Skipping this for now, filtering the repair output properly is going to be a lot of work, and I'd rather do a generic unlinked inode list recovery test instead of tying it to O_TMPFILE. With the first patch already reviewed and the second updated those should be ready to apply now.