How to get crt0.o without syscalls (solved?)

[Randall Nortman]

On Mon, Oct 17, 2005 at 05:39:31PM -0400, Jeff Johnston wrote:
> Randall Nortman wrote:
> >On Mon, Oct 17, 2005 at 11:45:05AM -0400, Jeff Johnston wrote:
> >[...]
> >
> >>When you specify --disable-newlib-supplied-syscalls, the code in 
> >>libgloss/arm is still compiled and there is a crt0.o there.
> >
> >
> >This does not seem to be the case in 1.13.0.  I get rdimon-crt0.o,
> >rdpmon-crt0.o, and redboot-crt0.o, but I'm not using any of these
> >monitors.  I will try with the CVS code when I have some time, since
> >apparently I ought to be using that anyway.
> 
> The missing crt0.o appears to be a typo on my part.  The crt0.o is only 
> being built when newlib-supplied-syscalls are enabled, but this doesn't 
> make sense since libc/sys/arm provides a crt0.o in that case.  I have 
> just made a patch to the CVS sources.  Please try it out.

One more thing has to be done to libgloss/arm/configure.in; the
AC_SUBST macros for the relevant variables are missing, and so they
are never substituted in the Makefile.  Just add these lines and I
think it will work:

AC_SUBST(BUILD_CRT0_TRUE)
AC_SUBST(BUILD_CRT0_FALSE)

I tried this myself but couldn't manage to regenerate configure from
configure.in ("aclocal && autoconf" didn't work), so I just
hand-edited configure and added the substitution code, and then
everything worked.