> How so? I did not see this yet, at least. I need to stop making so vague commit messages... I was thinking of the rebuild-indices pass, where ReadAttribute() is called on every inode for every index we're rebuilding, so there are lots of B_ENTRY_NOT_FOUND's which aren't really errors. Maybe it should be worked around some other way (or left alone). > Or is it a temporary problem because of the not yet supported move of > attribute directories? Moving attribute directories is supported now. :) > But this just means that the block allocator failed to allocate that > many blocks in a single block_run - you are actually supposed to call it > again with the rest of what you need. We need exactly those blocks at the end of the old log area, no others will do. Though it does assume that if the block allocator can fulfill our maximum request, it will, which isn't really guaranteed. I can rewrite it to repeatedly try to allocate from the area until we have all the blocks or allocation fails, if you think it's better. //Andreas