GregBowyer opened a new pull request #8948: URL: https://github.com/apache/arrow/pull/8948
@alamb @andygrove No idea if this is the root cause, but I figured that if its reading an incorrect number as a boolean its a messy transmute or similar. As such I ran miri over the specific failing test and got two hits, one on the hash function and one on the bit packing. The first commit states that the bit packing is unaligned which I think is ok to take (although it might be worth revisiting the bit-packing code). The second replaces the hash function which is stated as being unaligned. I dont know if wyhash is the perfect choice (xxHash, t1ha or others might be equally good choices) but it does not suffer from irritating miri. Thoughts? Ideas? ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
