Stop on solib load

[Vladimir Prus]

On Monday 31 December 2007 01:59:51 Hamish Rodda wrote:
> On Mon, 31 Dec 2007 09:30:22 am Andreas Pakulat wrote:
> > On 31.12.07 08:45:57, Hamish Rodda wrote:
> > > On Sun, 30 Dec 2007 06:55:34 pm Vladimir Prus wrote:
> > > > Hamish,
> > > > on IRC, you've asked:
> > > >
> > > >    <blackarrow> volodya: can we remove the stop at each shared lib
> > > > loading stuff now?
> > > >
> > > > In fact, I believe it was removed in revision 748808. There was:
> > > >
> > > >    queueCmd(new GDBCommand(GDBMI::GdbSet, "stop-on-solib-events 1"));
> > > >
> > > > which is now gone. I'm aware of one remnant of that logic -- where we
> > > > still ignore any error GDB reports for -break-insert operation -- while
> > > > now we should not.
> > > >
> > > > Do you have some other code in mind?
> > >
> > > Actually I was wondering about kdev 3.5...
> >
> > No way, that stuff still needs a cvs version of gdb, we can't even
> > require gdb 6.7 at the moment as thats simply too new - IMO.
> 
> Sure, that's why I was thinking we would check the version first; I presume 
> the pending breakpoints stuff already does that...?

KDE 3.5 sends -list-features command right at the start. When the list of
features includes 'pending-breakpoints', we add -f option to -break-insert,
and don't try to stop on solib load. If that feature is not present, we get
back to emulating pending breakpoint, as we did before.

(The -list-features command was introduced by myself in order to check specifically
for MI features -- version string is less reliable, since version in CVS is
not telling much, and since features can well be backported)

- Volodya