Re: Bug generating libc.so.lds

[Daniel Jacobowitz <drow at mvista dot com>]

On Tue, Apr 01, 2003 at 12:54:31AM -0800, Roland McGrath wrote:
> > And in configure.in right by "dnl No \ in command here because it ends
> > up inside ''." there's a duplicated check for
> > libc_cv_gcc_dwarf2_unwind_info.  I assume the first should be removed.
> 
> THe two checks are different, and actually I think this might be trying to
> address your problem.  The first check does not use -lgcc_eh.  However, its
> result is always ignored and overwritten by the second test.  The following
> patch fixes that so the first test's result of "static" will stick if the
> later tests fail.  Is that the right result for your case?  

No, look a little higher up in the file.  There's actually 

> Are you saying that the check should find the functions in -lgcc without
> -lgcc_eh, or that you really don't have any __register_frame_info function
> available in an installed library when you configure?  If the latter, then
> I think you are SOL.  You might just have to force the test if you don't
> want to rebuild glibc after you have a complete gcc installation.

I've been forcing the test, but here's the situation as I understand
it:

  - The correct outcome for the targets I'm working with is
    no_registry_needed.
  - __register_frame and __register_frame_info are in -lgcc_eh if
    --enable-shared, and -lgcc if --disable-shared.

I have a patch to let the test report no_registry_needed even if the
routines are in libgcc.a.  I'm not enough of an EH guru to know if that
would have unwelcome consequences, though.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer