[SashaMIT]

format_time(1.2) came back as 00:00:01,199. format_time(2.3) came back as 00:00:02,299. The fractional seconds sit just under the millisecond, and int() drops it. format_time(1.5) is still 00:00:01,500. format_time(0.9999) is still 00:00:00,999.

Red: format_time(1.2) was 00:00:01,199. Green: tests.test_utils.TestFormatTime, including test_tenth_of_a_second.

[boxerab]

thanks, the bug is real. the epsilon plus the nested carry is more than this needs though. converting to whole milliseconds once handles the carry by itself:

total_milliseconds = round(s * 1000)
hours, remainder = divmod(total_milliseconds, 3_600_000)
minutes, remainder = divmod(remainder, 60_000)
seconds, milliseconds = divmod(remainder, 1000)

this rounds to the nearest millisecond instead of truncating, so test_rounding_edge should expect 00:00:01,000 for 0.9999, and test_large_value can assert 23:59:59,999 exactly.