HP-UX

Thanks

Code:

(
cd dir1
find * -type d | cpio ( create_archive_options ) 
)| (
 cd dir2
 cpio ( unpack_archive_options )
 )

OR check the cpio pass option on the man page. Usually I feed cpio flat files and ask it to make dirs as necessary, but I know it can move empty dirs. "cp -rp" moves everything. In fact, I think scp, scp2, rcp, cp, *cp -rp all use cpio under the covers. The -p and cpio properly optioned preserves group, user, permissions, dates. I do not know the cpio options offhand because I immediately put them in scripts and forget them, and do not have UNIX here. Wait, I have cygwin; the first script takes in file names and makes a bzip2 compressed archive, and the second does somewhat the reverse:

Code:

$ pg ../*/mysrc_sol/b*cpio|cat
::::::::::::::
../DavidGPickett/mysrc_sol/bcpio
::::::::::::::
#!/usr/bin/ksh

cpio -oaH crc $1 |bzip2 $2

::::::::::::::
../DavidGPickett/mysrc_sol/buncpio
::::::::::::::
#!/usr/bin/ksh

bunzip2 <$2 |cpio -idmH crc $1
$

Hi,

As posted in another thread, check this link: https://www.unix.com/302460613-post23.html

Regards!

The ususal method is to use for example "find . -xdev -type d -print" piped to for example "cpio -pdmv /target_dir" . This does not require an intermediate archive (it writes directly to the target directory) and is just about the fastest way to copy a directory tree complete with permissions.
There are examples in both "man cpio" and "man find" . I haven't posted a formal example because there is some variation and we don't know what Operating System you have.




Be very cautious of the DGPickett post. You MUST use relative pathnames for this sort of operation.

Hi Felipe and all




Thanks

The script in filipe.vinturin post does not contain any of the design or coding errors which we see in your post.

I don't endorse your design based on the output from "find" piped to "ls -ld" because you lose the exact directory name. It becomes very awkward to deal with directory names containing space characters from that point on.

Using filipe.vinturin construct with "find" piped to "while read" is much better.

I was tempted to debug your script but I have just noticed that $destination is a directory under $source which makes it just too awkward because you generate an infinite loop by picking up "/tmp/dest" in the initial find.

There are also multiple design issues in the script which means that it does not do what you expect at all ... and if run after getting the syntax errors out could harm your system.
You certainly don't want to issue "chmod -R" in this circumstance - just use "chmod" for individual directories.


It is still much easier with "find" piped to "cpio" but that is also a non-starter if destination directory is under source directory.

Yes, sharp knives can cut the cook, too. It is very common to move subtrees around relatively pathed in archives, like for instance every bit of open source app source. You can make a file from the first cpio and do a cpio table of contents to see if you like the archive.

You are afraid you will hit a mount (-xdev)?

I mentioned the cpio pass option to do it in one step.

I use the 2 cpio on a pipe because it does not slow things down or take any more space than -p pass, but you can pipe it off-host using ssh2/ssh/rsh/remst/rexec and not leave any temp file clutter around, and compress it, and put the intermediate product in a file to reuse it. So, cpio pass mode does not do much of substance but expand the cpio man page. But I was not on that committee!