I use Linux. There is a pesky ^M (Windows cariage return) somewhere hidden in thousands of configuration files, and I have to find it, because it makes the server fail.

How do I find ^M among a directories hierarchy full of configuration files?

I think I can not enter ^M on the bash command line. But I have it in a text file that I called m.txt

grep -r $'\r' *

Use -r for recursive search and $'' for c-style escape in Bash.

More, if you are sure it's text file, then it should be safe to run

tr -d $'\r' < filename

to remove all \r in a file.

If use GNU sed, -i can perform in-place edit, so you won't need to write back:

sed $'s/\r//' -i filename
  • 10
    @Nicolas: You can enter a ^M at the command line by pressing ^V^M, but it's better to use $'\r'. – Dennis Williamson Oct 1 '10 at 5:54
  • Great, it works! Thanks for the ^V^M trick too :-) – Nicolas Raoul Oct 1 '10 at 5:56
  • 5
    Under Cygwin, -U is needed to make this work. And -n will tell you the line number: grep -r -U -n -e $'\r' – Rainer Blome Jan 3 '13 at 16:16
  • 4
    Add an -l to the grep command to just view the filenames. Else you might be bombarded with matching lines. – Brendan Byrd Mar 25 '14 at 20:28
  • 1
    @uprego not sure if you understand them now, but fyi and other's, search $' read the first hit in manpage bash(1), basically, you can see it as if you were writing C literal string. As for the command < filename, the use of < or > is called redirection, this is first time I have seen anyone called it greater expression. Search REDIRECTION in bash(1). – livibetter Sep 23 '15 at 2:03

When I tried, I could tell it was sort-of working, but the lines were printing blank. Add in the option:


If you get this issue, I think it's the escape characters for color highlighting interfering with the \r character.


If your server does not have a bash shell, an alternative is to use the -f option on grep, in combination with a prepared file containing \r.

To create the file:

$ echo -ne '\r' > /tmp/cr                    --or--                   $ printf '\r' > /tmp/cr

$ od -c /tmp/cr
0000000  \r

To actually do the search

$ grep -f /tmp/cr *.html *.php *.asp *.whatever

or you can be a little lazy and just type *,

$ grep -f /tmp/cr *

The -f filename option on grep is used to specify a file that contains patterns to match, one per line. In this case there's only one pattern.


To use grep on end-of-line characters, I guess you have to tell grep the file is binary.

-l (letter L) is for printing only the filename

-P is for perl regexp (so \x0d is tranformed to \r or ^M)

grep -l --binary -P '\x0d' *

If I understand your question correctly, what you really want is to normalize all line-endings to the Unix LF (\x0a) standard. That is not the same as just blindly removing CRs (\xod).

If you happen to have some Mac files around which use just CR for newlines, you will destroy those files. (Yes, Macs are supposed to use LF since almost 20 years, but there are still (in 2019) many Mac apps which use just CR).

You could use Perl's \R linebreak escape to replace any sort of newline with \n.

perl -i.bak -pe 's/\R/\n/g' $your_file

This would replace in-place any sort of linebreak with \n in $your_file, keeping a backup of the original file in ${your_file}.bak.


If you are on a Mac and use homebrew, you can do:

brew install tofrodos
fromdos file.txt

to remove all the Windows carriage returns from file.txt

To switch back to Windows carriage returns,

todos file.txt
  • to search in a folder and clean all files coming from dos, run this command : find . -type f -name "*.java" | xargs fromdos – Taiko May 19 '14 at 3:38

In regular expression style, various newlines:

Windows (CR LF)

Unix (LF)

Since the \r\n sequence is fairly unique, I think you should be able to search for it that way?

To make things worse Macs used to have just '\r' in place of newline. I cannot verify this, but I don't think MacOSX generations does that any more.

Older Macs (CR)

  • In the directory that contains m.txt, grep "\r\n" * gives no result. No result either for egrep -e "\r\n" * nor grep -E "\r\n" * – Nicolas Raoul Oct 1 '10 at 5:54
  • @nicolas ah, I misunderstood.. you meant CR only \r my bad. A full windows newline is indeed \r\n or CRLF – Jeff Atwood Oct 1 '10 at 5:58

Following up on previous answers, the 'tr' method is good:

533$ if [[ -n "tr -cd "\r" <~/.bashrc" ]]; then echo "DOS"; else echo "UNIX"; fi


534$ if [[ -n "tr -cd "\r" <dosfile.txt" ]]; then echo "DOS"; else echo "UNIX"; fi


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.