[Python][CI] Hypothesis tests are failing (test-conda-python-3.11-hypothesis crossbow build) #40379
Description
See eg https://github.com/ursacomputing/crossbow/actions/runs/8104553122/job/22151381641
They started failing between 2024-02-29 and 2024-02-28. First failure https://github.com/ursacomputing/crossbow/actions/runs/8089011272/job/22104197020 shows an error about
________________________ test_array_to_pylist_roundtrip ________________________
@h.given(past.all_arrays)
> def test_array_to_pylist_roundtrip(arr):
opt/conda/envs/arrow/lib/python3.11/site-packages/pyarrow/tests/test_convert_builtin.py:2209:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
opt/conda/envs/arrow/lib/python3.11/site-packages/pyarrow/tests/strategies.py:316: in arrays
value = st.binary(min_size=ty.byte_width, max_size=ty.byte_width)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
> ???
E ValueError: Less than one byte
E while generating 'arr' from arrays(type=one_of(one_of(one_of(one_of(one_of(one_of(one_of(just(DataType(null)), just(DataType(bool)), one_of(one_of(sampled_from([DataType(int8), DataType(int16), DataType(int32), DataType(int64)]), sampled_from([DataType(uint8), DataType(uint16), DataType(uint32), DataType(uint64)])), sampled_from([DataType(halffloat), DataType(float), DataType(double)]), builds(<cyfunction decimal128 at 0x7f8d784937a0>, precision=integers(min_value=1, max_value=38), scale=integers(min_value=1, max_value=38)), builds(<cyfunction decimal256 at 0x7f8d78493890>, precision=integers(min_value=1, max_value=76), scale=integers(min_value=1, max_value=76))), one_of(sampled_from([DataType(date32[day]), DataType(date64[ms])]), sampled_from([Time32Type(time32[s]), Time32Type(time32[ms]), Time64Type(time64[us]), Time64Type(time64[ns])]), builds(<cyfunction timestamp at 0x7f8d78492e40>, tz=one_of(none(), timezones(), timezones()), unit=sampled_from(['s', 'ms', 'us', 'ns'])), builds(<cyfunction duration at 0x7f8d78493110>, sampled_from(['s', 'ms', 'us', 'ns'])), just(DataType(month_day_nano_interval))), one_of(just(DataType(binary)), just(DataType(string)), just(DataType(large_binary)), just(DataType(large_string)), builds(<cyfunction binary at 0x7f8d78493b60>, integers(min_value=0, max_value=16)))), one_of(one_of(builds(<cyfunction list_ at 0x7f8d78498140>, one_of(just(DataType(null)), just(DataType(bool)), one_of(one_of(sampled_from([DataType(int8), DataType(int16), DataType(int32), DataType(int64)]), sampled_from([DataType(uint8), DataType(uint16), DataType(uint32), DataType(uint64)])), sampled_from([DataType(halffloat), DataType(float), DataType(double)]), builds(<cyfunction decimal128 at 0x7f8d784937a0>, precision=integers(min_value=1, max_value=38), scale=integers(min_value=1, max_value=38)), builds(<cyfunction decimal256 at 0x7f8d78493890>, precision=integers(min_value=1, max_value=76), scale=integers(min_value=1, max_value=76))), one_of(sampled_from([DataType(date32[day]), DataType(date64[ms])]), sampled_from([Time32Type(time32[s]), Time32Type(time32[ms]), Time64Type(time64[us]), Time64Type(time64[ns])]), builds(<cyfunction timestamp at 0x7f8d78492e40>, tz=one_of(none(), timezones(), timezones()), unit=sampled_from(['s', 'ms', 'us', 'ns'])), builds(<cyfunction duration at 0x7f8d78493110>, sampled_from(['s', 'ms', 'us', 'ns'])), just(DataType(month_day_nano_interval))), one_of(just(DataType(binary)), just(DataType(string)), just(DataType(large_binary)), just(DataType(large_string)), builds(<cyfunction binary at 0x7f8d78493b60>, integers(min_value=0, max_value=16))))), builds(<cyfunction large_list at 