Another --disable-checking bootstrap failure

[Jeffrey A Law]

On Mon, 2004-09-13 at 21:35, Diego Novillo wrote:
> On Mon, 2004-09-13 at 23:23, Jeffrey A Law wrote:
> > On Mon, 2004-09-13 at 21:05, Diego Novillo wrote:
> > > Is anyone seeing this on a --disable-checking bootstrap?
> > > 
> > > Jan, you have been making changes in this area.  Do you recognize this? 
> > > We seem to be freeing SSA_NAMEs too early.  This doesn't trigger with
> > > --enable-checking nor with -v -save-temps.
> > What varray are we walking in frame #7?
> > 
> ssa_names, that's why I have the impression we may be freeing them too
> early.  I'm not completely sure, though.
> 
> It's also fairly recent (last couple of days), the infinite loop we have
> now with --disabled-checking occurs in stage3 and is not related to
> this.
Actually, someone might be forgetting to call free_df...

Frame 7 is the walk down the SSA_NAMEs varray
Frame 6 should be walking through a SSA_NAME tree node
Frame 5 is walking through the SSA_NAME annotation (didn't that
        code change recently?)
Frame 4 is walking through the dataflow information -- why oh why
        do we have dataflow_d's still live at this point?
Frame 3 is some kind of tree node, presumably a statement.
Frame 2 is a basic block

Jeff