May 2005 - mcclim-cvs@common-lisp.net - mailman3.common-lisp.net

[mcclim-cvs] CVS update: mcclim/Backends/beagle/README.txt mcclim/Backends/beagle/load-beagle.lisp mcclim/Backends/beagle/package.lisp
by drose＠common-lisp.net 17 May '05

17 May '05

Update of /project/mcclim/cvsroot/mcclim/Backends/beagle In directory common-lisp.net:/tmp/cvs-serv28707/beagle Modified Files: README.txt load-beagle.lisp package.lisp Log Message: Removed *DEFAULT-BEAGLE-FRAME-MANAGER*. Use *DEFAULT-FRAME-MANAGER* instead. Updated README accordingly. Date: Tue May 17 22:26:37 2005 Author: drose Index: mcclim/Backends/beagle/README.txt diff -u mcclim/Backends/beagle/README.txt:1.9 mcclim/Backends/beagle/README.txt:1.10 --- mcclim/Backends/beagle/README.txt:1.9 Tue May 17 22:12:36 2005 +++ mcclim/Backends/beagle/README.txt Tue May 17 22:26:37 2005 @@ -5,7 +5,7 @@ . README . INSTALLATION . CONFIGURATION - . default frame manager + . frame manager . multiple ports . KNOWN LIMITATIONS / TODO LIST . FIXED STUFF PREVIOUSLY ON THE KNOWN LIMITATIONS / TODO LIST @@ -85,18 +85,17 @@ CONFIGURATION -default frame manager: ----------------------- +frame manager: +-------------- The Beagle back end defines two frame manager objects; one is the aqua look and feel, the other is the 'standard' McCLIM look and feel. If you want to configure a specific frame manager to be used, set the following parameter:- -BEAGLE::*DEFAULT-BEAGLE-FRAME-MANAGER* <defined in 'port.lisp'> - -> 'beagle::beagle-aqua-frame-manager [default] - -> 'beagle::beagle-standard-frame-manager +CLIM:*DEFAULT-FRAME-MANAGER* + -> 'beagle:beagle-aqua-frame-manager [default] + -> 'beagle:beagle-standard-frame-manager -Should use CLIM:*DEFAULT-FRAME-MANAGER* for this! Note that as yet, no native (aqua) look and feel panes have been defined, so it doesn't matter which frame manager you use. @@ -227,10 +226,6 @@ things got before blowing up) but now it just looks messy. -17. *BEAGLE-DEFAULT-FRAME-MANAGER* should be replaced with the standard - *DEFAULT-FRAME-MANAGER* instead. - - 19. Menus don't work in CLIM-FIG (or anywhere else!). No idea why not... This is because (I think) the menu popups don't operate in a flipped coord system (unlike NSViews). [Command menu that is drawn across @@ -358,6 +353,11 @@ allocated. Some are stack-allocated and cause errors about 'bogus' objects once they go out of scope. At least, I think (and hope) that's the reason 'cause that's easy to fix. RESOLVED 17.JUL.04 + +-17.- *BEAGLE-DEFAULT-FRAME-MANAGER* should be replaced with the standard + *DEFAULT-FRAME-MANAGER* instead. + FIXED 17.MAY.2005 - *beagle-default-frame-manager* is no more (well... + it's still there but it can be ignored to all intents and purposes). -18.- Note about force-quit; appended to (5). Index: mcclim/Backends/beagle/load-beagle.lisp diff -u mcclim/Backends/beagle/load-beagle.lisp:1.2 mcclim/Backends/beagle/load-beagle.lisp:1.3 --- mcclim/Backends/beagle/load-beagle.lisp:1.2 Tue May 17 00:13:08 2005 +++ mcclim/Backends/beagle/load-beagle.lisp Tue May 17 22:26:37 2005 @@ -7,5 +7,6 @@ ;;; want 'beagle::beagle-aqua-frame-manager, you don't need to set this since that ;;; is the default). ;;;(setf beagle::*default-beagle-frame-manager* 'beagle::beagle-standard-frame-manager) -(setf beagle::*default-beagle-frame-manager* 'beagle::beagle-aqua-frame-manager) +;;;(setf beagle::*default-beagle-frame-manager* 'beagle::beagle-aqua-frame-manager) +(setf clim:*default-frame-manager* 'beagle:beagle-aqua-frame-manager) (format t "~%Done.~%") Index: mcclim/Backends/beagle/package.lisp diff -u mcclim/Backends/beagle/package.lisp:1.3 mcclim/Backends/beagle/package.lisp:1.4 --- mcclim/Backends/beagle/package.lisp:1.3 Tue May 17 00:13:09 2005 +++ mcclim/Backends/beagle/package.lisp Tue May 17 22:26:37 2005 @@ -112,5 +112,7 @@ #:send-super #:slet #:with-cstrs - #:with-nsstr)) + #:with-nsstr) + (:export #:beagle-standard-frame-manager + #:beagle-aqua-frame-manager))

