;
`sun.java2d.uiScale` is the Windows path. Gradle's daemon also
swallows `JAVA_TOOL_OPTIONS` before forked run-JVMs — scaling tests
must use the direct `bin/incido` binary + xrdb, never `gradlew run`.
Phase-2 in-app scale setting: revisit with Compose 1.8 per-window
scale (CMP-5295) or self-apply content density — 1.7.3 has no knob.
3. **First tag run — deliberately NOT done.** `v*.*.*` also fires
`release.yml` (publishes to the user's real store channel); needs a
user decision, see below.
Remaining pending-MANUAL list (was four):
1. One real Esc / Ctrl+N keypress at the desk — compositor drops EVERY
synthetic route incl. uinput; only physical keys count.
2. ~~HiDPI eyeball~~ CLOSED above (Xft.dpi route, RMSE 0.0707).
3. First real tag run of `desktop-release.yml` — waiting on user.
4. Carried from M3: phone→desktop recovery-key leg (physical device).
**Next up: Phase 2 selection (§8)** — streaming backup formatVersion 2 **Next up: Phase 2 selection (§8)** — streaming backup formatVersion 2
driven by the seed numbers above, voice notes on desktop, OS-keyring driven by the seed numbers above, voice notes on desktop, OS-keyring
+11 -2
View File
@@ -121,7 +121,9 @@ app version `0.4.2.0` (code 38).
(no onClick); menu members are capitalized `Item`/`Separator`/`CheckboxItem`. (no onClick); menu members are capitalized `Item`/`Separator`/`CheckboxItem`.
* KWin under Wayland drops xdotool XTEST keys aimed at the XWayland window even * KWin under Wayland drops xdotool XTEST keys aimed at the XWayland window even
when focus probes show it focused — desktop key handling can't be injected when focus probes show it focused — desktop key handling can't be injected
headless here; verify real keypresses by hand. headless here; verify real keypresses by hand. (06/10: a kernel-level
`/dev/uinput` virtual keyboard — libinput-visible, the ydotool mechanism —
is ALSO dropped. No synthetic route works on this stack.)
* The jlink runtime inside the app-image lacks `jdk.unsupported`; DataStore's * The jlink runtime inside the app-image lacks `jdk.unsupported`; DataStore's
protobuf crashes with `NoClassDefFoundError: sun/misc/Unsafe` ONLY inside the protobuf crashes with `NoClassDefFoundError: sun/misc/Unsafe` ONLY inside the
image (`:desktopApp:run` on a full JDK never sees it). Shipped via image (`:desktopApp:run` on a full JDK never sees it). Shipped via
@@ -130,6 +132,13 @@ app version `0.4.2.0` (code 38).
* Compose 1.7.3's `nativeDistributions` DSL has no `iconFile` property; icons * Compose 1.7.3's `nativeDistributions` DSL has no `iconFile` property; icons
travel through `packaging/incido.desktop` (app-image) or the deb install travel through `packaging/incido.desktop` (app-image) or the deb install
step, not the DSL. step, not the DSL.
* HiDPI on Linux X11 = `Xft.dpi` (xrdb) ONLY. `sun.java2d.uiScale` does nothing
for Compose density here even when the jpackage launcher JVM picks it up, and
1.7.3 `Window` exposes no scale arg at all (javap-verified — its absence
surfaces as the usual overload-ambiguity lambda error). Gradle's daemon also
swallows `JAVA_TOOL_OPTIONS` before forked run-JVMs. Scale tests: direct
`bin/incido` + `xrdb`, never `gradlew run` (06/10; 2× render proven by pixel
RMSE 0.071 vs 0.250 unscaled control).
## 4. M2 — the backup bridge (next up; the product's reason to exist) ## 4. M2 — the backup bridge (next up; the product's reason to exist)
@@ -247,7 +256,7 @@ Exit criteria: one shared UI codebase; Android visually unchanged
(`assembleDebug` + on-device smoke); desktop shows the real list, editor and (`assembleDebug` + on-device smoke); desktop shows the real list, editor and
settings wearing Incido's theme. settings wearing Incido's theme.
## 7. M5 — packaging and integration — **DONE** (M5.1–M5.5, run logs in `DESKTOP_FEASIBILITY.md`; pending-manual list recorded there: real keypress, uiScale eyeball, first tag run of `desktop-release.yml`, M3's phone→desktop recovery-key leg) ## 7. M5 — packaging and integration — **DONE** (M5.1–M5.5 + HiDPI correction, run logs in `DESKTOP_FEASIBILITY.md`; pending: real keypress — only physical keys work, uinput also dropped; first tag run of `desktop-release.yml` — awaiting user; M3's phone→desktop recovery-key leg. HiDPI check CLOSED: Xft.dpi route, pixel-verified)
`packageDistributionForCurrentOS` per OS on a GitHub Actions matrix (no `packageDistributionForCurrentOS` per OS on a GitHub Actions matrix (no
cross-compiling); icons, `.desktop` entry, tray + single-instance, first-run cross-compiling); icons, `.desktop` entry, tray + single-instance, first-run