On Tue, Sep 08, 2026 at 11:53:03AM +0200, Vincent Lefevre wrote: > On 2026-09-08 11:31:37 +0200, [email protected] wrote: > > On Tue, Sep 08, 2026 at 11:19:36AM +0200, Vincent Lefevre wrote: > > > Other applications are also affected by the issue, but in different > > > and surprising ways: > > > > [...] > > > > all of them are consistent with the hypothesis that the > > application is getting both an IPv6 (AAAA record) and > > IPv4 (A) address when trying to resolve the name, and > > for some reason first try the IPv6 (and for some other > > don't fall back to IPv4 on failure). > > I think that their use of getaddrinfo is different. It appears > that only AF_UNSPEC is broken (see my other message about this). > > > You said you already edited your /etc/gai.conf? > > No, not on *this* machine. This was something I had to do in 2017 > on a machine at home (not the same network) because my ISP had an > IPv6 routing issue for a few weeks.
Then, it is a good chance to try something along the lines of precedence ::ffff:0:0/96 100 in there (that line should be already there, but commented out). This gives precedence to IPv4 in getaddrinfo (::ffff:0:0 is an IPv4 in IPv6 disguise [1]). Usually, IPv6 has precedence. Cheers [1] https://en.wikipedia.org/wiki/IPv6_address#Transition_from_IPv4 -- t
signature.asc
Description: PGP signature