1 0

[mcclim-cvs] CVS update: mcclim/Backends/beagle/native/lisp-window-delegate.lisp
by drose＠common-lisp.net 17 May '05

17 May '05

Update of /project/mcclim/cvsroot/mcclim/Backends/beagle/native In directory common-lisp.net:/tmp/cvs-serv27747/beagle/native Modified Files: lisp-window-delegate.lisp Log Message: Fixed annoying (and long term) issue with key focus. Focus is now (or at least, appears to now be) set correctly. Updated README.txt accordingly. Date: Tue May 17 22:12:38 2005 Author: drose Index: mcclim/Backends/beagle/native/lisp-window-delegate.lisp diff -u mcclim/Backends/beagle/native/lisp-window-delegate.lisp:1.1 mcclim/Backends/beagle/native/lisp-window-delegate.lisp:1.2 --- mcclim/Backends/beagle/native/lisp-window-delegate.lisp:1.1 Tue May 17 00:13:17 2005 +++ mcclim/Backends/beagle/native/lisp-window-delegate.lisp Tue May 17 22:12:37 2005 @@ -47,20 +47,6 @@ ;;; in the events.lisp file. ;;; -#|| - -(define-objc-method ((:void :window-did-expose notification) lisp-window-delegate) - (nslog (format nil "LISP-WINDOW-DELEGATE: received windowDidExpose event")) - (nslog (format nil "Notification name is: ~A" (description (send notification 'name)))) - ;; Window extents should be non-nil in this case... We can get an NSExposedRect out of the userInfo - ;; dictionary in the notification and pass this back to CLIM. Should speed things up a little... - #+nyi (slet ((rect (send (send notification 'object) 'frame))) - (add-notification-to-queue (send notification 'object) notification - (pref rect :<NSR>ect.origin.x) - (pref rect :<NSR>ect.origin.y) - (pref rect :<NSR>ect.size.width) - (pref rect :<NSR>ect.size.height)))) -||# (define-objc-method ((:void :window-did-resize notification) lisp-window-delegate) ;;; (nslog (format nil "LISP-WINDOW-DELEGATE: received windowDidResize event")) @@ -73,7 +59,14 @@ (pref rect :<NSR>ect.size.width) (pref rect :<NSR>ect.size.height)))) +(define-objc-method ((:void :window-did-become-key notification) lisp-window-delegate) + (add-notification-to-queue (send notification 'object) notification)) + +;;; No notifications below this point that we're interested in. #|| + +;;; ... although this one might be useful. Especially wrt native cut&paste. + (define-objc-method ((:void :window-will-close notification) lisp-window-delegate) (nslog (format nil "LISP-WINDOW-DELEGATE: received windowWillClose event")) (nslog (format nil "Notification name is: ~A" (description (send notification 'name)))) @@ -86,9 +79,6 @@ ||# -(define-objc-method ((:void :window-did-become-key notification) lisp-window-delegate) - (nslog (format nil "LISP-WINDOW-DELEGATE: received windowDidBecomeKey event")) - (add-notification-to-queue (send notification 'object) notification)) #|| (define-objc-method ((:void :window-did-change-screen notification) lisp-window-delegate) @@ -148,6 +138,20 @@ ;;; (nslog (format nil "LISP-WINDOW-DELEGATE: received windowDidWillResize event")) ;;; size) ;;; (send-super :window-will-resize ns-window :to-size size)) + + + +;;;(define-objc-method ((:void :window-did-expose notification) lisp-window-delegate) +;;; (nslog (format nil "LISP-WINDOW-DELEGATE: received windowDidExpose event")) +;;; (nslog (format nil "Notification name is: ~A" (description (send notification 'name)))) +;;; ;; Window extents should be non-nil in this case... We can get an NSExposedRect out of the userInfo +;;; ;; dictionary in the notification and pass this back to CLIM. Should speed things up a little... +;;; #+nyi (slet ((rect (send (send notification 'object) 'frame))) +;;; (add-notification-to-queue (send notification 'object) notification +;;; (pref rect :<NSR>ect.origin.x) +;;; (pref rect :<NSR>ect.origin.y) +;;; (pref rect :<NSR>ect.size.width) +;;; (pref rect :<NSR>ect.size.height))))

