[ltk-user] The cutting edge at http://ltk.rplay.net/svn/branches/ltk/repl/

[Peter Herth]

Hi Jussi,

thank you very much for your thorough testing of LTk. This really
helps getting the
little bugs worked out, especially as I have no access to a Vista machine to
perform testing.

> I took a look at the forthcoming version of LTK at the address above. I
> understand that this is work-in-progress
> and that some of my observations may be the result of how Windows or Tk
> behaves so that nothing can be done.
> I also know almost nothing of Tk, very little of LTK and Lisp are a bit
> rusty, so bear with me...
>
>
> 1. tktable.lisp
>
> When I did (compile-file "tktable") I got the following warning:
>      ;   In an anonymous lambda form at position 2382: Initarg :ROWS occurs
> in both ROWS and COLS slots
> I guess this is a valid error message and the code should be corrected like
> this (the original row is commented out):
> (defclass scrolled-table (frame)
>   ((table :accessor table)
>    (hscroll :accessor hscroll)
>    (vscroll :accessor vscroll)
>    (rows :accessor rows :initarg :rows :initform nil)
>    ;(cols :accessor cols :initarg :rows :initform nil)
>    (cols :accessor cols :initarg :cols :initform nil)
>    (titlerows :accessor titlerows :initarg :titlerows :initform nil)
>    (titlecols :accessor titlecols :initarg :titlecols :initform nil)
>    (data :accessor data :initarg :data :initform nil)
>    ))

Interesting, sbcl compiles this without complaining... using
the same initarg twice is a bit silly of course...


>
> When I run (tabletest) I get a neat-looking table with 200 rows and 40
> columns. The cells have light-gray (?) background
> and the foreground color is black (?). Very clear, no strain to the eye.
>
> But when I click a cell with the mouse, the background first changes to
> blue(?) then changes back to
> the original light-gray (?) and the foreground color changes to white (?).
> This combination
> of colors makes it very hard to see what the cell contents are.
>
> Scrolling by dragging the thumbs OK. Vertical scroll with either PgDn-key or
> by clicking the vertical scroll area below
> the thumb brings new rows into view but leaves one row out. The reason (as
> an example) is this: when the table is
> at first displayed it shows the first 34 rows in whole and a tiny slice of
> the 35th row also. The system "thinks" that
> the row 35 was shown although it was not wholly visible and starts showing
> the next rows from the 36th row onward.
>
> If I squeeze the tiny slice out by diminishing the window height a little
> the vertical scroll starts to function correctly. It would
> be nice if Ltk or Tk could squeeze the slice out instead.
>
> Still on the vertical scroll: If in the starting view (i.e. the rows 1-34
> showing) I click the cell of (row 9, col 5), for instance,
> to get it focused and then do a vertical scroll of one pagefull and scroll
> back (PgDn and PgUp, for instance), the focus
> (i.e. the text caret) is in the cell (row 1, col 1) and the cell (row 9, col
> 5) is displayed with a blue background.
>
>
> 2. ltk.lisp
>
> THE TEST i.e. (ltktest) OK as far as I understand. The Tk styles for Windows
> (winnative, xpnative) work beautifully.
> As the program starts the default selection of the radio button is on and
> the theme is xpnative as it should be on Vista.
> If, however, I choose another theme and then select the default again, the
> theme is not xpnative but Motif I think.
>
> EDIT: It later occurred to me that maybe "default" here means "Tk default"
> and not "Windows XP/Vista default" but still
> the behavior is not consistent.

Probably my radio button handling in ltktest is not correct....

>
> (ltkeyes) OK. All other tests except defwidget-test work. Running
> (defwidget-test) gives the error msg
> "Class named TW not found.".
>

yes, this should be test-widget instead

>
> 3. ltk-tile.lisp
>
> Running (tile-test) gives several times the error 'Tcl/Tk error: invalid
> command name "style" '. After adding
> "ttk::" to the beginning of "style" like this:
>
> (defun theme-names ()
>   (send-wish "senddatastrings [ttk::style theme names]")
>   (ltk::read-data))
>
> (defun use-theme(name)
>   (format-wish "ttk::style theme use ~a" name))
>
> the program runs OK except that it has the same problem with the default
> theme as ltktest above.

As tile is now part of ltk, ltk-tile can be removed completely.

>
>
> 4. ltk-mw.lisp
>
> Compiling gives:
>      ((SETF DATA) :AFTER (T LIST-SELECT)): Unused lexical variable VAL
>
> Apparently it concerns the last method in the list-select box widget:
>
> (defmethod (setf data) :after (val (select list-select))
>   (listbox-clear select)
>   (listbox-append select (mapcar (lambda (item)
>
>      (list-select-display select item))
>                                  (data select))))
>

Again a difference between sbcl and clozure, but easy to fix

> The test (tooltip-test) functions, sort-of. The tooltip is shown but at the
> screen position (0,0)!
> While debugging by spreading prints around I by accident discovered a silly
> way of correcting
> the behavior. This is how the last lines of the tooltip method show look
> like now when it
> is operating correctly:
>    (when (and txt (> (length txt) 0))
>       (setf (text (tooltip-label tooltip)) txt)
>       ;(print (geometry tooltip ))
>       (set-geometry-xy tooltip (truncate x)  (truncate y))
>       ;(print (geometry tooltip ))
>       (geometry tooltip )
>       (normalize tooltip)
>       ;(print (geometry tooltip ))
>       (raise tooltip))))
>
> So the extra call to geometry does the trick and naturally I have no idea
> what is going on here.

You are not alone there :). But I will look into it to find out what
is the problem...

>
> Running (gtree-demo) produces errors about :borderwidth and height in:
>      (defmethod gtree-render-node ((d gtree-demo) node )
>          (make-instance 'label :master d :text node :borderwidth 3 :relief
> :raised :background :grey :height 1 :width 10))
> After the removal of these the run succeeds.

The tile based widgets which LTk now uses by default do not support many
of the old configuration options, so they should not be used anymore as in
this case.

Thanks for taking the time and reporting your test results so
thoroughly, especially
as you are using a different OS and Lisp than I do, so many things come up which
never show in my daily work :)

Peter