Re: Extracting a single directory from a partial incremental dump?

"Felix E. Klee" <[email protected]> wrote:

> At Tue, 18 Oct 2005 15:07:09 +0200,
> Joerg Schilling wrote:
> > Note that you need to start with creating a _new_ level 0 backup as
> > the bug did cause the inode list for a directory to get corrupted on
> > the aresultant archive.
> > 
> > The probability for this to happen was approx.: 1 : 40,000,000 this is
> > why it took so long to be discovered.
>
> Unfortunately, either the bug persists in version 1.5a67 or we're
> subject to a misunderstanding.  When I execute the script below on my
> system (Linux kernel 2.4.29, format 3.6 Reiser FS) I get output such as
> the following, i.e. star panics:
>
>   > ./star-test 
>   Creating archive...
>   Type of this level 0  dump: full
>   Date of this level 0  dump: Fri Oct 28 18:27:02 2005
>   Date of last level 0  dump: the epoch
>   star: 1 blocks + 0 bytes (total of 10240 bytes = 10.00k).
>   Dump record  level 0  dump: Fri Oct 28 18:27:02 2005 written
>   Done.
>   
>   Extracting archive...
>   Validating this dump against restored filesystem...
>   Dump level 0 on empty filesystem, starting restore.
>   star: pfind_node(): Not a directory 'B' for 'B/a' flags 0 i_dir 0.
>   star: padd_node(): caller 'sym_addrec'.
>   star: padd_node(): Not a directory 'B' for 'B/a' flags 0 i_dir 0.
>   star: padd_node(B/a, 6, 0, 0) = NULL
>   star: WARNING: No old inode number for 
>   star: WARNING: No new inode number for 
>   star: WARNING: No new inode number for /B
>   star: Panic: cannot add node 'B/a'.

Did you really follow my advise and use a _new_ archive created with
a _recent_ star, when dong untar tests?

On the other side, there are rumors that Reiserfs is not POSIX compilant.
In this case, Reiserfs may be the reason for the problems.

Another problem may be caused by a bug on the compiler.... I did run all
my tests with star binaries compiled by the Sun Studio 10 compiler.

Anyway, if you believe that you have a small test case that creates a small
tar archive, I would like to look at that archive. I see no other way to
proceed, unless you like to debug the problem yourself....

Jörg

-- 
 EMail:[email protected] (home) Jörg Schilling D-13353 Berlin
       [email protected]		(uni)  
       [email protected]	(work) Blog: http://schily.blogspot.com/
 URL:  http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily