Re: Patch: 'show args' -vs- '--args'

[Daniel Jacobowitz <drow@false.org>]

On Sun, Feb 03, 2008 at 01:03:10PM -0700, Tom Tromey wrote:
> I use "gdb --args" a lot, and I've noticed a weird behavior.  If I
> start gdb this way, "show args" will initially say that there are no
> arguments, even if there are.  However, if I run "show args" twice in
> a row, the second one will print the correct answer.
> 
> I tracked this down to notice_args_read.  This calls get_inferior_args
> to set the correct value, but too late.  This is why it fails the
> first time (it uses the old value) but succeeds the second time (the
> first call sets the value).
> 
> This patch fixes the problem by using the (apparently as-yet-unused)
> 'pre_show_hook' of the show command.

This is from your original --args patch as posted on gdb-patches:

@@ -1785,8 +1836,10 @@
   	      "Set argument list to give program being debugged when
	      it is started.\n\
 Follow this command with any number of args, to be passed to the
 program.",
		   &setlist);
-  add_show_from_set (c, &showlist);
   c->completer = filename_completer;
+  c->function.sfunc = notice_args_set;
+  c = add_show_from_set (c, &showlist);
+  c->pre_show_hook = notice_args_read;
 
   c = add_cmd
     ("environment", no_class, environment_info,

It was removed in a typo later and eventually ended up readded as
a show_func rather than a pre_show_hook.

Anyway, now that it's a show_func...

> +notice_args_read (struct cmd_list_element *c)
>  {
> -  deprecated_show_value_hack (file, from_tty, c, value);
>    /* Might compute the value.  */
>    get_inferior_args ();
>  }

wouldn't just swapping those two function calls fix it?


-- 
Daniel Jacobowitz
CodeSourcery