0x7f8d78498230>, one_of(just(DataType(null)), just(DataType(bool)), one_of(one_of(sampled_from([DataType(int8), DataType(int16), DataType(int32), DataType(int64)]), sampled_from([DataType(uint8), DataType(uint16), DataType(uint32), DataType(uint64)])), sampled_from([DataType(halffloat), DataType(float), DataType(double)]), builds(<cyfunction decimal128 at 0x7f8d784937a0>, precision=integers(min_value=1, max_value=38), scale=integers(min_value=1, max_value=38)), builds(<cyfunction decimal256 at 0x7f8d78493890>, precision=integers(min_value=1, max_value=76), scale=integers(min_value=1, max_value=76))), one_of(sampled_from([DataType(date32[day]), DataType(date64[ms])]), sampled_from([Time32Type(time32[s]), Time32Type(time32[ms]), Time64Type(time64[us]), Time64Type(time64[ns])]), builds(<cyfunction timestamp at 0x7f8d78492e40>, tz=one_of(none(), timezones(), timezones()), unit=sampled_from(['s', 'ms', 'us', 'ns'])), builds(<cyfunction duration at 0x7f8d78493110>, sampled_from(['s', 'ms', 'us', 'ns'])), just(DataType(month_day_nano_interval))), one_of(just(DataType(binary)), just(DataType(string)), just(DataType(large_binary)), just(DataType(large_string)), builds(<cyfunction binary at 0x7f8d78493b60>, integers(min_value=0, max_value=16)))))), builds(<cyfunction list_ at 0x7f8d78498140>, one_of(just(DataType(null)), just(DataType(bool)), one_of(one_of(sampled_from([DataType(int8), DataType(int16), DataType(int32), DataType(int64)]), sampled_from([DataType(uint8), DataType(uint16), DataType(uint32), DataType(uint64)])), sampled_from([DataType(halffloat), DataType(float), DataType(double)]), builds(<cyfunction decimal128 at 0x7f8d784937a0>, precision=integers(min_value=1, max_value=38), scale=integers(min_value=1, max_value=38)), builds(<cyfunction decimal256 at 0x7f8d78493890>, precision=integers(min_value=1, max_value=76), scale=integers(min_value=1, max_value=76))), one_of(sampled_from([DataType(date32[day]), DataType(date64[ms])]), sampled_from([Time32Type(time32[s]), Time32Type(time32[ms]), Time64Type(time64[us]), Time64Type(time64[ns])]), builds(<cyfunction timestamp at 0x7f8d78492e40>, tz=one_of(none(), timezones(), timezones()), unit=sampled_from(['s', 'ms', 'us', 'ns'])), builds(<cyfunction duration at 0x7f8d78493110>, sampled_from(['s', 'ms', 'us', 'ns'])), just(DataType(month_day_nano_interval))), one_of(just(DataType(binary)), just(DataType(string)), just(DataType(large_binary)), just(DataType(large_string)), builds(<cyfunction binary at 0x7f8d78493b60>, integers(min_value=0, max_value=16)))), integers(min_value=0, max_value=16)))), struct_types()), builds(<cyfunction dictionary at 0x7f8d784985f0>, sampled_from([DataType(int8), DataType(int16), DataType(int32), DataType(int64)]), one_of(just(DataType(bool)), one_of(sampled_from([DataType(int8), DataType(int16), DataType(int32), DataType(int64)]), sampled_from([DataType(uint8), DataType(uint16), DataType(uint32), DataType(uint64)])), sampled_from([DataType(float), DataType(double)]), just(DataType(binary)), just(DataType(string)), builds(<cyfunction binary at 0x7f8d78493b60>, integers(min_value=0, max_value=16))))), map_types()), one_of(one_of(builds(<cyfunction list_ at 0x7f8d78498140>, (deferred@140245131593456)), builds(<cyfunction large_list at 0x7f8d78498230>, (deferred@140245131593456))), builds(<cyfunction list_ at 0x7f8d78498140>, (deferred@140245131593456), integers(min_value=0, max_value=16)))), struct_types(item_strategy=(deferred@140245131593456))))
pyarrow/types.pxi:279: ValueError
so that might be related to #39592 ( although I don't directly see how)
Later (the last two days), more failures were appearing, most likely due to #40160, which edited a hypothesis strategy, but we forgot to trigger the hypothesis tests in that PR.