caac/at7 fails badly #64
Description
Looking at the spreadsheet of audio-only tests, there is one wild card, caac/at7 as you can see here.
Most of the other caac tests pass on all 4 tests in the December test run with the updated end tolerance (like at6 shown here).
The number of failures is significant, not just 1-2 missing segments.
- FAIL: Audio segments failed at the following timestamps: 0ms 20ms 420ms 440ms 520ms 540ms 560ms 580ms 600ms 620ms 640ms 660ms 680ms 700ms 720ms 740ms 760ms 780ms 800ms 820ms ...too many failures, reporting truncated. Found 25 segments are missing. 2 start segment failed, tolerance is 3. 2 end segment failed, tolerance is 4. 21 mid segment failed, tolerance is 10.
- FAIL: Audio segments failed at the following timestamps: 0ms 20ms 140ms 160ms 180ms 200ms 220ms 240ms 260ms 280ms 300ms 320ms 340ms 360ms 380ms 400ms 420ms 440ms 460ms 480ms ...too many failures, reporting truncated. Found 40 segments are missing. 2 start segment failed, tolerance is 3. 1 end segment failed, tolerance is 4. 37 mid segment failed, tolerance is 10.
- FAIL: Audio segments failed at the following timestamps: 0ms 20ms 140ms 160ms 180ms 200ms 220ms 240ms 260ms 280ms 300ms 320ms 340ms 360ms 380ms 400ms 420ms 440ms 460ms 480ms ...too many failures, reporting truncated. Found 40 segments are missing. 2 start segment failed, tolerance is 3. 1 end segment failed, tolerance is 0. 37 mid segment failed, tolerance is 10.
This is such a wildcard that it makes me think something is wrong with the content.