1 0

[mcclim-cvs] CVS update: mcclim/Backends/beagle/input/events.lisp
by drose＠common-lisp.net 17 May '05

17 May '05

Update of /project/mcclim/cvsroot/mcclim/Backends/beagle/input In directory common-lisp.net:/tmp/cvs-serv27747/beagle/input Modified Files: events.lisp Log Message: Fixed annoying (and long term) issue with key focus. Focus is now (or at least, appears to now be) set correctly. Updated README.txt accordingly. Date: Tue May 17 22:12:37 2005 Author: drose Index: mcclim/Backends/beagle/input/events.lisp diff -u mcclim/Backends/beagle/input/events.lisp:1.2 mcclim/Backends/beagle/input/events.lisp:1.3 --- mcclim/Backends/beagle/input/events.lisp:1.2 Tue May 17 19:51:14 2005 +++ mcclim/Backends/beagle/input/events.lisp Tue May 17 22:12:37 2005 @@ -28,7 +28,7 @@ #|| -$Id: events.lisp,v 1.2 2005/05/17 17:51:14 drose Exp $ +$Id: events.lisp,v 1.3 2005/05/17 20:12:37 drose Exp $ All these are copied pretty much from CLX/port.lisp @@ -302,24 +302,17 @@ (sheet (%beagle-port-lookup-sheet-for-view *beagle-port* (send window 'content-view)))) ;; We don't get exposure notifications when the window has a (Cocoa) backing store. (cond - ;; I'm not sure this is right; we need to find the sheet in the hierarchy of this window - ;; that can accept key events. Can there be multiple sheets this counts for? I'd guess - ;; so, in which case how do we know which one to set as port key focus? Suspect we can - ;; only know this by tracking the last view that can be key that received a mouse down? - ;; *Should* be able to do this by doing a :hit-test on the content view of the receiving - ;; window after manually grabbing the pointer coordinates. ((send (send notification 'name) :is-equal-to-string #@"NSWindowDidBecomeKeyNotification") (setf return-event nil) (when (send window 'is-visible) ; only do if window is on-screen... - ;; NB: this isn't in the right coordinate system! Convert screen -> content view coords. - (slet ((pointer-loc (send (@class ns-event) 'mouse-location)) - (loc-in-window (send window :convert-screen-to-base pointer-loc))) - (let* ((content-view (send window 'content-view)) - (target-view (send content-view :hit-test loc-in-window)) - (target-sheet (%beagle-port-lookup-sheet-for-view *beagle-port* target-view))) - (unless (null target-sheet) - (format *debug-io* "Setting focus in *beagle-port* onto (hopefully correct) sheet: ~S~%" target-sheet) - (%set-port-keyboard-focus target-sheet *beagle-port*)))))) + (let* ((content-view (send window 'content-view)) + (target-sheet (%beagle-port-lookup-sheet-for-view *beagle-port* content-view)) + (frame (pane-frame target-sheet)) + ;; Works out which sheet *should* be the focus, not which is currently... + ;; or at least, so I think. + (focus (climi::keyboard-input-focus frame))) + (unless (null target-sheet) + (setf (port-keyboard-input-focus *beagle-port*) focus))))) ((send (send notification 'name) :is-equal-to-string #@"NSWindowDidExposeNotification") (setf return-event (make-instance 'window-repaint-event :timestamp (incf timestamp) @@ -703,19 +696,28 @@ ;;; (oops, we lose the timestamp here.) ;;; Cocoa note: the Frame (NSWindow) must be made key for us to receive events; but they -;;; must then be sent to the Sheet that has focus. +;;; must then be sent to the Sheet that has focus. Whilst there are Cocoa mechanisms to +;;; do this, it's probably best to let CLIM decide on the appropriate sheet and we just +;;; send all key events to it. -;;; NB. when the method was renamed it appears that the argument order was also changed. (defmethod %set-port-keyboard-focus ((port beagle-port) focus &key timestamp) (declare (ignore timestamp)) - (let ((mirror (sheet-mirror focus))) - (when mirror - (let ((window (send mirror 'window))) - (when window - (setf (beagle-port-key-focus port) focus) - (if (send window 'is-key-window) - (send window :order-front nil) - (send window :make-key-and-order-front nil))))))) + (if (eq (beagle-port-key-focus port) focus) + (format *trace-output* "Attempt to set keyboard focus on sheet ~a which already has focus.~%" + focus) + (let ((mirror (sheet-mirror focus))) + (if (null mirror) + (format *trace-output* "Attempt to set keyboard focus on sheet ~a which has no mirror!~%" + focus) + (let ((window (send mirror 'window))) + (if (eql window (%null-ptr)) + (format *trace-output* "Attempt to set keyboard focus on sheet ~a with no NSWindow!~%" + focus) + (progn + (setf (beagle-port-key-focus port) focus) + (unless (send window 'is-key-window) + (send window 'make-key-window))))))))) + ;;; Not sure we need to do this... apparently we do. I have stopped flushing ;;; the window after every drawing op, and now things don't get output

