-symbol-list-lines

[Daniel Jacobowitz]

On Mon, Feb 16, 2004 at 03:55:58PM -0500, Bob Rossi wrote:
> On Mon, Feb 16, 2004 at 03:45:16PM -0500, Daniel Jacobowitz wrote:
> > On Mon, Feb 16, 2004 at 03:40:35PM -0500, Bob Rossi wrote:
> > > I am seeing 2 potential problems,
> > > 
> > > Several lines come back as {pc="0x08048406",line="0"},
> > > should the value "0" be documented? or is it a bug?
> > > 
> > > Also, for some reason, a value is repeated
> > > {pc="0x0804849c",line="38"}. Is this ok?
> > 
> > I can't say anything about what is appropriate (or useful) for the MI
> > interface but I can tell you what's going on.  This matches the
> > structure of the internal line table precisely.
> > 
> > - line 0 means end-of-sequence.  I.E. PCs greater than 0x8048406 that
> > don't have a new starting line do not belong to this line number table.
> 
> I personally can't see why this would be useful to front ends. Can
> anyone else? If it isn't, I will submit a patch to not print this data.
> 
> > - duplicated line numbers are emitted by the compiler to mark, mostly,
> > the ends of prologues.  Sometimes the prologue is zero-sized.
> 
> Would keeping a list internally of numbers that have been sent out
> already be reasonable? or is it repeating the data unimportant?
> 
> I would also be willing to fix this.

I don't know.  I think I'd prefer for frontends to tolerate this data;
just because I can't think of a use for it now doesn't mean someone
else won't later.  I was recently reminded that "front-end" can be a
pretty broad concept, not just user interfaces.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer