>>>	Your patch: set seconds to 0 when seconds < 0, 
>>>	BIND 4.9.7 and 8.2.2: set seconds to 1 when seconds <= 0, 
>Either behavior is OK with me.  The important idea is that the
>requested timeout must eventually take place.  Hanging indefinitely is
>not OK.
>>	I checked the current tree, and found that our code is exactly the
>>	same as BIND 4.9.7 and BIND 8.2.3.  so I must ask.
>The code in src/dist/bind/lib/libresolv is OK and contains the fix
>that Vixie implemented based on my report (and a similar report from
>Chrisos Zoulas).  If that's the code that we use to build libresolv.a,
>then libresolv.a probably works correctly.  It's the resolver in libc
>that has the bug described in my PR...sorry if that wasn't clear in my
>most recent posting.

	sorry i was confused.  your fix has multiple items in it:
	(1) timer setup code improvement before "wait" label
	(2) timer setup code improvement after EINTR
	BIND 4.9.7 code includes (1), not (2).

itojun