Re: [Bug-tar] Incremental extract improvement

[Helmut Waitzmann]

Ian Turner <address@hidden> writes:

>On Wednesday 10 May 2006 21:18, Sergey Poznyakoff wrote:

>> > IMNSHO it would be better if the incremental
>> > restore works analogously to the non-incremental restore: That is, to not
>> > delete stuff not mentioned in the archive.
>>
>> Than it would not be incremental restore. It's main purpose is to
>> restore the filesystem in exactly the same state it had when dumped.

Yes.  Sergey is right:  If you don't want to restore the filesystem in
exactly the same state it had when dumped, then simply don't use the
"--listed-incremental" parameter when extracting an incremental archive.

>Eh, a more common understanding of incremental extraction is like patch: 
>"Apply the difference between this archive and this other archive as stored 
>in this archive here." 

"... stored in this archive here."  That is, files, which are stored in
the incremental archive?

>I don't think that says anything about what should happen to other,
>unrelated files. GNU tar is the only tool I'm aware of that goes
>deleting unrelated files.

What are "unrelated files"?  Files, which are not stored in the
incremental archive?

>Look at it this way: If the user really just wants to start from
>scratch, it is easy to blow away an existing directory, then do all the
>appropriate restores.

No.  It's not easy.  Imagine, you want to restore the filesystem to a
state which has been archived by a sequence of two backups: a full
backup, "full.tar", and a following incremental backup,
"incremental.tar".

So, how would you do that, if tar, when extracting incremental, wouldn't
delete files?

First, delete all files:

$ rm -rf -- filesystem

Then, extract the full backup:

$ tar -x -f full.tar --listed-incremental=/dev/null

Now, the filesystem reflects the state it had, when the full dump was
made.

Now, imagine, that in the filesystem, after the full dump, but before the
incremental dump had been made, the file "filesystem/new" had been
created and the file "filesystem/obsolete" had been deleted, whereas the
file "filesystem/unchanged" had been retained.

Then, extract the incremental backup:

$ tar -x -f incremental.tar --listed-incremental=/dev/null

If tar didn't remove "filesystem/obsolete", what command would you start
to remove "filesystem/obsolete" while retaining "filesystem/unchanged",
if all information you had about the saved filesystem states would be the
tape archives mentioned above?

>Look at it this way: If the user really just wants to start from
>scratch, it is easy to blow away an existing directory, then do all the
>appropriate restores.

Please, tell me the commands to execute in order to blow away all
existing files that should be deleted according to the state of the
filesystem recorded in the "incremental.tar" archive, while retaining all
files which were dumped into the "full.tar" archive and were not dumped
again into the "incremental.tar" archive because they hadn't changed.

On the other hand, if you don't want tar to remove "filesystem/obsolete"
when extracting "incremental.tar", all you have to do is not to supply
the "--listed-incremental" parameter:

$ tar -x -f incremental.tar

>But if the user does *not* want this, it requires many more operations
>to make the restore to some other place and then copy it into the
>directory structure s/he is interested in.

Could you please explain, why

$ tar -x -f incremental.tar --no-overwrite-dir

wouldn't do that task?
-- 
Wenn Sie mir E-Mail schreiben, stellen  | When writing me e-mail, please
Sie bitte vor meine E-Mail-Adresse      | precede my e-mail address with
meinen Vor- und Nachnamen, etwa so:     | my full name, like
Helmut Waitzmann <address@hidden>, (Helmut Waitzmann) address@hidden