1 0

[mcclim-cvs] CVS update: mcclim/Backends/beagle/README.txt
by drose＠common-lisp.net 17 May '05

17 May '05

Update of /project/mcclim/cvsroot/mcclim/Backends/beagle In directory common-lisp.net:/tmp/cvs-serv27747/beagle Modified Files: README.txt Log Message: Fixed annoying (and long term) issue with key focus. Focus is now (or at least, appears to now be) set correctly. Updated README.txt accordingly. Date: Tue May 17 22:12:36 2005 Author: drose Index: mcclim/Backends/beagle/README.txt diff -u mcclim/Backends/beagle/README.txt:1.8 mcclim/Backends/beagle/README.txt:1.9 --- mcclim/Backends/beagle/README.txt:1.8 Tue May 17 00:13:08 2005 +++ mcclim/Backends/beagle/README.txt Tue May 17 22:12:36 2005 @@ -5,10 +5,10 @@ . README . INSTALLATION . CONFIGURATION - . debug . default frame manager - . listener + . multiple ports . KNOWN LIMITATIONS / TODO LIST + . FIXED STUFF PREVIOUSLY ON THE KNOWN LIMITATIONS / TODO LIST . WISH LIST . APPLICATIONS @@ -76,7 +76,7 @@ loop. Note #3: If you'd rather run with the CLX back end, load CLX - instead here. It is possible to with multiple ports simultaneously + instead here. It is possible to run multiple ports simultaneously so that both a CLX and a Beagle Listener can be run side by side for comparative purposes (or just because the Listener is actually usable for something useful when running under CLX). @@ -137,11 +137,18 @@ large output history. Paolo's stress test takes 26 seconds and conses 16MB on my (admittedly slow) iMac compared to 1.5 seconds on a 2.4GHz Pentium IV and unknown (to me) consing. - Should be able to speed things up by performing fewer focus lock / unlocks, - and by not setting drawing options unless necessary. I don't know how - far this will get us though... - UPDATE - 21.AUG.2004 - performing fewer NSWindow flushes makes no difference - to speed. + + NB: the number of subclasses of T when running Beagle is about twice the + number when running CLX (because the Cocoa bridge introduces every + Cocoa object into the Lisp image, and even just the core + appkit + frameworks are large). Additionally, I think allocated Cocoa objects + are included in the cons measurement of the Lisp image which accounts + for a chunk of the memory usage. + + Should be able to speed things up by not setting drawing options unless + necessary. A hammer-like approach to do this has been implemented but could + do with a little finessing. + UPDATE - 25.APR.2005 - When the mirror transformation is set (sheet is scrolling) we dispatch a repaint on the 'untransformed mirror region' (using the mirror transformation as the 'untransformation') @@ -153,33 +160,30 @@ no redraw happens from CLIM generally. Also, Cocoa appears to get really slow at rendering text when the output history gets too large (maybe this is CLIM again, it's hard to know). Need to profile. - TODO: cache lisp-bezier-path instances and reuse them. Use approximation for - text sizing (there's as much overhead working out how big rendered text - would be as there is to rendering the text itself, or almost). -2. When running the Listener (and probably other applications), the resize + TODO: Use approximation for text sizing (there's as much overhead working out + how big rendered text would be as there is to rendering the text itself, + or almost). + + TODO: Use pixmap for mirror contents and use blitting to copy areas? + + +2. When running the Listener (and other applications), the resize handle is not visible; it's there, but you can't see it. Grab and drag with faith and it should work anyway. + 3. There are not yet any aqua look and feel panes. Sorry, I'm trying to get everything else working first! + 5. Mouse down / up on buttons appears not to work very well unless the frame containing the buttons is the only active frame. Actually, this ^^^ seems to work fine, but the highlighting for button gadgets looks screwy under OS X. + (Think there is a problem with tracking rectangles not being set for + panes. Investigation needed.) -6. Swapping between key windows (the window accepting the keyboard input) - is a little flakey; as an example, if a second Listener is started from - the first, clicking between the windows transfers key focus (as - expected). However, if the first is then 'Exit'ed, the second will not - get the key focus until some other (non-McCLIM) window has been given - the keyboard focus first (i.e. click on the OpenMCL Listener window, - then back on the McCLIM Listener window). - Additionally, clicking on a scroll-bar (for example) makes the window - key, so clicking on a view that accepts keyboard input (interactor) - within this window won't then allow keyboard input. - We should stop scroll-bars being able to get keyboard input... 