Re: RFC: The unwind specification draft for GNU/Linux/ia64

["H. J. Lu" <hjl at lucon dot org>]

On Fri, Jul 16, 2004 at 04:58:42PM -0700, Jim Wilson wrote:
> On Fri, 2004-07-16 at 14:05, H. J. Lu wrote:
> > This proposal is the result of discussion between Mark, David and Intel
> > compiler people.
> 
> That is a reasonable list.  What about ia64-hpux?  Is Mark representing
> just the C++ ABI, or that and ia64-hpux?  If he isn't representing
> ia64-hpux, then we need to get someone else to make sure this is OK for
> ia64-hpux.
> 

My proposal shouldn't affect ia64-hpux.

> As currently set up, the ia64-hpux port always uses the system libunwind
> library.  This is a different library than David's.
> 

Unless ia64-hpux uses or plans to the same ABI as Linux, I don't see
how my proposal should apply to ia64-hpux.

> > Here is my unfinished patch for gcc. The stub, XXX, in libgcc just
> > calls __libunwind_XXX in libunwind.
> 
> That is a pretty big complicated patch.  I haven't tried to figure out
> what all it is doing.  It would be nice if you included an explanation
> of what it is doing.  Also, we might need doc changes for some of this
> stuff.  The IA-64 install docs mention --enable-libunwind-exceptions for
> instance.

It does a few things for Linux targets:

1. It always uses libunwind when and only when the unwind ABI calls for
libunwind.
2. It uses the system libunwind when and only when
--with-system-libunwind is used.
3. If the system libunwind isn't used, an ABI compliant libunwind will
be built.
4. When libunwind is used, the existing unwind functions in
libgcc_s.so.1 will just call those in libunwind, which can be either
the system libunwind or the one come with gcc.

I will update the IA-64 install docs.


H.J.