Mats Erik Andersson <[email protected]> writes:

> Hello there,
>
> since Simon Josefsson and Tim Rühsen are both involved in libidn2,
> this bug is doubly relevant here.
>
> The following call
>
>    host = "::1";
>
>    idna_to_ascii_lz(host, &newhost, 0);

Hi.  Sorry for slow response.  The real problem here is that IDNA is
only specified for domain names, and ::1 is clearly not a domain name so
InetUtils should not be calling libidn/libidn2 functions on strings like
that.

These days I believe the general recommendation is to use system
getaddrinfo() which has been updated to support IDNA on the platforms we
care about, and linking directly to libidn/libidn2 does not really help
except to trigger bugs.  Of course, we should test that only relying on
getaddrinfo() works the same as our existing uses of libidn/libidn2
though!  There are legitimate uses of calls to libidn/libidn2 so not
everything should be removed.

/Simon

Attachment: signature.asc
Description: PGP signature

Reply via email to