I have a customer that has encountered what I believe to be https://issues.apache.org/jira/browse/ARROW-9114
They are running Windows. They receive an illegal instruction exception on pyarrow.parquet.read_table. Their processor (i5-3470) does not support BMI2. The customer is using the pypi distribution (installed with pip). The fix for ARROW-9114 appears to have been a change made to snappy-feedstock which patches snappy before building it. However, I think this only applies to the conda distribution. Will there be a corresponding change made to the pypi distribution of pyarrow? Or is this something the pypi distribution should not have exhibited anyways?
