Created attachment 35635 [details]
Snapshot of revised gperfmeter preferences sheet

Created attachment 35636 [details]
Revised gperfmeter.glade from Calum based on initial work from Eugenia.

Note that we should be starting from Eugenia's work 
and then taking the modifications as supplied by Calum.

Further comment from Calum:

  Haven't had time to finish this because I'm having to 
  use glade over vpn today, which is rather impractical... 
  but it'll show you the sort of thing I had mind for the 
  Hosts and Chart tabs. The Metrics tab is probably fine 
  as it is, if you don't want to wait until I've had a 
  look at it.

I've been wondering about the Local/Remote radio buttons. Is 'Local'
superfluous? Could there be one edit field be preceded by 'Hostname'
(gnome-netinfo uses 'Network address', gnome-remote-desktop uses 'Computer',
gnome-remote-shell uses 'Host') where the user could fill in 'localhost' if they
chose to? I think most networking apps behave this way.

(perhaps this should be its own bug, though I thought I'd mention it while there
was a proposal for a new layout)

I think a change like that could be handled when we 
implement bug #162649. 

The localhost idea could work, and indeed presumably does today if you type
'localhost' into the 'remote host' field, but using it does rely on some (albeit
simple) knowledge from the sysadmin domain... for Joe User, I think I'd be
happier to keep the options separate-- it's not exactly the busiest tab in the
dialog anyway :)

With that in mind, as this is supposed to be a tool for everyone it probably
*would* be better to use a term more consistent with one of the other tools Ryan
mentions... possibly "This Computer" and "Remote Computer", although "This" can
get a bit ambiguous in *nix environments as you could easily be running the app
remotely.

I think you're right about how separate options makes things easier for Joe User
which is where GNOME is going.

Fwiw, if someone has sshed to a remote machine and then launched gperfmeter from
the command-line, their literacy probably makes 'This Computer' unamiguous.
(unless they've launched multiple meters and forgot which is which, or launched
the meter so long ago they can't remember. Which is a reminder that  the title
bar and hostname meter label remain identical to whatever is chosen here for the
localhost label, as they are now.)

Evolution uses "On This Computer" to organize local data.

A Sun support forum post seems to indicate that JDS also uses "This Computer":
http://supportforum.sun.com/sjds/index.php?t=msg&goto=2129&rid=0#msg_2129
but you guys would know for sure. ;)

Nautilus and the Hal Device Manger use "Computer"

Created attachment 38518 [details]
Combined new gperfmeter.glade

With Calum's help, I've taken what he provided and merged
it into the gperfmeter.glade file that's in CVS HEAD.

This new gperfmeter.glade has been checked into CVS HEAD.
Currently it's broken though. It's failing with:

...
Gtk-CRITICAL **: file gtkcombobox.c: line 4625: assertion `GTK_IS_LIST_STORE
(combo_box->priv->model)' failed
aborting...
t@1 (l@1) signal ABRT (Abort) in _lwp_kill at 0xfdebd5ec
0xfdebd5ec: _lwp_kill+0x0008:           bad opcode
Current function is g_logv
  488                 abort ();
t@1 l@1 <2> where 10                                                         
current thread: t@1
  [1] _lwp_kill(0x0, 0x6, 0x839a4, 0xfdeb6238, 0xfdeebb20, 0x0), at 0xfdebd5ec 
  [2] raise(0x6, 0x6, 0x5, 0x6, 0x32684, 0x0), at 0xfde5d884 
  [3] abort(0xffbfcf08, 0xa, 0xfe0608c0, 0xa6544, 0x0, 0xfdee4278), at 0xfde3ddf8 
=>[4] g_logv(log_domain = 0xfecbb030 "Gtk", log_level = G_LOG_LEVEL_CRITICAL,
format = 0xfecbb034 "file %s: line %d: 
assertion `%s' failed", args1 = 0xffbfd490), line 488 in "gmessages.c"
  [5] g_log(log_domain = 0xfecbb030 "Gtk", log_level = G_LOG_LEVEL_CRITICAL,
format = 0xfecbb034 "file %s: line %d: 
assertion `%s' failed", ...), line 507 in "gmessages.c"
  [6] gtk_combo_box_append_text(combo_box = 0x2265f8, text = 0x285978
"jurassic"), line 4625 in "gtkcombobox.c"
  [7] set_remote_host_list(combo_box = 0x2265f8, list = 0x69aa8), line 319 in
"props.c"
  [8] CreateProps(), line 374 in "props.c"
  [9] PmShowProps(), line 808 in "props.c"
  [10] menu_handler(widget = 0x163220), line 771 in "perfmeter.c"
...

This is in the set_remote_host_list() routine in props.c:

static void
set_remote_host_list(GtkWidget *combo_box, GList *list)
{
    gint i, len;
    gchar *str;

    if (list != NULL) {
        len = g_list_length(list);

        for (i = 0; i < len; i++) {
            str = g_strdup((gchar *) g_list_nth_data(list, i));

            gtk_combo_box_append_text(GTK_COMBO_BOX(combo_box),  str);  <===
FAILING HERE.
        }
    }
}

I'll continue to investigate.

Hopefully all fixed now. Changes checked into CVS HEAD.
I've bumped the version number in configure.in to 2.0.43.

Several problems:
  - One of the Glade attributes for the Hosts_Combo wasn't being set.
  - The LogFile_Entry widget wasn't connected up.
  - All the accessibility information in the Glade file had been lost.
  - Needed to check for a null widget being passed into the SetSensitive
    function in props.c.