On Wed, Wed, 22 Jun 2022 14:13:47 -0400, Taylor Blay wrote: > It shouldn't in practice. We'll recognize that the .rev file is garbage > if it's checksum doesn't match the MIDX's, and we squashed the bug where > changing the object *order* (but not the objects themselves) doesn't > change the MIDX checksum (it does now). > > We won't write over an existing .rev file with the right name, but we'll > always prefer to read the .rev data from the MIDX itself, if it exists. Yes, this is what I want to say. On the patch itself, I think it may worth to remove the tmp rev file at earlier time because there may reduce doubt about whether the file remaining after the interruption will have side effects (but we know it won't for now , because regenerating midx and rev will ensure an strict match based on the checksum). Thanks.
