TimeSeries.fetch_open_data() fails when end time is at or just before a multiple of 4096

Created by: GNiklasch

TimeSeries.fetch_open_data() raises an unexpected ValueError when the requested end time is an exact multiple of 4096 seconds, thus at a GWOSC file boundary, or a fraction of a second before such a boundary.

It works as expected when the end time is a full second before a multiple of 4096, or when it is just past a multiple of 4096. (For my present purposes, both alternatives are viable workarounds - just wanted to report it for the record.)

Self-contained reproducible test case:

from gwpy.timeseries import TimeSeries

def load_strain(ifo, start, end) -> None:
    try:
        strain = TimeSeries.fetch_open_data(ifo, start, end)
    except Exception as ex:
        print(type(ex))
        print(ex.args)
        #raise # uncomment to get a backtrace on the first failure
        return
    outcome = """Loaded strain, {0} - {1}, got {2}.
    """.format(start, end, strain.duration)
    print(outcome)
    return

# test case:
ifo, t_start, t_end = 'H1', 1268121592.0, 1268121600.0

# alternate test case:
# ifo, t_start, t_end = 'V1', 1257189368.0, 1257189376.0

load_strain(ifo, t_start, t_end) # fails
load_strain(ifo, t_start, t_end - 1/64.) # fails
load_strain(ifo, t_start, t_end - 1.) # works
load_strain(ifo, t_start, t_end + 1/64.) # works

Output:

<class 'ValueError'>
[1268121592 ... 1268121600)
<class 'ValueError'>
[1268121592 ... 1268121599.984375)
Loaded strain, 1268121592.0 - 1268121599.0, got 7.0 s.
    
Loaded strain, 1268121592.0 - 1268121600.015625, got 8.015625 s.
    

This is reproducible with Python 3.9.6 + GWpy 3.0.4 (latest) + gwosc 0.7.1 (latest) (and numpy 1.24.3, astropy 5.2.2).

Uncommenting the re-raise in line 9 reveals that the ValueError comes from line 179 in gwpy/timeseries/io/losc.py:

    keep = file_segment(url) & span