[mcclim-devel] with-output-as-gadget, again

[Max-Gerd Retzlaff]

Hello,

I've tried to get with-output-as-gadget working but I was not
successful. The problem being that I have not understood recording and
several further parts of CLIM, yet.  In this mail I want to present my
observation about the problem, in the hope that someone with more
experience and knowledge might see a solution.  I would really like to
have it working for a file selector that I need for several projects.

Christophe reported the problem already in July:
http://common-lisp.net/pipermail/mcclim-devel/2005-July/004100.html
You might want to read that mail first.


Basically the panes of the whole gadget are only adopted to the
stream's sheet (stream being the stream that is given as an argument
to the call of WITH-OUTPUT-AS-GADGET) when the output-record of the
sheet is replayed (or what is the right term?). The adoption happens
via note-output-record-got-sheet method calls that are made when
output-records about "parent/child-relationships" are processed;
there is specialization on (record gadget-output-record) for that
method, in which the adoption takes place.

The problem is that the output methods of the clx-backend
(window-clear-area (called from window-clear), medium-draw-text*
(called from write-string), .. ) try to access the
native-transformation (and later native-region) of the gadget's panes
quite early, and that does not work if the panes are not yet adopted.

Either the adoption has to happen earlier, or the clx output method
have to be delayed via output-records as well. I suspect that the clx
backend needs an general overhaul to make it be (more)
"output-recording aware". But perhaps a single, deliberately
inserted line would work a miracle..


Consider Christophe's example:

 (clim:with-output-as-gadget (*standard-output*)
    (clim:with-look-and-feel-realization
        ((clim:frame-manager clim:*application-frame*)
         clim:*application-frame*)
      (clim:labelling (:label "Hello")
        (clim:make-pane 
         'clim:text-field
          :activate-callback (lambda (g) (princ (clim:gadget-value g)))))))

The flow to the error NO-APPLICABLE-METHOD on sheet-native-transformation when
called with an argument of NIL is:

In WITH-OUTPUT-AS-GADGET a instance of 'GADGET-OUTPUT-TYPE is created
(with the evaluated body of the call as its :child argument).

There is a initialize-instance :after method for instances of this
class, in which ALLOCATE-SPACE for the child (the label-pane) is
called.

ALLOCATE-SPACE calls MOVE-SHEET for the first child of the label-pane
(the text-field-pane).

MOVE-SHEET SETFs the SHEET-TRANSFORMATION of the text-field-pane.

There is a method (setf sheet-transformation) :after (transformation
(sheet basic-sheet)). It calls (note-sheet-transformation-changed
sheet), which calls UPDATE-MIRROR-GEOMETRY.

Inside UPDATE-MIRROR-GEOMETRY (sheet-native-transformation
(sheet-parent sheet)) is called, but as the sheet is not adopted yet,
that gives nil and sheet-native-transformation is not defined for nil.

=> NO-APPLICABLE-METHOD on sheet-native-transformation when called
   with an argument of NIL


It is quite the same for simpler examples where the WRITE-STRING is
called with a pane of the gadget as output stream. There it is:
WRITE-STRING -> %WRITE-STRING -> .. -> MEDIUM-DRAW-TEXT* ->
(sheet-native-transformation (medium-sheet medium))
Inside sheet-native-transformation the same function is called again
for the sheet-parent. As the sheet is not adopted yet, that gives nil
and sheet-native-transformation is not defined for nil.


Naively, I tried something like:

(let ((my-stream (find-pane-named *application-frame* 'clim-listener::interactor)))
    (replay (STREAM-CURRENT-OUTPUT-RECORD my-stream) my-stream))

right after WITH-OUTPUT-AS-GADGET and before the WRITE-STRING, in the
hope that this would force recorded adoption records to take place
earlier. Is this completely wrong, should it have been
replay-output-record instead? Well, it would only be an ugly workaround
at it doesn't help against the problems with labelling and scrolling.


I could confirm my assumption that the delayed adoption is the problem
as with neatly placed calls to SHEET-ADOPT-CHILD and SHEET-DISOWN-CHILD
I got my file-selector kind of running (only the first ACCEPTING-VALUES
iteration draws correctly and there is no scrolling for the embedded
pane:
         http://bl0rg.net/~mgr/flux/file-selector-dialog.png

(In the end it should become a accept method for pathname
presentations in gadget-dialog-view.)


There are additional problem when the stream for WITH-OUTPUT-AS-GADGET
is an ACCEPTING-VALUES-STREAM:

When attempting to read the slot's value (slot-value), the slot
CLIM-INTERNALS::HEIGHT is missing from the object
#<CLIM-INTERNALS::ACCEPTING-VALUES-STREAM {92BF441}>.
   [Condition of type SIMPLE-ERROR]

It would be especially nice to get this running as well..


Regards,
Max

-- 
Max-Gerd Retzlaff <m.retzlaff at gmx.net>

For your amusement:
I don't make the rules, Gil, I only play the game.
                -- Cash McCall

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <https://mailman.common-lisp.net/pipermail/mcclim-devel/attachments/20050826/fbd2d0e0/attachment.sig>