Re: AI_V4MAPPED/AI_ALL

[Roland McGrath <roland at redhat dot com>]

> Ok, AI_V4MAPPED and AI_ALL are specified in POSIX, rfc2553 and
> rfc3493. So I doubt that it will be changed in the near future
> again, I think something specified in POSIX should be stable 
> enought to include in glibc.

Agreed.  I'm not sure the RFCs always indicate a stable interface, but
what's in 1003.1-2001 we definitely want to implement and these are there.

Can you add a test case that is broken by the current code and fixed by
your patch?  We can add an xtests case that requires some particular DNS
zone data to do the test.  At least for forward zone data, I can set up a
glibc-test.gnu.org subdomain with test data if needed.

> I append a patch which adds both defines to resolv/netdb.h and
> implements it in getaddrinfo(). (Looking at the ugly getaddrinfo
> code, it seems we really need a new, clean implementation :( ).

Definitely true.  It was close to the top of my list before you fixed the
AF_UNSPEC behavior.  I can probably find time before the next release to do
the rewrite.  If you have any list of known issues with the current
implementation (aside from its internal structure), and especially if we
can soup up the tests to cover more details, that would be a help.


Thanks,
Roland