Resolution differences for getaddrinfo versus host/dig/delv

[Phil Mayers]

On 18/11/15 21:26, Stephane Bortzmeyer wrote:
> On Wed, Nov 18, 2015 at 12:19:57PM +0000,
>   Phil Mayers <p.mayers at imperial.ac.uk> wrote
>   a message of 44 lines which said:
>
>> I suspect getaddrinfo isn't parsing the DNS response for some reason.
> ...
>> Obviously the *.thing on the RHS of the first CNAME is weird, but is it
>> illegal?
>
> Yes, for a *host* name (no for a *domain* name). See Tony Finch's
> explanation.
>
> In the GNU libc, the relevant code is in resolv/res_comp.c and
> includes this function, which tests that a *host* name is
> [a-z0-9\.\-]+ :

Thanks for the replies all; I guess the missing bit in my knowledge was 
that glibc and BSD libc apply the hostname-validity rules to CNAME 
targets (entirely reasonably).

FWIW, that hostname resolves OK on Windows (and as someone else noted, 
MacOS X) so here's hoping the lack of resolution never becomes an issue 
- given it "works" on "their" platform I don't see Microsoft being 
persuaded they're doing something wrong, based on past interactions with 
their 365 ops team :o/