On Thu, 14 Apr 2005 12:05:24 +0200 Jean-Christophe Michel <[EMAIL PROTECTED]> wrote:
> Sorry for my english being hard to understand. > I was told that an error on a tar achive can make it impossible to > restore a file from a non corrupted part of the archive. Is it true ? * If you run tar without on-the-fly compression (-z) you can simply read all the archive up to the read error. After that it depends if tar can find the next superblock (or whatever it calls the structural metainformation). I tested this with deliberate data corruption halfway through the archive and hb quite mixed results. After that I found a tar-fixing utility that might help. * If you ran tar through gzip (-z) everything after the error is gone. tar pipes the whole data stream unblocked through gzip and gzip can not recover from a read error. * It might be slightly better to run everything through bzip2 (-j on recent gnu tars) as bzip2 automatically blocks and a read error in one block does not affect later blocks. I didn't try. * I did try afio which runs each single file through gzip and was quite robust against a few changed bytes in data area as well metadata. I don't think afio can handle incremental, but you have to feed it a list of files which could be a list of files changes since last backup time (find ... -mtime ...). HTH, K.-H. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
