Subject: Re: NetBSD/pc98
To: None <kiyohara@kk.iij4u.or.jp>
From: Izumi Tsutsui <tsutsui@ceres.dti.ne.jp>
List: current-users
Date: 08/21/2006 22:20:37
kiyohara@kk.iij4u.or.jp wrote:

> I put sources and diff.
>   ftp://ftp.netbsd.org/pub/NetBSD/misc/kiyohara/pc98/pc98-20060819.diff
>   ftp://ftp.netbsd.org/pub/NetBSD/misc/kiyohara/pc98/pc98-20060819.tar.gz

- it looks a bit ugly to use flags of config(9) to denote
  NE2000 variants in a config file.
  Is there no way to probe them properly?
  How FreeBSD/pc98 detects them?
- isn't it better to attach legacy devices (dcom, pic, wdc etc.)
  at cbus (which is a dumb bus anyway) rather than mainbus?
  I have a handmade two port dcom cbus board..
- how will you handle symlinks in include dir on import?
  which dir will machine@ symlink point? (no pc98/include/Makefile)
- is there any plan to implement pc98 specific disksubr.c and disklabel.h?

Some other humble comments:

- needs some KNF (no parentheses after return, uintNN_t etc).
- bus_space.c has some retained comments about ISA 
- todr(9) functions should be separated from clock.c
  (though they are not implemented yet)
- include "ioconf.h" rather than "extern struct cfdriver foo_cd"
- is it worth to use MALLOC(9) in intr.c? (I know it's taken from x86 :-)
- __UNCONST() in cbus_intr_typename() looks ugly.
  why not to make it return (const char *)?
- "pc98" strings in dev/cbus seems ugly
  (I know luna68k had optional cbus extension slot, but
   it may be better to move them under arch/pc98/cbus for now?)
- please remove magic numbers as much as possible

---
Izumi Tsutsui