cvs commit: src/sys/kern kern_exec.c

[Maxime Henrion]

Kris Kennaway wrote:
> On Sat, Mar 13, 2004 at 12:04:39AM +0200, Ruslan Ermilov wrote:
> > On Fri, Mar 12, 2004 at 01:54:50PM -0800, Kris Kennaway wrote:
> > > On Fri, Mar 12, 2004 at 01:06:20PM -0800, Ruslan Ermilov wrote:
> > > > ru          2004/03/12 13:06:20 PST
> > > > 
> > > >   FreeBSD src repository
> > > > 
> > > >   Modified files:
> > > >     sys/kern             kern_exec.c 
> > > >   Log:
> > > >   Do what the execve(2) manpage says and enforce what a Strictly
> > > >   Conforming POSIX application should do by disallowing the argv
> > > >   argument to be NULL.
> > > >   
> > > >   PR:             kern/33738
> > > >   Submitted by:   Marc Olzheim, Serge van den Boom
> > > >   OK'ed by:       nectar
> > > 
> > > Is this likely to break any third party software (e.g. ports)?  If so,
> > > please bump __FreeBSD_version.
> > > 
> > It'll break only broken software, so I don't think the bump is in order.
> 
> Well, that doesn't follow.  When third party software suddenly stops
> running on FreeBSD, we need a way to predict in advance whether or not
> it will run on a given kernel -- whether the software is "broken" in
> an ideological sense or not.  For example, we need to mark the port
> BROKEN on those versions of FreeBSD so the user doesn't waste their
> time worrying about or reporting a known problem.
> 
> __FreeBSD_version bumps have negligible cost, so if we're both wrong
> and this has no effect, then no harm is done.

I believe that what Ruslan wanted to say is that since software that
will break after this commit are already broken, we can just patch the
application via the port and have it working whatever version of FreeBSD
you're running.  This isn't like if that change was an API change that
forces us to do stuff in one way for some __FreeBSD_version and in
another way for another __FreeBSD_version.  We can, and should, do it
in just one way, the correct one.

So I don't think a __FreeBSD_version bump is needed either.

Cheers,
Maxime