Fix eight review findings on the render-bounds diagnostics

Corrects a false comment example, fixes two tests that couldn't detect their
own regressions, adds two more read-back checkpoints around Main_OnCommand so
a drift report self-locates, guards a spurious zero-vs-zero coincidence match,
and softens two sentences that overclaimed cause or defect.
This commit is contained in:
2026-08-02 14:45:42 -04:00
parent 292d14d14c
commit 0ab4673887
6 changed files with 68 additions and 19 deletions
+10 -5
View File
@@ -100,12 +100,17 @@ BoundsVerdict checkRenderedBounds(const std::string& renderedPath,
if (renderHonoredBounds(expectedFrames, actualFrames)) return v;
// Says whether this shortfall has the one shape two live short renders already
// matched to the frame, so every refusal from here on adds to (or breaks) that
// evidence instead of needing the arithmetic done by hand. A count coincidence
// only — it does not establish how the render resolved anything.
// matched to the frame: the END alone floored to the millisecond. Checked against
// the END only -- a refusal whose START is also off-grid and independently floored
// would not match this shape, and this note's silence on that refusal is this
// check not covering it, not the coincidence breaking. Excludes 0, which every
// sub-millisecond window (a legitimate day-one capture) also floors to, and which
// would otherwise match a render that produced nothing. A count coincidence only —
// it does not establish how the render resolved anything.
const long long msFlooredEnd =
msFlooredEndFrameCount(request.startSeconds, request.endSeconds, rate);
const std::string msNote =
actualFrames == msFlooredEndFrameCount(request.startSeconds,
request.endSeconds, rate)
(msFlooredEnd > 0 && actualFrames == msFlooredEnd)
? " Those are exactly the frames this window holds with its end floored to"
" the millisecond -- a match on the count, not a measured cause."
: std::string();