2008/12/30 Paolo Bonzini <[email protected]>:
>
>> I don't think so, because the bug does not happen with sed from the
>> heirloom toolchest.
>
> It probably implements its own regex matcher instead of using libc's.
Maybe. But two other GNU programs which probably use libc's regex
matcher, bash and grep, do not behave like sed.
~$ x=fooˢbar
~$ echo ${x//[^a-z]/X}; echo ${x//[a-z]/X}
fooXbar
XXXˢXXX
~$ egrep -o '[^a-z]' <<< $x
ˢ
~$ sed 's/[^a-z]/X/g' <<< $x
fooˢbar
Note: Both bash and grep consider that the 3700 problematic characters
listed in the above attached file are not letters nor digits.
Thomas