7. Keyboard events are not handled "properly" as far as any OS X user will be concerned; only the ASCII characters are recognised, along with @@ -187,16 +191,26 @@ Emacs-like CTRL-B (backward char), CTRL-F (forward char), CTRL-A (start of line), CTRL-E (end of line), CTRL-D (delete char). + 8. There is no +flipping-ink+ implementation. Anything drawn in flipping ink shows up bright red with about 50% transparency (which is why the cursor looks so strange). + NB: moving to 10.4 (Tiger) would fix this since there's an XOR mode + on NSBezierPath in that OS version. So too would implementing pixmaps + for mirrors, since we can do XOR image compositing. My gut feel is I'd + like to avoid having pixmap caches for mirrors though. + + 9. Foreign memory management is non-existent. This is fine whilst running in the same thread as the OpenMCL Listener (since we use its autorelease pool) but when running in a separate thread lots of warning messages - are generated. + are generated. You might find it necessary to force-quit OpenMCL after + you've finished. + + +12. There's some debug output remaining in some corner cases. -12. There's probably some debug output remaining in some corner cases. 15. Popup menus don't work quite the same way as they do in the CLX back end. Cocoa doesn't support pointer grabbing so disposing of menus when @@ -204,23 +218,27 @@ a command, or mouse-click on a different window to get rid of them). Additionally, highlighting the menu item the mouse is currently over is rather intermittent, although the correct menu item appears to always - be chosen on mouse-click. + be chosen on mouse-click (related to the same tracking rectangle issue + mentioned in (5)?). + 16. Windows are put on screen very early in the realization process which wasn't a bad thing during early development (could see how far through things got before blowing up) but now it just looks messy. + 17. *BEAGLE-DEFAULT-FRAME-MANAGER* should be replaced with the standard *DEFAULT-FRAME-MANAGER* instead. -18. The back end doesn't clear up after itself very well. You might find it - necessary to force-quit OpenMCL after you've finished. 19. Menus don't work in CLIM-FIG (or anywhere else!). No idea why not... This is because (I think) the menu popups don't operate in a flipped - coord system (unlike NSViews). + coord system (unlike NSViews). [Command menu that is drawn across + top of window has 'child menus' drawn in bottom-left corner of screen) + TODO: make use of graft native transformation to flip coords rather - than the NSView 'isFlipped' method. + than the NSView 'isFlipped' method? + 20. Bounding rectangles are slightly off (this can be seen in CLIM-FIG again). It's only a matter of a pixel, maybe 2 in the worst case I've seen. @@ -228,10 +246,12 @@ int -> float conversion and ordinate manipulation (in cocoa 0.0, 0.0 falls 'between pixels' - 0.5, 0.5 is 'center of pixel'). + 21. Highlighting on mouse overs isn't quite right; artefacts are left on the display after the mouse has moved out of the target object bounding rectangle (most easily visible in CLIM-FIG again). + 22. Sending key-down / key-up events for modifiers-changed events doesn't look to help get the pointer documentation pane to show the correct prompt. For example, in the Listener, issue a 'help commands' and @@ -242,6 +262,7 @@ Presentation'. Need to check CLX implementation to see if this is the same... + 23. Large output histories: the transformations and geometry calculations go wrong when the output takes up more than 2^16 pixels; the medium should be used to account for this (it does in CLX) but for some @@ -250,6 +271,25 @@ but this will fail eventually (i.e. with a large enough output history), so it needs sorting properly. + +24. Testing + further work on patterns and stencils. + + +25. Bounding rects for commands entered in interactor panes are way out; + looks like the baseline of the text is being used as the bottom of + the bounding rect! + + +26. Minimising frames and then restoring them leads to the frame not + being drawn properly; it looks like 'drawRect' is invoked (as + expected), but nothing to tell McCLIM to redraw the whole frame. + Suspect a notification is sent... needs investigation. + +%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% + +FIXED STUFF PREVIOUSLY ON THE KNOWN LIMITATIONS / TODO LIST + + -4.