Re: reversed behavior with nsswitch.conf

Hi,

I have run into this as well.  I am a student in a class that required 
us to set up sendmail in an environment where we don't have access to 
the local DNS, so all our resolution is done via /etc/hosts.  I had to 
be able to send email between our three *nix boxes, and it kept breaking 
on the DNS resolution. Initially, I used mailertable entries to 
circumvent the issue since every other strategy I found to turn off DNS 
resolution for sendmail didn't work.  However, upon reading this, I 
switched the order of the
"hosts" settings in nsswitch.conf and got rid of the mailertable entries 
and it worked.  So yes, I have run into the issue as well.  I am using 
6.0-RELEASE.

Hope this helps.
Joel

Alexander Leidinger wrote:

>Hi,
>
>yesterday I debugged a mail problem where the reverse DNS of the receiver is
>somewhat fucked up (for one receiver the DNS resolving chain was: HostA ->
>IP-A -> HostB -> IP-B -> HostB, for the other receiver there was no reverse
>DNS).
>
>Sendmail doen't like this, but there are ways to circumvent this. So I added
>the hosts in question to /etc/hosts (nsswitch contains "hosts: files dns"),
>and thought this will solve it. It didn't.
>
>A temporary (as in "to get those 10 mails out") work-around of "hosts: files"
>proved that the solution of adding the hosts to /etc/hosts works as expected.
>
>The current solution for the general case is to use "hosts: dns files". It
>gets the correct values for the buggy remote hosts from /etc/hosts, and
>averything else from DNS.
>
>So it seems the order of the use of the entries in nsswitch.conf is reversed.
>
>I noticed this on a 6.0 system. I can't test this on a -current system right
>now (maybe tomorrow or at the weekend), and I don't have a RELENG_6 system.
>
>So if someone can confirm if this bug is visible on -current and on RELENG_6:
> - a fix can be MFCed if it is fixed in -current
> - someone could try to fix it if the same problem exists in -current
> - an errata can be added to the release notes of 6.1 if it isn't
>   fixed until the release (that's the reason why I CCed re_at_)
>
>Bye,
>Alexander.
>
>  
>