Debugging warning_at and friends

[Richard Biener]

On Mon, Feb 16, 2026 at 11:57 AM Jakub Jelinek <jakub@redhat.com> wrote:
>
> On Mon, Feb 16, 2026 at 10:34:52AM +0100, Richard Biener via Gcc wrote:
> > Hi,
> >
> > Ever since the change of now passing the diagnostic option ID as
> > struct (now diagnostics::option_id) I struggle with what previously
> > was as simple as
> >
> > (gdb) b warning_at if id == OPT_Wuninitialized
> >
> > doing
> >
> > (gdb) b warning_at if option_id == OPT_Wuninitialized
> >
> > causes
> >
> > Error in testing condition for breakpoint 5.1:
> > Internal error: incompatible overload candidates proposed
> >
> > and I'm facing gazillion calls of warning_at that in the end are
> > not emitted and not relevant to what I am trying to debug.
> >
> > Can I ask for a sensible workaround / tip as to where I can
> > break to catch an actually emitted diagnostic reliably,
> > eventually filtered by the option kind?
>
> Just use break-on-diagnostics?  I think that function doesn't
> have OPT_W* value passed to it in any way, so one would need
> to query value from an outer frame.

Ah, thanks.  Didn't know about that (will try to remember).  That seems
to work for me.

Richard.

>
>         Jakub
>