Incorrect RC3339 encoding
fgm opened this issue · comments
I seem to have stumbled upon an edge case with this timestamp: 2023-12-14T21:59:43.10-05:00
It is correctly deserialized but, when reserialized, becomes 2023-12-14T21:59:43.1-05:00
AFAICS, the current tests do not check what happens when the seconds are a multiple of 10.
Actually this does not appear to be incorrect. Stdlib does the same, even documenting that RFC3339Nano drops zeroes so causing data not to sort. So, it's not actually a bug, just a known annoying feature.