Re: [RFC] -thread-select double print stack frame

[Nick Roberts <nickrob@snap.net.nz>]

 > > But why do you think that while file and line details are currenly
 > > duplicated in MI, they are not in CLI?  (I mean what is the underlying
 > > logic/cause, not the just the code differences).
 > 
 > Here's what I get from the CLI:
 > 
 > [Switching to thread 1 (Thread 46912506009296 (LWP 28747))]#0  main
 > (argc=1, argv=0x7fffffffe178)
 >     at /space/fsf/commit/src/gdb/gdb.c:28
 > 28      {
 > 
 > The first line is the frame description and becomes the new-thread-id
 > and the frame={} tuple.  The second line ("28 {") comes from
 > print_source_lines (from lack of ui_source_list).  I think we need
 > both, because we use print_source_lines in other contexts too, but I'm
 > not sure about that.
 > 
 > So the short answer is that they are duplicated in the CLI too.  The
 > two copies are just formatted so differently that it isn't obvious.

OK.  I think I understand the second call now.

I was looking specifically at this clause in print_source_lines_base:

  if (desc < 0)
    {
      last_source_error = desc;

      if (!noerror)
	{
	  char *name = alloca (strlen (s->filename) + 100);
	  sprintf (name, "%d\t%s", line, s->filename);
	  print_sys_errmsg (name, errno);
	}
      else
	ui_out_field_int (uiout, "line", line);
      ui_out_text (uiout, "\tin ");
      ui_out_field_string (uiout, "file", s->filename);
      ui_out_text (uiout, "\n");

      return;
    }

I think this code may only be reached when there is duplicated MI output.

 > This does make me wonder about the patch though.  Denis, could you
 > hold off on committing it?  Which duplicate copy are you eliminating?
 > Maybe we should diff two testsuite runs to see what else changes.

No, I think this change is alright (I haven't run the testsuite though.).

It's your decision but the next release is a long way off and if these changes
are not the right ones they are not far off.  So I would suggest committing
them all i.e this one, yours and Denis's -thread-* commands now.  That way
if there are problems we will quite likely discover them before the release.


-- 
Nick                                           http://www.inet.net.nz/~nickrob