- Pixmap support is not implemented; this means clim-fig drawing doesn't work. This is getting there, although not very efficiently; we are missing a @@ -264,6 +304,20 @@ hacked for non-tiled patterns, not looked at for stencils). +-6.- Swapping between key windows (the window accepting the keyboard input) + is a little flakey; as an example, if a second Listener is started from + the first, clicking between the windows transfers key focus (as + expected). However, if the first is then 'Exit'ed, the second will not + get the key focus until some other (non-McCLIM) window has been given + the keyboard focus first (i.e. click on the OpenMCL Listener window, + then back on the McCLIM Listener window). + Additionally, clicking on a scroll-bar (for example) makes the window + key, so clicking on a view that accepts keyboard input (interactor) + within this window won't then allow keyboard input. + We should stop scroll-bars being able to get keyboard input... + FIXED 17.MAY.2005 DAR - the problem was we never invoked + 'setf (port-keyboard-input-focus' in handling + the 'did become key' notification. -10.- Text sizes aren't calculated correctly; when multiple lines are output together, the bottom of one line can be overwritten by the top of the @@ -277,12 +331,17 @@ Perhaps Cocoa thinks the dirty region includes that text or something. It's annoying whatever. Still, I'm going to mark this as fixed for now and maybe will come back to it later. + TODO: I think this is to do with the way width and height (rather than text-size) is used to calculate bounding rectangles might be wrong (i.e. getting the wrong information from Beagle). --11.- Line dash patterns haven't been implemented. + TODO: Also note that when a graph is output, there's no (significant) + problem with bounding rects etc. I suspect Beagle may be drawing + with y-align :bottom when CLIM is expecting :baseline, or + something. +-11.- Line dash patterns haven't been implemented. -13.- Some Apropos cases fail; for example 'Apropos graft' fails (although '(apropos 'graft)' does not). The same problem prevents the address @@ -291,13 +350,16 @@ this is happening, but it should be possible to track it down]. RESOLVED 21.AUG.2004 - it appears MACPTRs are output using the family (for a text style) of :fixed - which didn't exist (only :fix). Not - sure if this is a specification violation or not... + sure if this is a specification violation or not... both are bound + in McCLIM, so suspect :fixed is added for compatability with one of + the vendor CLIMs? -14.- Not all foreign objects we keep hold of in the back end are heap- allocated. Some are stack-allocated and cause errors about 'bogus' objects once they go out of scope. At least, I think (and hope) that's the reason 'cause that's easy to fix. RESOLVED 17.JUL.04 +-18.- Note about force-quit; appended to (5). %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% @@ -315,16 +377,19 @@ 4. Look again at the build process and reintegrate with Bosco. -5. Possibly migrate to being a Carbon, rather then a Cocoa, application to - remove OpenMCL version dependencies. +-5.- Possibly migrate to being a Carbon, rather then a Cocoa, application to + remove OpenMCL version dependencies. *Don't bother doing this. Note that + in 10.4, Cocoa apps take advantage of GPU caching (performed by the OS), + but Carbon apps do not. It's possible in some future OS version that + Cocoa drawing will actually be faster than Carbon drawing.* -6. Reduce focus locking in NSViews (I think this will give a not +-6.- Reduce focus locking in NSViews (I think this will give a not insignificant speed increase). 7. Documentation 8. Code tidying, and lots of it! Refactoring. Need to implement many - abstractions (which should also help in the Cocoa -> Carbon move, + abstractions. (which should also help in the Cocoa -> Carbon move, when it happens). 9. Release resources on exit. @@ -335,7 +400,10 @@ particular. 12. Look again at sheet hierarchy stuff; I'm pretty sure this only works - when the graft is in the default orientation. + when the graft is in the default orientation. (There's lots of + problems in this area, most of which are in Beagle, some of which + are in McCLIM (by design? But not sure which are by design), e.g. + sheets being mirrored but with no medium attached). 13. I'd like to see the silica functionality in a separate package; I think (need to check!) that silica + back-end implementation should

