[PATCH]: Rewrite reassociation pass => Ada bootstrap failure on x86-linux

[Jeffrey A Law]

On Wed, 2006-01-04 at 17:37 +0000, Graham Stott wrote:
> Hi jeff,
> 
> 
> --- Jeffrey A Law <law@redhat.com> wrote:
> 
> [snip]
> > This has been bootstrapped and regression tested on i686-pc-linux-gnu
> > (ada included).
> > 
> > 
> > 
> > 
> > > 	* tree-cfg.c (bsi_replace): Remove the original statement
> > 	from the EH throw statement table.
> > 
> > Index: tree-cfg.c
> > ===================================================================
> > *** tree-cfg.c	(revision 109334)
> > --- tree-cfg.c	(working copy)
> > *************** bsi_replace (const block_stmt_iterator *
> > *** 2869,2875 ****
> >       {
> >         eh_region = lookup_stmt_eh_region (orig_stmt);
> >         if (eh_region >= 0)
> > ! 	add_stmt_to_eh_region (stmt, eh_region);
> >       }
> >   
> >     delink_stmt_imm_use (orig_stmt);
> > --- 2869,2878 ----
> >       {
> >         eh_region = lookup_stmt_eh_region (orig_stmt);
> >         if (eh_region >= 0)
> > ! 	{
> > ! 	  remove_stmt_from_eh_region (stmt);
> > ! 	  add_stmt_to_eh_region (stmt, eh_region);
> > ! 	}
> >       }
> >   
> >     delink_stmt_imm_use (orig_stmt);
> > 
> 
> This looks a bit odd to me.
> 
> The ChangeLog entry says "remove original stmt" but you appear
> to be removing the new stmt and not the orignal!
Crap.  You're absolutely right.   I used orig_stmt on a devel box;
and just entered the change by hand in my real tree and mucked it
up in the process.  Grrr.  I'll get a fresh bootstrap test started. 

I also need to update the copyright date for tree-cfg.c, but that's
slightly less important.


> Also if we really should be removing the orignal stmt shouln't
> we be doing it independently of the "preserve_eh_info".
I'm less sure of this -- I can almost envision asking for a
statement replacement where we don't muck with the EH info.  For
example, we're moving a statement to a different point in the 
same EH region and replacing the original with a copy.

Jeff