Use assert_approx_equals in MediaSource tests
#17266
Conversation
The `threeDecimalPlaces` helper using `number.toFixed(3)` allowed for a maximum difference of 0.001 for a pair like 0.0005 / 0.00149999, but had no lower limit on the difference, a pair like 0.0004999 / 0.0005 would fail. This may be the cause of a Safari failure `!EQ(6.548, 6.549)`: https://wpt.fyi/results/media-source/mediasource-endofstream.html?run_id=255090001
|
One of the tests I touched is flaky in Chrome:
But that isn't plausibly caused by the changes made here. From the wpt_report.json.gz I see that when the test passed it took about 500 ms, so it's not that it's too slow. Perhaps there's a race condition in the test where it runs faster the 2nd and following times, making it more prone to time out then. |
|
Reloading it over and over locally I can't get it to time out. |
|
The wpt.fyi checks show that this fixes a failing subtest in both Chrome and Safari which previously failed as "assert_equals: SegmentInfo duration should still roughly match mediaSource duration expected 6.548 but got 6.549": Apart from that results are unaffected. @wolenetz would you mind reviewing this now? |
| 0.001, | ||
| "sourceBuffer.buffered range begins where expected before EOS"); | ||
| assert_less_than_equal(sourceBuffer.buffered.end(0), | ||
| expectedBufferedRangeMaxEndTimeBeforeEOS + 0.001, |
There was a problem hiding this comment.
Note the + 0.001 to keep the same behavior as before. This is the only change that isn't a straight replacement with assert_approx_equals.
The `threeDecimalPlaces` helper using `number.toFixed(3)` allowed for a maximum difference of 0.001 for a pair like 0.0005 / 0.00149999, but had no lower limit on the difference, a pair like 0.0004999 / 0.0005 would fail. This may be the cause of a Safari failure `!EQ(6.548, 6.549)`: https://wpt.fyi/results/media-source/mediasource-endofstream.html?run_id=255090001
The
threeDecimalPlaceshelper usingnumber.toFixed(3)allowed fora maximum difference of 0.001 for a pair like 0.0005 / 0.00149999, but
had no lower limit on the difference, a pair like 0.0004999 / 0.0005
would fail.
This may be the cause of a Safari failure
!EQ(6.548, 6.549):https://wpt.fyi/results/media-source/mediasource-endofstream.html?run_id=255090001