1 0

[mcclim-cvs] CVS update: mcclim/Backends/beagle/windowing/port.lisp
by drose＠common-lisp.net 17 May '05

17 May '05

Update of /project/mcclim/cvsroot/mcclim/Backends/beagle/windowing In directory common-lisp.net:/tmp/cvs-serv19755/beagle/windowing Modified Files: port.lisp Log Message: Add auto release pool to function that creates the event loop to stop OS X barking all the time. Date: Tue May 17 19:56:20 2005 Author: drose Index: mcclim/Backends/beagle/windowing/port.lisp diff -u mcclim/Backends/beagle/windowing/port.lisp:1.2 mcclim/Backends/beagle/windowing/port.lisp:1.3 --- mcclim/Backends/beagle/windowing/port.lisp:1.2 Tue May 17 19:51:15 2005 +++ mcclim/Backends/beagle/windowing/port.lisp Tue May 17 19:56:20 2005 @@ -159,6 +159,7 @@ (setf (port-event-process port) (clim-sys:make-process (lambda () + (ccl::create-autorelease-pool) (loop (with-simple-restart (restart-event-loop "Restart CLIM's event loop.")

1 0

[mcclim-cvs] CVS update: mcclim/Backends/beagle/windowing/port.lisp
by drose＠common-lisp.net 17 May '05

17 May '05

Update of /project/mcclim/cvsroot/mcclim/Backends/beagle/windowing In directory common-lisp.net:/tmp/cvs-serv19635/beagle/windowing Modified Files: port.lisp Log Message: Apply Cyrus Harmon's changes to Beagle key handling posted on 08-MAR-2005. Not sure if these were never applied or if they have been clobbered since. Date: Tue May 17 19:51:15 2005 Author: drose Index: mcclim/Backends/beagle/windowing/port.lisp diff -u mcclim/Backends/beagle/windowing/port.lisp:1.1 mcclim/Backends/beagle/windowing/port.lisp:1.2 --- mcclim/Backends/beagle/windowing/port.lisp:1.1 Tue May 17 00:13:21 2005 +++ mcclim/Backends/beagle/windowing/port.lisp Tue May 17 19:51:15 2005 @@ -135,6 +135,8 @@ ;;; the port-server-path? Not sure...) to permit the user to make use of screens other ;;; than the main screen. (defmethod initialize-beagle ((port beagle-port)) + + (ccl::create-autorelease-pool) ;; CLX port gets some options here and uses those to set stuff up. We should probably do ;; this too, in the future ::FIXME::

1 0

[mcclim-cvs] CVS update: mcclim/Backends/beagle/input/events.lisp
by drose＠common-lisp.net 17 May '05

