The following reply was made to PR port-m68k/30924; it has been noted by GNATS.

From: YAMAMOTO Takashi <yamt@mwd.biglobe.ne.jp>
To: cube@cubidou.net
Cc: christos@zoulas.com, gnats-bugs@netbsd.org
Subject: Re: port-m68k/30924: m68k copyinout_t breakage
Date: Fri, 12 Aug 2005 19:00:10 +0900

 hi,
 
 > > |  I have two options:
 > > |  
 > > |   - put copyinout_t in its own header, and include it when needed
 > > |   - put copyinout_t in types.h
 > > 
 > > There is also the:
 > >     - fix clockframe #define abuse so that it can be put in <system.h>
 > 
 > So, that part is fixed,
 
 thanks for doing it.
 
 > but yamt says it's a bad idea to have systm.h
 > included from time.h.
 
 what i said was about including systm.h from type.h.
 i think it's a bad idea because type.h is
 more "primitive" than systm.h, IMO.
 
 from time.h, it's ...acceptable. :)
 
 > types.h looks reasonable to me.
 
 it's also acceptable to me for now.
 
 > I don't want to move timer_create1's
 > prototype away as it would only solve the issue for that one, although
 > it might reproduce with another header.
 
 i think sys/time.h is one of the worst headers in the respect of
 circular dependency.
 it has both of "primitive" types like timespec and
 high-level prototypes like timer_create1.
 while the latter is not needed for the most part of kernel,
 sys/param.h includes it for the former.
 it also propagates bogus dependencies on headers like sys/callout.h
 throughout kernel.
 
 YAMAMOTO Takashi