解压缩之前，tar归档文件中列出的备份文件可简化回滚

Question

Before unpacking a compressed tar file, what is the best way to backup the files that will be overwritten?

For example, given the input archive files.tar.gz containing:

path/to/fileA.txt
path/to/fileB.txt
path/to/fileC.txt

and currently on the filesystem we have:

path/to/fileA.txt
path/to/fileC.txt

Since fileA.txt and fileC.txt will be overwritten, we would like to have files-backup.tar.gz which contains only those two original files.

Currently I'm doing this manually to have a crude rollback mechanism if things don't work out. With more than a handful of files in several locations this can become a pain.

Would love to script it up or learn a new hidden parameter for the tar command.

UPDATE : The best so far I've come up with is:

tar ztf files.tar.gz | xargs tar zcvf /tmp/file-backup.tar.gz

Answer 1

What you had is a good start, but one thing to be aware of is that xargs will process stdin lines in groups. (That's its whole purpose in life afterall: to take an unlimited number of input lines and execute a command multiple times, each time with a subset of input lines as command line arguments.) When files.tar.gz has a huge number of files, xargs will do tar zcvf /tmp/file-backup.tar.gz on a subset of them at a time! The resulting /tmp/file-backup.tar.gz will contain only the last subset of files!

I will assume that since you specified paths as path/.../... that all paths will be in relative.

Here's a better solution (see "notes" below on use of tar ...|while read... ):

(tar ztf files.tar.gz|while read f; do if [ -f "$f" -o -h "$f" ]; then echo "$f" ; fi; done)|tar Tcfz - /tmp/`date '+%Y%m%d_%H%M%S'`.file-backup.tar.gz

The tar command produces a file named /tmp/yyyymmdd_hhmmss.file-backup.tar.gz containing the list of files supplied on its stdin, which is the list of files and symbolic links in files.tar.gz .

Finally, you might want to consider a couple of alternatives that might make hosuekeeping easier:

1. Only save off files that have actually changed, instead of saving off every file.
2. Or, instead of saving off files into a tar/gzip file, rename files in place; this way, you'd see right away which files have newer version(s) without needing to go look into several different /tmp/yyyymmdd_hhmmss.file-backup.tar.gz files first:

(TAG=`date '+%Y%m%d_%H%M%S'` ; tar ztf files.tar.gz|while read f; do if [ -f "$f" -o -h "$f" ]; then mv "$f" "$f.$TAG" ; fi; done)

NOTE: the tar ztf ...|while read f ensures that filenames containing whitespace(s) won't cause a problem. The simpler

for f in `tar ztf files.tar.gz`; do...

won't work when there are files in files.tar.gz with whitespace(s) in their name.