The Apple HUD froze its clock offset at connect, so every host-anchored number lied — and the tvOS present floor is closed on sound evidence now #208

Merged

enricobuehler merged 2 commits from worktree-appletv-present-depth into main 2026-08-13 21:21:43 +00:00

Conversation 0 Commits 2 Files Changed 9

+366 -53

2 Commits

Author	SHA1	Message	Date
enricobuehler	90d13de81e	fix(apple): the stats overlay lied three ways — a frozen clock offset, silently trimmed impossible samples, and Int -1 fallbacks printing as NaN ... Field 2026-08-13, Apple TV vs Bazzite VM host, two sessions minutes apart on the same wire: hostnet_p50 read 17-21 ms, then a physically impossible 4.4 ms (host-side encode alone is ~4.7). Root causes, each its own defect: - The client consumed the CONNECT-TIME skew offset and froze it: cached in a Stage2Pipeline field, in a StreamPump let, and in a ContentView closure CAPTURE LIST feeding the hostnet meter and the host/network splitter. The core keeps a live estimate (punktfunk_connection_clock_offset_now_ns, ABI v10, re-synced every 60 s + on suspected wall-clock steps) and its own doc says the connect-time value 'silently corrupts every capture-clock comparison' after an NTP step — a VM host steps. Now PunktfunkConnection.clockOffsetNs IS the live read (an atomic load behind the FFI) and every consumer reads it at use: per record, per AU, per enqueue. The Swift audio plane's AvSync observation gets the live value through the same property. - LatencyMeter's impossible-sample guard (≤ 0 after offset correction) dropped samples SILENTLY, so a wrong offset didn't invalidate a window — it trimmed the impossible half of the shifted distribution and presented the surviving tail as a plausible small number ('e2e 0-3 ms p50 / 23 ms p95' on a session whose true hostnet was ~18 ms; also the historical '0 ms network / 0 ms e2e' readings). The refusals are now counted and drained separately from Stats — deliberately, because a fully-poisoned window drains to nil and a count inside Stats would vanish with it. The HUD shows an orange 'clock offset suspect' line and the stats line grew skew_trim=N; nonzero means disregard e2e/hostnet. - Every invalid-field fallback in the 1 Hz stats line was a bare -1: in the variadic CVarArg context the ternary does NOT unify to Double, the literal goes in as Int, and %f reads Int64(-1)'s all-ones bit pattern — which is a quiet NaN. Latent since the line existed; stage-1 (the first rung with invalid fields while frames flow) printed nan for every one. All fallbacks are now typed -1.0 / Double()-wrapped.	2026-08-13 23:15:30 +02:00
enricobuehler	1abf5c91b9	feat(apple): the tvOS present-floor verdict rested on a property readback, so open the two untested doors ... The 2026-08-13 field ladder closed 'the tvOS two-refresh present floor is immovable' on the strength of a 'link granted latency 1.00 frames' HUD line. But that line reads back preferredFrameLatency — a plain read-write float (CAMetalDisplayLink.h carries no doc contract) that echoes whatever we stored. A readback is not a grant; the measured vend lead (1.95 refresh periods) was the only truth-teller, and two levers were never actually pulled. This commit also carries the ladder instrumentation that run used: the 1 Hz stats mirror to stdout (the only log channel that exists on an Apple TV), the PresentLinkInfo HUD plumbing, the stage-4 drawable-pool clamp to 2, and the tvOS fixed-rate range pin. - PUNKTFUNK_FRAME_LATENCY makes the ask a lever (float 0...4, default 1) so an on-device ladder can prove whether the property does ANYTHING on tvOS: ask=2 growing the vend lead to ~3 means it works and the floor is ~ask+1; a lead pinned at ~2 means it is inert and the compositor regime is fixed. ask=0.5 is the in-regime win probe (the property is a float for a reason). Ask + readback go to the HUD line and the stats line (link_ask/ link_readback) so the ladder reads HUD-off over stdout. - PUNKTFUNK_PRESENTER=stage1 now resolves on Release builds (env only; the persisted picker stays DEBUG-gated — an env var is never a leftover, it takes a devicectl/Xcode launch to exist). Stage-1 presents on the hardware video plane (AVSampleBufferDisplayLayer + DisplayImmediately) instead of through the GPU compositor — the only rung that can dodge the two-refresh regime — and the field A/B silently ran stage-4 because the gate keyed on build config. The pump gains stage-1's only latency instrument: capture→enqueue into the e2e meter (offset-corrected, displayed frames only), so cross-rung runs can pin any felt difference on the present tail.	2026-08-13 23:15:30 +02:00