Re: CVS commit: src/sys/kern

["Mindaugas R." <rmind%NetBSD.org@localhost>]

Elad Efrat <elad%bsd.org.il@localhost> wrote:
> therefore, I kindly ask you again to revert.
I would certainly revert if at least one more people would ask for that (I
re-asked them to take a look at this). But I still have no objections on this.

> later on, we lookup the table itself. according to your change, if it
> doesn't exist, we return ENOENT. this is *after* a file_free() call that
> (lacking return value :) succeeded in removing a file from the table.
Heh, so rather let to dereference NULL pointer?
Yes, in a normal case, instead of this check, there could be simply KASSERT,
but in this case, when there is a lack of locking (this code is not MP
safe,by the way) - I would really rather put the checks in such places. To be
honest, as long I look at the code, I would tend to agree with Jared to mark
fileassoc as experimental, while somebody fixes/finishes it.

Anyway, your primary claim that "it is wrong" is not true.

-- 
Best regards,
Mindaugas
www.NetBSD.org