Open kapouer opened 4 years ago
Are you passing dns.ADDRCONFIG
to dns.lookup()
?
While I can confirm the issue w.r.t. to the test suite, I don't think it's something we need or want to fix. The test suite assumes the availability of IPv4 and that seems fine to me.
I understand about the test suite being written to be run in an env with IPv4 all right. However i'm not sure that this issue won't show up in various other osx/linux environments ? Reading the perl bug report (link above), it seems assuming using ADDRCONFIG for loopback addresses poses some kind of bug.
Fair enough. I've removed the test
label.
glibc and musl do AI_ADDRCONFIG
detection differently:
That difference in approach makes it complicated to fix on node's end. https://github.com/libuv/libuv/issues/2225 is also relevant.
Node's habit of defaulting hints
to AI_ADDRCONFIG
papers over some of the differences. I worry that removing it will end up breaking more than it fixes. For sure it broke our CI last time we tried it.
In a IPv6 only setup that can be reproduced with:
and making sure
/etc/hosts
has that linemany tests fail E.g. parallel/test-http-localaddress throws
Error: bind EINVAL 127.0.0.2
.There is similar problem in Perl, in particular see this message: https://bugs.debian.org/962318#12