busybox compile error form ARM core???

[Rob Landley]

On Tuesday 30 May 2006 1:45 pm, Bernhard Fischer wrote:
> On Tue, May 30, 2006 at 10:01:50AM +0200, Peter S. Mazinger wrote:
> >On Tue, 30 May 2006, Rob Landley wrote:
> >> On Tuesday 30 May 2006 2:20 am, yangm wrote:
> >>
> >>
> >>   if (!(opt & 2)) {
> >> #if defined(__uClinux__)
> >
> >for uClibc-svn non-MMU the condition to use the internal version would be
> >
> >#if defined __UCLIBC__ && !defined __ARCH_USE_MMU__
> >
> >Peter
> >
> >>     /* reexec for vfork() do continue parent */
> >>     vfork_daemon_rexec (0, 0, argc, argv, "-f");
> >> #else
> >>     daemon (0, 0); /* bb_xdaemon? */
> >> #endif /* uClinux */
> >>   } else {
>
> Landley, what do you think?
>
> in platform.h,
> /* uclibc does not implement daemon for no-mmu systems. */
> #if defined __UCLIBC__ && !defined __ARCH_USE_MMU__
> #define BB_DAEMON(no_chdir, no_close, argc, argv, str) \
> 	vfork_daemon_rexec(no_chdir, no_close, argc, argv, str)
> #else
> #define BB_DAEMON(no_chdir, no_close, argc, argv, str) \
> 	daemon(no_chdir, no_close)
> #endif
>
> Alternatively, we can of course introduce an int bb_daemon(d,c,ac,av,s){}
> to libbb which does the same (may be bigger than the proposal above, didn't
> try).

Actually, what I'd like is a bb_daemon(a, b) that (for the BB_NOMMU case) uses 
approximately the same mechanism as the standalone shell when it needs to 
exec stuff.  Our argc and argv should be made available as globals so this 
thing can access them.

Speaking of which, we still need a config entry that lets people change the 
standalone shell's default "/proc/self/exe" to point to wherever their copy 
of busybox is installed if they don't have /proc.  Break that out into some 
kind of bb_exec_self(argc, argv) or some such.

These are both existing todo items of mine, but if somebody else wanted to 
take a stab at it first...

Rob
-- 
Never bet against the cheap plastic solution.