does tar cheat when no one is looking (at its output)?

eaberry · 02-01-2022, 01:48 PM

Having burned a dvd (with ext2 filesystem) and wanting some way to check its integrity, I thought tar might be a good option. If I tar the entire disk, tar will have to read every file which will result in errors if there are any. Since I don't care about the output, tar to stdout and redirect that to /dev/null so I can watch stderr:

# tar -c /mnt/cdrom > /dev/null
tar: Removing leading `/' from member names
(finishes immediately, no error)
#
No way it could read the whole disk, much less tar all the files, in a jiffy!
Try this:

# tar -c /mnt/cdrom |tar -tv
tar: Removing leading `/' from member names
(runs for 20 minutes, with several error messages)
tar: Exiting with failure status due to previous errors
#
So it seems like tar looks at the pipe. he knows /dev/null doesn't care what you send him, and skips a lot of work. But if the output is going somewhere real, then he does his job.
Does anyone have a more rational explanation (and what is it)?
# tar --version
tar (GNU tar) 1.26

(But I seem to remember this behavior from around 2000 and before, old red-hat 7.3)

# fsck -n /dev/sr0
fsck from util-linux 2.23.2
e2fsck 1.42.9 (28-Dec-2013)
/dev/sr0 was not cleanly unmounted, check forced.
Pass 1: Checking inodes, blocks, and sizes
Inode 153 has illegal block(s). Clear? no
Illegal indirect block (3948281850) in inode 153. IGNORED.
Error while iterating over blocks in inode 153: Illegal indirect block found
/dev/sr0: ********** WARNING: Filesystem still has errors **********
e2fsck: aborted

boughtonp · 02-01-2022, 02:31 PM

Quote:

Originally Posted by https://www.gnu.org/software/tar/manual/tar.html

When the archive is being created to ‘/dev/null’, GNU tar tries to minimize input and output operations.

...

shruggy · 02-01-2022, 06:04 PM

Not only tar. Not so long ago, DistroWatch Weekly ran an article in their Tips and Tricks column describing how to find the right words with grep. You know, sort of the task danielbmartin sometimes posts here.

So, I set out to compare how different grep-like (and not-so-like) tools fare against each other doing this task.

First, I tried something like

Code:

time sh -c "grep -P '(?=.*a)(?=.*e)(?=.*i)(?=.*o).*u' /usr/share/dict/words>/dev/null"

But quickly noticed that some tools did it suspiciously fast, and so I ended up piping the output to wc to get reasonable results.

Code:

time sh -c "grep -P '(?=.*a)(?=.*e)(?=.*i)(?=.*o).*u' /usr/share/dict/words|wc -l"

I guess those grep-like tools stop the search after the first successful match when redirected to /dev/null.

I uploaded the full script here if anybody is interested.