[uClinux-dev] Re: fork on uClinux

[Rob Landley]

On Thursday 25 May 2006 12:48 pm, Shaun Jackman wrote:
> On 5/24/06, Greg Ungerer <gerg at snapgear.com> wrote:
> > I suspect you will find a number of changes in the uClinux-dist
> > version of busybox that have not been submitted back through
> > to the busybox original sources. (And undoubtly most of them
> > will be in some way uClinux specific).
> >
> > Almost certainly the fork define above will be one of them.
>
> I see. Who heads up the development of the uClinux-dist version of
> busybox? Does this person distribute an easy-to-read diff of the
> patches necessary to run busybox on uClinux? Does he or she have any
> plans to send this work upstream?

I'm interested in merging this, by the way.  I got forwarded a copy of the 
uClinux patch a few months ago, but the version was _ancient_.

I recently added bb_spawn() and bb_xspawn(), which work through vfork.  (And 
they work through vfork for _everybody_.  If we can use vfork anywhere, we 
should be able to use it everywhere.  Not that this is _easy_, but I used to 
do threaded programming under OS/2 and that's actually slightly worse, so...)  
I plan to swat the separate vfork_daemon_rexec() and give bb_spawn() the 
ability to call daemonize on the new thing (as soon as I figure out how).

I also have various related issues queued up for things like bbsh.  (For 
example, the lash code conditionally does vfork.  It's conditional because 
right afterwards it fiddles with signals.  I need to figure out if there's a 
vfork variant (possibly calling clone directly?) that gives the child 
separate signals)...

I'm interested in making mainstream busybox work better with uClinux.  
Eventually, I'd like you to be able to use vanilla and submit patches 
directly to the busybox list...

Rob
-- 
Never bet against the cheap plastic solution.