Re: "Calculated and written ECMA-119 tree end differ" under very specific circumstances

[Thomas Schmitt]

Hi,

Valtteri Vuorikoski wrote:
> The bad news is that the full backup job with some 140k files
> fails using an invocation that worked previously (apparently because
> zisofs was only applied to files from the local filesystem).
> ...
> find /Mail -type f -pending_data -exec set_filter --zisofs --

So now /Mail comes from NFS ?
How many files are in Mail ? Most of the 140k ?


>   libisofs: WARNING : Calculated and written ECMA-119 tree end differ: 37525 
> <> 37458
> The error in the calculated/written tree difference is now the other way
> around,

Probably this is another bug. 67 blocks of mismatch is a lot.


> Switching -for_backup to "-md5 on -hardlinks on -acl on" makes the
> operation succeed.

Does it work with -xattr "on" ?

Did you try whether it really works with -for_backup if you omit the
setting of zisofs filters ?
(You may use -dev "stdio:/dev/zero" instead of -dev "/dev/sr0".
It's only of interest whether the warning message appears.)


> Since the resulting ISO now
> contains my maildir I can't send that for inspection.

How ist it structured, roughly ? (How many subdirectories with how many
files ?)
How long are the file names ?

I will make experiments with larger trees which get long xattr from
xorriso command -setfattr_r.


Have a nice day :)

Thomas