2015-11-22 21:24:05 -0800, Shivanshu Goyal:
[...]
> I think I found a bug which did not exist in version 2.14, but does seem to
> exist in versions 2.16 and 2.22. I have not tested any other versions.
>
> Say there is a file with the following contents:
>
> shivanshu@thetis:tmp$ cat temp | xxd
> 0000000: 68e2 8093 680a h...h.
>
> The following is the grep 2.14 command and output:
>
> shivanshu@thetis:tmp$ cat temp | grep -P '\xe2\x80\x93'
> h–h
>
> The following is the grep 2.16/2.22 command and output:
>
> shivanshu@thetis:tmp$ cat temp | grep -P '\xe2\x80\x93'
> d1y8@thetis:tmp$
[...]
If you read the pcrepattern man page, you'll see that \xe2
doesn't match the byte e2, but the character of code e2.
If you're in a UTF-8 locale, \xe2 would match the character of
Unicode code point e2 (LATIN SMALL LETTER A WITH CIRCUMFLEX)
which in UTF-8 is written as the bytes c3 a2.
The sequence e2 80 93 is actually the one character U+2013 (EN
DASH). So, here, you either want:
LC_ALL=C grep -P '\xe2\x80\x93'
That is use a locale where characters are single-byte and their
code is the byte value, or assuming the current locale is UTF-8,
use:
grep -P '\x{2013}'
Or, regardless of the locale:
grep -P '(*UTF8)\x{2013}'
--
Stephane
