posix_fadvise (Re: CVS commit: src/sys/kern)

[yamt%mwd.biglobe.ne.jp@localhost (YAMAMOTO Takashi)]

[ moved from source-changes@ ]

> On Mon, Jan 28, 2008 at 07:29:40AM +0900, YAMAMOTO Takashi wrote:
> > why do you bother to keep the broken one?
> > given the purely advisory nature of the syscall,
> > i don't think it's worth to do.
> 
> Because that is how I understood core, after you pointed out that the old
> version actually worked on i386 - I guess there is some misunderstanding
> somewhere. I have no problem making the old syscal just UNIMPL.
> 
> Martin

ok, i see.

then i hereby suggest to make the old version of posix_fadvise
an alias of nullop().

because:

- the effect of this particular function is purely advisory.
  ie. nullop is fine.

- applications do not likely complain on successes.
  ie. nullop can be better than the current broken one.

- it's easily pullup-able for netbsd-4.

- the complication from another obvious solution (#ifdef i386 etc)
  is not worth for this particular case, IMO.

YAMAMOTO Takashi