resolver change?

[Robert Watson]

On Sun, 27 Jan 2008, Dag-Erling Smørgrav wrote:

> Hajimu UMEMOTO <ume at mahoroba.org> writes:
>> Dag-Erling Smørgrav <des at des.no> writes:
>>> OK, so the resolver now uses a process-internal cache?  Is there any way 
>>> to turn it off?
>> No, our resolver doesn't have a process-internal cache at all.
>
> So what's going on?
>
> Looking back through my logs, it was working correctly as late as January 
> 13, so something broke between then and January 21.

A casual glance suggests no pertinent changes in that timeframe -- however, 
were you just updating your kernel, or also userspace, and in particular, 
libc?

> My name server does *not* forward queries, it goes straight to the source. 
> Everything looks fine if I run host(1) multiple times, it only seems to fail 
> when successive lookups are made from the same process.

What happens if you back out getaddrinfo.c:1.86:

----------------------------
revision 1.86
date: 2007/09/05 18:08:14;  author: jinmei;  state: Exp;  lines: +19 -2
ensure the head entry of addrinfo chain has non-NULL ai_canonname to be
compliant with RFC3493.

PR: standards/114910
Approved by:   ume (mentor)
Approved by:   re
MFC after:     1 week
----------------------------



Robert N M Watson
Computer Laboratory
University of Cambridge