nss_wrapper support for musl-libc

[Andreas Schneider]

On Tuesday, 28 February 2017 12:20:00 CET Dennis Schridde wrote:
> Hi Andreas!
> 
> On Montag, 20. Februar 2017 17:08:33 CET Dennis Schridde wrote:
> > > About tests/test_getaddrinfo.c:
> > > 
> > > I think you should report those to the developers of musl-libc. These
> > > tests
> > > pass correctly on Linux, Solaris and BSD. All have different libc
> > > implementations.
> > > 
> > >  	rc = getaddrinfo(NULL, "echo", &hints, &res);
> > > 
> > > -	assert_int_equal(rc, EAI_NONAME);
> > > +	if (rc != EAI_SERVICE) { // musl-libc returns EAI_SERVICE
> > > +		assert_int_equal(rc, EAI_NONAME);
> > > +	}
> > > 
> > > I would argue that EAI_NONAME is the correct error code here as it is
> > > NULL.
> > > EAI_SERICE is obviously the wrong return code.
> > 
> > Is EAI_NONAME correct, because "echo" is specified in /etc/services as
> > 7/tcp or 7/udp? I.e. because a remote "node" address is necessary to
> > communicate via TCP or UDP and getaddrinfo is supposed to deduct that
> > from
> > /etc/services?
> 
> From reading freeaddrinfo(3p) (redirect from getaddrinfo(3p)) it is not
> clear to me what the expected behaviour is. It says about these two errors:
> 
> [EAI_NONAME]
>      The name does not resolve for the supplied parameters.
> 
>      Neither nodename nor servname were supplied. At least one of these
> shall be supplied.
> 
> [EAI_SERVICE]
>      The service passed was not recognized for the specified socket type.
> 
> It seems that EAI_NONAME is about resolving the nodename only, while
> EAI_SERVICE indicates whether the provided servname and hints represent a
> sane combination.
> 
> The test specifies no nodename, only a servname. Thus musl's behaviour of
> indicating that "the service passed was not recognized for the specified
> socket type" seems to be spot on. Going by the specs, it seems that either
> GNU libc's behaviour is wrong or the specification is too lax in this
> regard.
> 
> Another question would be what we are actually testing here: Conformance of
> libc to POSIX, or whether nss_wrapper works as intended?

Normally I try to mimic the behaviour of glibc. Then I test on other platforms 
BSD, OpenSolaris if they differ. They often did the same as glibc.

I'm currently checking why BSD and Solaris nightly build are not working.


	Andreas

-- 
Andreas Schneider                   GPG-ID: CC014E3D
Samba Team                             asn at samba.org
www.samba.org