Re: Updating INSN_STMT records.

[Jan Hubicka <jh at suse dot cz>]

> On Mon, Jun 03, 2002 at 06:48:20PM +0200, Jan Hubicka wrote:
> > Actually I was thinking about it more, and we will definitly want to
> > behave differently in some other cases, so probably adding third
> > argument usually used as INSN_STMT (after) sounds more sensible to me.
> 
> I thought that's what you meant the first time.  ;-)

What I am nervous about is when I continue on storing line number,
column number whatever in the insn, I will need some way to pass this
into the function. If I create INSN_STMT, INSN_LINE_NUMBER, INSN_COLUMN
and so on, it will be ugly.

I am still thinking about the idea of "statement records" containing
nesting, line number, file, column and whaever else we store about the
staement for debug output and using just single pointer in INSN
container. We've disucssed this earlier, but you argued that it will end
up eating more memory.

I am not 100% sure it will, as single statement can (usually does?)
produce multiple instructions and we will save the record just once and
not copy it in each INSN.
What I was thinking about is that we even can just store pointer to STMT
tree node, once all targets are function at time.
This will require more involved changes to way how the debug info in RTL
is gneerated :(

Honza
> 
> 
> r~