V1.1 validation: target-machine preview and performance gates

[ashmitvoid]

Validation tranche for #69 after #70/#71 land.

Gate real Ryoku/Hyprland behavior, helper lifecycle/idle expiry, rapid-selection cancellation, large/pathological/corrupt preview inputs, theme/reduced-motion/HiDPI behavior, removable/slow storage, and recovered idle/performance baselines before any release candidate is cut.

[ashmitvoid]

Repository-side preparation for the V1.1 target-machine gate is complete after PR #89 merged. Issue #72 was reopened because the merge closed it before the required real-machine evidence existed.

Current initial validation candidate: 8f82b5c5ccfbea03babb67244b3e42baa43d4ad4 (main).

Harness/docs now on main:

• tools/validation/target-machine.sh
• docs/V1_1_TARGET_MACHINE_VALIDATION.md

CI Build #695 passed on the frozen harness head, including harness self-test, app/tests, FileChooser smokes, portalctl, preview foundation, and staged install. Packaging did not trigger because this was validation/docs-only and is outside its path filter.

The real gate is still pending: exact candidate build on Ryoku/Hyprland; visible-idle + post-stress CPU/PSS samples; <=2 helpers; helper idle expiry; rapid-selection/corrupt/oversized inputs; image/PDF/media/font/Properties UX; live theme/reduced-motion/HiDPI; removable/slow storage; and real FileChooser compatibility.

If any code changes after evidence collection starts, discard that evidence and restart from the new candidate SHA. Keep #69 open until #72 is genuinely complete.

[ashmitvoid]

Target-machine validation update for candidate 8f82b5c5ccfbea03babb67244b3e42baa43d4ad4:

• main application configure/build/test completed successfully: 29/29 tests passed.
• preview-helper configure was blocked before build/test because the target machine did not have the required Arch package poppler-qt6 installed (pkg-config could not resolve poppler-qt6).
• the subsequent ninja: ... build.ninja and ctest: No tests were found messages are consequences of the failed configure and are not product-test results.
• this is an environment prerequisite gap, not a Ryofiles code failure. Install/sync the package dependencies from packaging/arch/PKGBUILD, then rerun preview configure/build/tests on the same candidate SHA before continuing the manual gates.

[ashmitvoid]

Target-machine validation update for candidate 8f82b5c5ccfbea03babb67244b3e42baa43d4ad4:

• main application suite: 29/29 passed
• preview-helper configure/build: passed after installing missing target-machine prerequisite poppler-qt6
• preview-helper test suite: 4/4 passed (preview-scheduler, media-preview, font-preview, image-preview)
• observed compiler warning in src/preview/MediaProbe.cpp about narrowing QImage::bytesPerLine() (qsizetype -> int); non-blocking for this validation pass, but worth a narrow cleanup issue before/after RC if desired

This clears the local build/test prerequisites for the real Ryoku/Hyprland runtime gates. #72 remains open pending staged-run evidence: helper idle expiry, rapid-selection cancellation, visible-idle CPU/PSS, theme/reduced-motion/HiDPI, large-directory stability, removable/slow-storage behavior, and real-session FileChooser compatibility.

[ashmitvoid]

Target-machine retest on merged candidate f654952a9de5175db6a6aafee4c54f6f5747fb7f is clean through the local automated gates:

• main application tests: 29/29 passed
• preview helper tests: 4/4 passed
• root-QML startup smoke: passed (Ryofiles root QML instantiated successfully)

Headless PipeWire/PulseAudio connection warnings were observed during the offscreen smoke, but the application remained alive for the timeout and no QML root/type/property failure recurred. Proceeding to staged real-session launch and shell/performance validation on this exact SHA.

[ashmitvoid]

Target-machine real-session baseline on merged candidate f654952a9de5175db6a6aafee4c54f6f5747fb7f:

• shell-steady snapshot: app_count=1, helper_count=0
• app RSS: 258956 KiB (~252.9 MiB)
• app PSS: 129080 KiB (~126.1 MiB)
• app lifetime CPU snapshot: 4.70% (informational only; interval sampler is the performance source)
• helper RSS/PSS/CPU: 0
• 60s visible-idle-before sampler completed with PASS: preview helper count remained within the process-wide <=2 invariant

This establishes the initial real-session idle baseline and confirms helpers are absent at settled idle. Next gate: trigger one ordinary bounded preview and run the 15s helper idle-expiry check immediately afterward.

[ashmitvoid]

Target-machine performance/lifecycle update for exact candidate f654952a9de5175db6a6aafee4c54f6f5747fb7f:

• real-session steady snapshot: app_count=1, helper_count=0
• app RSS: 258956 KiB (~252.9 MiB)
• app PSS: 129080 KiB (~126.1 MiB)
• 60s visible-idle sample: PASS for process-wide <=2 helper invariant
• visible-idle sample summary: 48 samples, average interval CPU 0.02%, average PSS 129088.6 KiB, min/max PSS 129063/129106 KiB (43 KiB spread), max helpers 0
• helper idle-expiry gate: PASS, helper exited after 7.969s and remained absent long enough to satisfy the collector

These results clear the idle busy-loop/memory-stability baseline and helper idle-expiry gate on the target Ryoku/Hyprland machine. #72 remains open for rapid-selection/newest-selection-wins, corrupt/oversized previews, rich preview UX, theme/reduced-motion/HiDPI, large-directory stress, removable/slow storage, post-stress memory sample, and real-session FileChooser checks.

[ashmitvoid]

Validation is intentionally paused after the real-session UX audit. The candidate f654952a9de5175db6a6aafee4c54f6f5747fb7f has strong automated/idle/helper results, but the visible interaction shell is not release-quality: overlapping controls, broken arrow-key selection/focus, prototype-like hierarchy/iconography, slow/ambiguous preview feedback, and insufficient motion/polish were observed on the target machine. New UX-reset epic #92 with focused children #93–#96 now blocks this gate. Do not continue #72 or cut an RC until the reset lands and target-machine validation is restarted from the new canonical SHA.

[ashmitvoid]

UX reset #93–#96 is now fully merged. Final integrated candidate is 989e54dc494dcb4a1f2f0de3a367815c66f5839e on main (PR #100 exact head 3a4af4ad2b242a703f06bfb422f44321a46c50f3; Build #715 and Packaging #131 both green, including QML startup, preview foundation, exact-SHA package payload and runtime linkage checks).

Previous target-machine evidence was collected against the pre-UX-reset candidate and must not be used to satisfy this gate. Restart the real Ryoku/Hyprland validation from 989e54dc494dcb4a1f2f0de3a367815c66f5839e and keep #72 open until the full manual/performance evidence pass is complete.