Description of problem:

With a locale set to UTF-8, the '\w' regular expression should match any character that is considered alphanumeric in Unicode. This works fine as long as there is at least one alphanumeric ASCII character on the same line.

$ echo 'hääyöaie' | grep '\w'
hääyöaie

When a line contains no alphanumeric ASCII characters, Unicode letters are not matched anymore:

$ echo 'ä' | grep '\w'
$ echo 'ä*$ ?!' | grep '\w'
$ echo 'äæøüåő' | grep '\w'
$

To fix this issue, the '\w' pattern should match Unicode letters either always (if a UTF-8 locale is set) or not at all (otherwise); it should not depend on their context.


Version-Release number of selected component (if applicable):
grep-2.16-1.el7.x86_64


How reproducible:
100%


Further information:

Note that the whole word 'hääyöaie' in the example above is highlighted in red, meaning that every single character matched the pattern '\w'. Compare with the following cases where only letters -- whether ASCII or Unicode -- and digits are matched, as opposed to the non-alphanumeric asterisk symbols. This indicates that the Unicode letters are indeed matched by the pattern under certain conditions:

$ echo 'h*ä*ä*y*ö*a*i*e' | grep '\w'
h*ä*ä*y*ö*a*i*e
^ ^ ^ ^ ^ ^ ^ ^------ highlighted
$ echo 'ä*ö*1*2*3' | grep '\w'
ä*ö*1*2*3
^ ^ ^ ^------ highlighted