Re: [PATCH] Add autoload-breakpoints [3/7] ReportAsync-doc

[Hui Zhu <hui_zhu at mentor dot com>]

On 03/22/12 02:27, Eli Zaretskii wrote:
> > Date: Wed, 21 Mar 2012 15:23:02 +0800
> > From: Hui Zhu<hui_zhu@mentor.com>
> > CC:<gdb-patches@sourceware.org>,<stan_shebs@mentor.com>
> >
> > > > > > +The @value{GDBN} remote serial protocol includes @dfn{reportasync},
> > > > > > +packets that the stub can report status or do an operation in any time.
> > > > >
> > > > > I don't understand the "do an operation in any time" part.  What does
> > > > > it mean in practice?
> > > >
> > > > It is means that after gdb connected with the stub, the stub can report
> > > > when the inferior is running and GDB is waiting it, or inferior is
> > > > stoped and GDB just control it.
> > > > It make the stub can report the status for any time.
> > >
> > > You mean, as in "asynchronously"?
> >
> > Yes, that is my mean.
>
> Then please use
>
>   The @value{GDBN} remote serial protocol supports @dfn{reportasync}
>   that can be used asynchronously to report status or send commands.
>
> > > I still don't understand why you used the word "simple".  "Simple" as
> > > opposed to what other packet handling?
> >
> > Maybe "normal" is better. "Normal" as opposed to reportasync packets.
>
> "As usual" or "usual" would be better.  But see below.
>
> > > > >     After the handshake, the stub can send @samp{reportasync} packets to
> > > > >     @value{GDBN}, using the same packet format as in simple remote serial
> > > > >     protocol.
> > > > >
> > > > > Is this only for the failed handshake?  If so, perhaps we need to tell
> > > > > what happens when the handshake succeeds.
> > > >
> > > > No, this is for success.
> > >
> > > The problem here is that you again used "simple protocol", and before
> > > that you mentioned "simple packet" when the handshake failed.
> > >
> > > So this again boils down to the same question: why do you use the word
> > > "simple" in this context?
> >
> > Yes, it looks really odd even if change it to normal.  That is because
> > after shakes mode, the packet will use the normal format
>
> But you said
>
>    After the handshake, the stub can send @samp{reportasync} packets to
>    @value{GDBN}, using the same packet format as in simple remote serial
>    protocol.
>
> which actually tells me that the "normal" packets are _not_ in
> contrast with reportasync packets, since reportasync packets can be
> sent in the "normal format".  What am I missing?

Yes, after success shake hands, the stub will send packet in normal 
format to GDB.
If need, GDB will answer this packets in normal format.

For example:
static void
remote_reportasync_handler (struct remote_state *rs)
{
  int len;
  struct cleanup *old_chain
    = make_cleanup_restore_integer (&inside_reportasync_handler);

  /* Shake hands with remote part.  */
  serial_write (remote_desc, "^", 1);

  inside_reportasync_handler = 1;

  if (getpkt_sane (&rs->buf, &rs->buf_size, 0) < 0)
    return;

  if (remote_debug)
    fprintf_unfiltered (gdb_stdlog, "ReportAsync: %s\n", rs->buf);

  if (strncmp (rs->buf, "QBDP", strlen ("QBDP")) == 0)
    {
      if (parse_autoload_breakpoint_definition_to_breakpoints
	      (rs->buf + strlen ("QBDP")) == 0)
	putpkt ("OK");
      else
	putpkt ("E01");
    }
  else
    putpkt ("");

  do_cleanups (old_chain);
}

This function will be call when GDB got a ^ and is OK to shake hands.
It call serial_write (remote_desc, "^", 1); to shake hands.  Then call 
getpkt_sane to get the package from the stub.  And use putpkt reply.

Thanks,
Hui