[RFC] transfer files listed in a file.

[James A. Morrison]

On Thu, 2003-12-11 at 13:48, Wayne Davison wrote:
> On Wed, Dec 10, 2003 at 07:49:51PM -0500, James A. Morrison wrote:
> > e.g.
> > rsync -f rsync.files jim at server::jim/ .
> > and
> > rsync -s rsync.files jim at server::jim
> 
> In CVS, both directions use the same --files-from option, and this file
> can be read from either side of the connection.  e.g., these are all
> equivalent, with just the difference in transport (daemon vs ssh) and
> the difference of which side the rsync.files file is read:
> 
> rsync --files-from=rsync.files jim at server::jim/ .
> rsync --files-from=rsync.files server:/path/jim/ .
> rsync --files-from=server:/path/jim/rsync.files server:/path/jim/ .
> rsync --files-from=:rsync.files jim at server::jim/ .
> 
> and
> 
> rsync --files-from=rsync.files . jim at server::jim
> rsync --files-from=rsync.files . server:/path/jim
> rsync --files-from=:/path/jim/rsync.files . server:/path/jim
> rsync --files-from=server:rsync.files . jim at server::jim
> 
> The leading ':' is a short-hand for referring to the same server-name as
> the server in the transfer (since it has to be the same name, I
> recommend just prefixing the name with a colon to indicate that it is
> remote).
> 
> Some differences in our implementations:
> 
> The CVS version assumes the --relative option with --files-from so that
> the paths are preserved.  The --no-relative option was added to allow
> you to turn this off.
> 
> The CVS version tweaks -a (--archive) to not imply -r (--recursive) when
> used with --files-from.  Combined with the feature that any named
> directories are created on the destination, this allows the user to use
> the traditional "-av" options and get only the files and directories
> specified sent to the destination system and no others (specify -ar if
> you want to send named directories with recursion).
> 
> BUG:  It looks like combining --files-from with a daemon pull and a
> local file-list currently hangs during the transfer of file names (it
> succeeds in the other 3 modes of daemon transfer, and in all 4 modes
> of remote-shell transfer).  I'll fix this.

 Cool, thanks for this work, I'll be sure we get some testing of your
implementation.

> > Other than adding those options this patch also creates a new function
> > send_a_file() which makes send_file_list() a bit shorter and easier to
> > read.
> 
> I'll take a look at that.  Your patch also added an "(x of y)" display
> during the transfer which I will also consider (though I'd only want
> such extra output in the --progress mode).
> 
> ..wayne..

 Yes, it does, I just left it in there because it didn't hurt or
add anything.  If it's available as an option that would be great.

Jim