25

I'm copying lots of files that have changed from one server to another using rsync. I know I can use the -n option to do a dry run, so I can see what files have been changed. However is it possible to get rsync to print a diff of the file contents that's changed? I'd like to see what's happening before doing a copy? Something I can save to a file and the apply with diff(1) later?

Amandasaurus
  • 30,211
  • 62
  • 184
  • 246

7 Answers7

9

There might be a better way, but this might work, albeit not that efficiently:

 rsync -vrn / dest:/ > ~/file_list

Then edit test to remove the stats, then:

while read file; do
    diff $file <(ssh dest "cat $file")
done < ~/edited_file_list

Another Option:
You might also consider mounting the file system with something like sshfs/fuse, and then just using diff.

Kyle Brandt
  • 82,107
  • 71
  • 302
  • 444
  • Note: I didn't test those commands ;-) – Kyle Brandt Sep 04 '09 at 11:31
  • Good start, but there's loads of extra output from rsync, such as the statistics, and "sending incremental file list", etc – Amandasaurus Sep 04 '09 at 11:43
  • You could use --out-format="%f" – Kyle Brandt Sep 04 '09 at 11:49
  • If you use the out-format, drop the v, and grep -v 'skipping non-regular file' ... That should get it pretty clean – Kyle Brandt Sep 04 '09 at 11:51
  • Just checking if by chance there is a new / better method to `rsync --diff` two years later... – Déjà vu Jan 13 '13 at 12:30
  • The `while` only read the first line of the file. Solved by writing the `diff` commands to a temporary script: `while read file; do echo "diff $file <(ssh dest \"cat $file\")" >> /tmp/diffCheck; done < ~/edited_file_list; bash /tmp/diffCheck; rm /tmp/diffCheck`. – Roger Dueck Mar 23 '17 at 19:57
  • Given that `rsync` is copying from local to remote, it might be intuitive to reverse the `diff` arguments, so that the diff produced can be applied to the remote file, which is what `rsync` is doing. As it stands, you'll produce a reversed patch file. – Jim L. Sep 13 '19 at 18:15
6

For create patch:

rsync -arv --only-write-batch=patch new/ old/

For apply it:

rsync -arv --read-batch=patch dir/

or use auto-generated script:

./patch.sh

Sources:

Hongda Zhang
  • 3
  • 2
FarK
  • 159
  • 1
  • 3
2

rsync can't do this natively, but if there's a possibility of using unison you can produce diff style format from that.

Philip Reynolds
  • 9,751
  • 1
  • 32
  • 33
1

To expand on Kyle's answer, this automates the process. Note that it is totally untested, probably pretty fragile, and may delete your computer and kill your dog.

#!/bin/bash

REMOTE=${1?Missing Remote Path}
LOCAL=${2?Missing Local Path}

# Trim trailing slash since we'll be adding it as a separator later
REMOTE=${REMOTE%/}
LOCAL=${LOCAL%/}

#Break it down
RHOST=${REMOTE%:*}
RPATH=${REMOTE#*:}

while read FILE; do
    diff -u ${LOCAL}/${FILE} <(ssh $RHOST "cat ${RPATH}/${FILE}")
done < <(rsync -vrn $REMOTE/ $LOCAL/ | sed '1d;/^$/q')
tylerl
  • 14,885
  • 7
  • 49
  • 71
1

It's not possible natively because rsync only cares about binary differences between files.

You might be able to script it, using rsync's output. But it would be hackish.

I do believe it's natively possible with Unison though.

Dan Carley
  • 25,189
  • 5
  • 52
  • 70
1

Why not just use something like diff (for text files) or xdelta (for binary files) to generate the diffs? Why do you need to specifically get something out of rsync?

womble
  • 95,029
  • 29
  • 173
  • 228
0

The rsync algorithm works by comparing binary chunks of the file. Such binary diff is not meant to be printable. There is a command called rdiff that uses the rsync algorithm to generate a binary diff, but I don't think it'd be useful for what you describe, it is commonly used to implement incremental backups.

sagi
  • 707
  • 3
  • 9
  • 19