Re: [PATCH v3 2/2] trace-cmd: Trace timesync to find pids that map vCPUs

[Tzvetomir Stoyanov <tz.stoyanov@gmail.com>]

On Fri, May 14, 2021, 16:12 Steven Rostedt <rostedt@goodmis.org> wrote:
>
> On Fri, 14 May 2021 07:10:51 +0300
> Tzvetomir Stoyanov <tz.stoyanov@gmail.com> wrote:
>
> > > @@ -3250,10 +3282,18 @@ static char *parse_guest_name(char *gname, int *cid, int *port)
> > >         } else if (is_digits(gname))
> > >                 *cid = atoi(gname);
> > >
> > > -       read_qemu_guests();
> > > +       if (*cid < 0)
> > > +               read_qemu_guests();
> > > +
> > > +       if (*cid < 0)
> > > +               return NULL;
> >
> > This check is not needed. If cid is not part of the string, let
> > read_qemu_guests() to try discover the VMs, instead of returning NULL.
>
> I'm confused. That's basically what the above does.
>
>         *cid = -1;
>
>         if (<cid is in name>) {
>                 *cid = look_for_cid_via_tracing();
>         }
>
>         if (*cid < -1)
>                 read_qemu_guests();
>
> That is, if the cid isn't part of the name, then it will call
> read_qemu_guests().
>
> The second check is in case the read_qemu_guests() fails to find the cid
> either.
>
> What am I missing?

cid is not updated between both checks, it always fails - even though
read_qemu_guests() may succeed to find the VM with that name.

>
> -- Steve