Re: dynamically created /dev/null is a regular file

[christos%astron.com@localhost (Christos Zoulas)]

In article <30812CA8-FF63-4C48-B455-030384053BCE%nmsu.edu@localhost>,
Brook Milligan  <brook%nmsu.edu@localhost> wrote:
>On Jun 24, 2015, at 10:37 PM, David Holland wrote:
>>> Is this related to kern/33023?
>> 
>> It might be; but given that this mode gets used a fair amount, and
>> also that the device should be getting created in the top of the
>> unionfs, I would guess more likely not.
>
>I'm not actually using unionfs, but the PR suggested similar issues. 
>Perhaps just a red herring.
>
>> I would do a quick look for something (e.g. in your shell startup
>> files) that's clobbering /dev/null somehow. I had a horrible
>> intermittent problem with this at one point until I discovered that
>> less would do it for you if you linked .lesshst to /dev/null?
>
>The startup scripts are just what is in the release.  All I have done is
>untar them, set variables in rc.conf (and a few other things like create
>resolv.conf and edit mail/aliases) and boot.  None of this should be
>changing what the startup scripts do.
>
>There are indeed a bunch of redirections to /dev/null in the startup
>scripts.  Presumably some of these are running before init generates the
>device files and therefore create the errant file.  The init man page
>documents that the device files are created, but not when that happens. 
>Is it possible that this is happening after (at least some of) the
>startup scripts are run?  Should the timing of this be added to the man
>page for completeness?
>
>Thanks again for your help.

Why isn't MAKEDEV invoked with -f?

christos