Stewart Stremler wrote:
begin quoting Ralph Shumaker as of Wed, Sep 21, 2005 at 07:48:37PM -0700:
As suggested by some on the list, I tried cpio for copying whole
partitions (not to mention using dd) and parts of partitions. I did not
find it any better than cp with the appropriate switches, but in fact worse.
Interesting!
So far, I like cp the best. But there is primarily one thing that I do
*not* like. It doesn't keep the timestamps of links. I tried touching
What options are you passing to cp?
I don't remember for sure. I was doing all this from Knoppix, so I
don't have the command history. I also don't have around me the paper
on which I wrote the switches I chose to use. Oops. I spoke to soon.
The paper I most recently used (with knoppix v3.4) has:
cd [source directory]
cp -rvpd ./* [target directory]
One other thing that was a bit odd was that this command copies just
fine any dot files below the starting directory but ignores the dot
files (in my case they were dot directories and their contents) that are
actually in the starting directory.
a few of the destination links after the copy, but instead of the links'
timestamps changing, the timestamps were changed on the files to which
they link.
The simplest way that I can think of to verify all files copied over is
to "ll -aR > ll-aR.hde2" and "ll -aR > ll-aR.hdg2" and then "diff
ll-aR.hde2 ll-aR.hdg2 > ll-aR.hdg2-hda2.diff", but every link shows up
Er, what's the ll business? That's a common alias for "ls -l", but it's
not a standard command on any of my Linux boxen.
Sorry. I'm actually aware that most on the list use "ls -l" instead of
"ll" because that's the way they say it in their descriptions of how to
do things that include that. I just forgot that the "ll" alias,
although apparently standard in RH, is not universal. "Oops." is about
all I can say I reckon.
there because of the timestamp. And there are a whole *lotta* links. I
can pipe through a 'grep -v " --> "' and get a much smaller output. And
although this assures me that the objects of the links copied fine, it
eliminates the certainty that all *links* got copied. Is there a way to
make cp keep the timestamps of the copied links? As far as I can tell,
any attempt to modify the timestamp of the link just gets applied to its
target.
What if the target isn't there?
mv $target $target.__save__
touch $link blah blah blah
mv $target.__save__ $target
Perhaps I'm just really tired, but I don't get what you're trying to say
here. I get lines 1 and 3, but the purpose of line 2 eludes me. Also
eluding me is why you're using "mv". I don't see how that relates to my
question.
The other thing that is disconcerting about the cp, ll, diff I do is
that directory sizes don't necessarily stay the same. If the size
differs at all, usually the copy is taking less space. (I don't
remember seeing any that ended taking more.) But this occurrence is
much, much less frequent than the link dates changing.
Yes, that's expected.
I don't understand why, although I think I have an idea. But my idea is
immaterial. When it comes to the actual size of the directory copy, I
would prefer that the size matches, but am satisfied if all of that
directory's contents match.
Is there a way to get a reliable listing of the source files, do a byte
for byte comparison, and report only the ones that are different?
Isn't that what the cmp command does?
-Stewart "Never thought about looking at a link's timestamp before" Stremler
Yeah, well, I just feel better when the copy seems to be a perfect
copy. I would actually prefer to *know* that the copy is just as
reliable as the original. And if the dates or sizes are different, it
just *feels* like they are not reliably the same.
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
