Re: [patch] improve handling of ms_struct

[Roger Sayle <roger at eyesopen dot com>]

Hi Eric,

Re: http://gcc.gnu.org/ml/gcc-patches/2006-04/msg01064.html
Re: http://gcc.gnu.org/ml/gcc-patches/2006-05/msg01214.html
Re: http://gcc.gnu.org/ml/gcc-patches/2006-05/msg01282.html

On Fri, 2 Jun 2006, Eric Christopher wrote:
> > 2006-05-19  Eric Christopher  <echristo@apple.com>
> >
> > 	* stor-layout.c (start_record_layout): Initialize
> > remaining_in_alignment.
> > 	(debug_rli): Output value for remaining_in_alignment.
> > 	(update_alignment_for_field): Unconditionalize
> > 	ms_bitfield_layout_p code. Handle non-bitfield fields. Remove
> > 	extra alignment code.
> > 	(place_field): Don't realign if ms_bitfield_layout_p.
> > Unconditionalize
> > 	ms_bitfield_layout_p code. Rewrite handling of structure fields.
> > 	* tree.h (record_layout_info_s): Remove prev_packed.
> > 	* doc/extend.texi (ms_struct): Add documentation of format.
> >
> > 2006-05-19  Eric Christopher  <echristo@apple.com>
> >
> > 	* gcc.dg/attr-ms_struct-1.c: New.
>
> Ping?

This is OK for mainline.  For those not following the thread, this is
effectively a bug-fix for Microsoft-compatible structure layout and
is therefore suitable for stage3.  We should be on the look out for
problems from the cygwin and mingw folks, who might inadvertantly be
relying on GCC's current non-MS structure layout behaviour.

p.s. when you updated and reposted your patch on the 24th, you should
probably have updated the ChangeLog entries from "2006-05-19" to
"2006-05-24" to avoid any ambiguity when pinging/approving, and to
enable the correct post to be identified on gcc-patches if the posts
become unthreaded (as in this case).  This was likely just an oversight
this time, but to confirm it's the revised patch that I've approved:
http://gcc.gnu.org/ml/gcc-patches/2006-05/msg01282.html

Thanks (and to Joseph for reviewing the documentation changes),
Sorry this has taken so long to get into the tree.

Roger
--