AE#418 round 2: an axis offset composes, it is not owned by a decode run

[superuser404notfound]

Follow-up to #420. rrgomes retested 6.43.0 and relation (3) failed, in a case the fix did not anticipate: the decode-run re-anchor fired for a seek that had not left the re-aimed run, so the axis went to 0 while AVPlayer was still playing content 14 s below it, and the captions ran that far ahead of the picture again.

Reproduced on the lying-Cues fixture with play --picture-probe, deterministically: a resume at 53 s (gate re-aims and opens 9 s below its boundary) followed by a seek to 80 s reads capErr=-8.983, with axisErr unchanged at -9.000 throughout. The picture never moved; only the published axis did.

What the measurement says

Round 1 keyed the offset to "the decode run AVPlayer began here" and answered that from the fetch order. Every fetch out of sequence was called a fresh run. It is not one: AVPlayer asks for a segment below its target on a seek, and it asks out of order while continuing the run it is already playing.

What the axis actually turns on is placement, and the offset composes. AVPlayer places a segment at its advertised start read through the mapping its timeline already carries, so re-placing an overlong segment adds its offset again. Measured at re-aims of 0.5, 0.875, 1, 3, 5, 7, 9 and 11 s:

case	true axis	round 1 published
resume 53 (gate 9 s below its boundary)	-9.000	-9.000
... then a seek onto an axis-true segment	-9.000	0.000
... then a seek that re-places that same segment	-18.000	0.000
... then a seek whose restart re-aims 5 s more	-14.000	-5.000
tiered fixture, -11 re-placed	-22.000	0.000

One exception, also measured: AVPlayer discards a sub-second axis at a seek and snaps back to the playlist. -0.500 and -0.875 read axisErr=0.000 after one; -1.000, -1.083, -1.292, -1.500, -3, -4, -7, -9 and -11 all survive unchanged.

The change

• The axis moves when AVPlayer places a segment, by exactly what that segment carries below its advertised start, and the seam sits at that advertised start read through the axis in effect before it landed.
• The per-index record replaces the single anchor pair: several epochs can leave an overlong segment in the cache at once. A new epoch drops the entries at and above its own index, which are rewritten axis-true.
• The gate no longer publishes on its own. It records what the segment is worth, and the placement publishes it, so an epoch AVPlayer never fetches from cannot move the clock.
• The VOD seek path publishes the sub-second snap from the landing forward.

Verification

Thirteen arms of the fixture matrix, including the four that read -8.983 before, now read capErr=+0.017, one frame at 24 fps. The control fixture (Cues that ARE its sync samples) is untouched on both arms. swift test: 2100 tests in 296 suites, 0 failures.

🤖 Generated with Claude Code

https://claude.ai/code/session_01RbWZLwBVLGM1xUVXa9NeJ1