Re: getaddrinfo and AF_UNSPEC w/ IPv4/6 messes up ai_canonname?

[Pekka Savola <pekkas at netcore dot fi>]

On Sat, 24 Aug 2002, Ulrich Drepper wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Pekka Savola wrote:
> 
> > I noticed something strange when using getaddrinfo to get both IPv4 and 
> > IPv6 addresses w/ AI_CANONNAME.
> 
> Your example code fails to iterate over all results.  This is what I get 
> if I modify the output part of your code to iterate via ai_next:
> 
> host: ftp.kame.net, canon (if any): 2001:200:0:4819:210:f3ff:fe03:4d0
> host: ftp.kame.net, canon (if any): 3ffe:501:4819:2000:210:f3ff:fe03:4d0
> host: ftp.kame.net, canon (if any): kame220.kame.net
> 
> 
> Now you see that indeed the right canonical name is generated.  

Ok.

> Why 
> should the implementation not return the IPv6 entries?  This is what 
> your report seems to imply.  One could argue that returning the numeric 
> format should never happen.  

Exactly.

> But where is this written?  It's easy 
> enough to implement but I am not willing to make a change unless you can 
> motivate it.

Well,

 1) with PF_INET, this does not happen (CNAME entry is returned, not
addresses -- at least as far as I could see): thus the current behaviour
does not seem the same between address families.  

 2) (not really motivation, but still..) e.g. BSD libc does not return
addresses (the test works as "expected"); I haven't checked other
implementations.

Is this enough motivation?

-- 
Pekka Savola                 "Tell me of difficulties surmounted,
Netcore Oy                   not those you stumble over and fall"
Systems. Networks. Security.  -- Robert Jordan: A Crown of Swords