Re: [RFA/tracepoint] Make GDB can work with some old GDB server

[Jan Kratochvil <jan.kratochvil@redhat.com>]

On Fri, 24 Jun 2011 10:46:58 +0200, Hui Zhu wrote:
> Sending packet: $qTStatus#49...Ack
> Packet received: E22
> trace API error 0x2.
[...]
> This is because this kgdb(I think trunk have fixed this bug) don't
> support qtstatus and reply -0x22.
> So gdb throw a error.  Then all connect process stop.

Why not to just put it more close to the error invocation?  Still some calls
to target_get_trace_status would remain unprotected otherwise.


Thanks,
Jan


gdb/
2011-07-07  Jan Kratochvil  <jan.kratochvil@redhat.com>

	Work around kgdb.
	* remote.c (remote_get_trace_status): New variable ex.  Put
	remote_get_noisy_reply into TRY_CATCH.  Call exception_fprintf for it.

--- a/gdb/remote.c
+++ b/gdb/remote.c
@@ -10045,11 +10045,21 @@ remote_get_trace_status (struct trace_status *ts)
   char *p;
   /* FIXME we need to get register block size some other way.  */
   extern int trace_regblock_size;
+  volatile struct gdb_exception ex;
 
   trace_regblock_size = get_remote_arch_state ()->sizeof_g_packet;
 
   putpkt ("qTStatus");
-  p = remote_get_noisy_reply (&target_buf, &target_buf_size);
+
+  TRY_CATCH (ex, RETURN_MASK_ERROR)
+    {
+      p = remote_get_noisy_reply (&target_buf, &target_buf_size);
+    }
+  if (ex.reason < 0)
+    {
+      exception_fprintf (gdb_stderr, ex, "qTStatus: ");
+      return -1;
+    }
 
   /* If the remote target doesn't do tracing, flag it.  */
   if (*p == '\0')