Re: [OT] How to create a 'patch' with full files?

["Chester R. Hosey" <Chester Hosey gianteagle com>]

On Mon, 2005-07-25 at 15:45 +0200, wolf2k5 wrote:
> Hi all,
> 
> [snip]
>
> Is there any way to automate the last step of this process (figuring
> out the modified files and putting them in a directory)?
> 
> That could be probably done with some piping between diff, awk and cp,
> but I don't have the necessary skills for that.

Try "cut" -- it's not quite awk, but works in a pinch and is a little
easier to learn.

> Is there any script already available?

There is now ;)

> Thanks in advance.

Nobody's touched this, so I'll give it a stab.

You basically have two tasks -- to get the list of files, and to do
something productive with it.

You can pull out the list of files using either of:

diff -qr rev0 rev1 | cut -f 4 -d ' '   #second field, space-delimited
diff -qr rev0 rev1 | awk '{print $4;}'

I'd do something like the following:

#!/bin/sh
# run from directory containing rev0 and rev1

DEST=/tmp/newtreeroot

mkdir $DEST

for FILE in `diff -qr rev0 rev1 | awk '{print $4;}'`
do
	# Get the directory portion of the filename
	DIR=`dirname $FILE`
	
	# You could just _always_ run mkdir if you
	# didn't know how to test for $DIR...
	[ -d $DIR ] || mkdir -p $DIR

	# Copy the file to the destination tree
	cp $FILE $DEST/$DIR
done

#EOF


The most important constructs are the 'for' command and use of backtics
(they're in the ~ key -- just don't press shift) to insert command
output into the command line. You could extend it to take command-line
arguments, etc., but it's a quick solution. Let me know off-list if I
can help further.

Chet