Re: General host configuration fragment purge

[Richard Henderson <rth at redhat dot com>]

On Sun, Jan 13, 2002 at 02:03:40PM -0800, Zack Weinberg wrote:
> On Fri, Jan 11, 2002 at 09:25:18PM -0800, Douglas B. Rupp wrote:
> > Your change to move DO_GLOBAL_[CD]TORS_BODY from xm-interix.h to interix.h
> > has uncovered this.
[..]
> > libbackend.a(i386.o): In function `ix86_svr3_asm_out_constructor':
> > i386.c:12375: undefined reference to `init_section'
> > 
> > I think the code in ix86_svr3_asm_out_constructor needs to be moved into
> > a tm file, maybe i386.h.
> 
> I think, given the way interix DO_GLOBAL_CTORS_BODY seems to work,
> that you should be enabling init_section().  However, I don't fully
> understand this area.

That depends on how much startup code comes from the vendor and
how much is being supplied by gcc.  I don't know the answer to
that question for interix.

> Richard Henderson was the last person to touch this code; can you
> comment?

The function itself can't be moved to a header file -- it's part of
the target structure and must be addressed as a function.  But we do
need some tm.h key by which we can determine that we wish to make
use of the svr3 style of pushing the constructor list onto the stack.

Perhaps INIT_SECTION_ASM_OP is good enough here.


r~