17 May '05

Update of /project/mcclim/cvsroot/mcclim/Backends/beagle/input In directory common-lisp.net:/tmp/cvs-serv19635/beagle/input Modified Files: events.lisp Log Message: Apply Cyrus Harmon's changes to Beagle key handling posted on 08-MAR-2005. Not sure if these were never applied or if they have been clobbered since. Date: Tue May 17 19:51:15 2005 Author: drose Index: mcclim/Backends/beagle/input/events.lisp diff -u mcclim/Backends/beagle/input/events.lisp:1.1 mcclim/Backends/beagle/input/events.lisp:1.2 --- mcclim/Backends/beagle/input/events.lisp:1.1 Tue May 17 00:13:16 2005 +++ mcclim/Backends/beagle/input/events.lisp Tue May 17 19:51:14 2005 @@ -28,7 +28,7 @@ #|| -$Id: events.lisp,v 1.1 2005/05/16 22:13:16 drose Exp $ +$Id: events.lisp,v 1.2 2005/05/17 17:51:14 drose Exp $ All these are copied pretty much from CLX/port.lisp @@ -571,23 +571,24 @@ ;; We need to maintain the modifier flags state constantly to be able to ;; implement this; suggest a slot in beagle-port? (when (equal #$NSFlagsChanged event-type) - (format *debug-io* "In event-build (flags changed)~%") +;;; (format *debug-io* "In event-build (flags changed)~%") ;; Use the 'old' 'modifiers' in conjunction with the new 'modifier-state' ;; to work out if this is a key up or a key down... - (setf return-event (make-instance (if (current-mods-map-to-key-down (send event 'modifier-flags)) - 'key-press-event - 'key-release-event) - :key-name nil - :key-character nil - :x 0 - :y 0 - :graft-x 0 - :graft-y 0 - ;; Irrespective of where the key event happened, send it - ;; to the sheet that has key-focus for the port. - :sheet (beagle-port-key-focus *beagle-port*) - :modifier-state (beagle-modifier-to-modifier-state (send event 'modifier-flags)) - :timestamp (incf timestamp)))) + (setf return-event + (destructuring-bind (event-class key) + (current-mods-map-to-key (send event 'modifier-flags)) + (make-instance event-class + :key-name key + :key-character nil + :x 0 + :y 0 + :graft-x 0 + :graft-y 0 + ;; Irrespective of where the key event happened, send it + ;; to the sheet that has key-focus for the port. + :sheet (beagle-port-key-focus *beagle-port*) + :modifier-state (beagle-modifier-to-modifier-state (send event 'modifier-flags)) + :timestamp (incf timestamp))))) ;; #$NSHelpRequested- wonder if we can convert this into "user pressed help key" key event? ;; Then could pull up docs (or could do if there were any!) @@ -630,7 +631,7 @@ ;;; This is really, really horribly written. Hopefully it will just be ;;; temporary until everything is 'band-aided' (!?) at which point we'll ;;; look to migrate to Carbon and reimplement a lot of this stuff. -(defun current-mods-map-to-key-down (current-modifier-state) +(defun current-mods-map-to-key (current-modifier-state) (declare (special *-current-event-modifier-state-*)) ;; Are there modifiers in 'current-modifier-state' that don't exist in ;; *-current-event-modifier-state-* (key down) or vice versa (key up)? @@ -643,38 +644,38 @@ ;;#$NSAlternateKeyMask +super-key+ ;;#$NSAlphaShiftKeyMask +hyper-key+ (cond ((null *-current-event-modifier-state-*) - t) + '(key-release-event nil)) ((and (> (logand *-current-event-modifier-state-* +shift-key+) 0) (= (logand current-modifier-state #$NSShiftKeyMask) 0)) - nil) + '(key-release-event :shift)) ((and (= (logand *-current-event-modifier-state-* +shift-key+) 0) (> (logand current-modifier-state #$NSShiftKeyMask) 0)) - t) + '(key-press-event :shift)) ((and (> (logand *-current-event-modifier-state-* +control-key+) 0) (= (logand current-modifier-state #$NSControlKeyMask) 0)) - nil) + '(key-release-event :control)) ((and (= (logand *-current-event-modifier-state-* +control-key+) 0) (> (logand current-modifier-state #$NSControlKeyMask) 0)) - t) + '(key-press-event :control)) ((and (> (logand *-current-event-modifier-state-* +meta-key+) 0) (= (logand current-modifier-state #$NSCommandKeyMask) 0)) - nil) + '(key-release-event :meta)) ((and (= (logand *-current-event-modifier-state-* +meta-key+) 0) (> (logand current-modifier-state #$NSCommandKeyMask) 0)) - t) + '(key-press-event :meta)) ((and (> (logand *-current-event-modifier-state-* +super-key+) 0) (= (logand current-modifier-state #$NSAlternateKeyMask) 0)) - nil) + '(key-release-event :super)) ((and (= (logand *-current-event-modifier-state-* +super-key+) 0) (> (logand current-modifier-state #$NSAlternateKeyMask) 0)) - t) + '(key-press-event :super)) ((and (> (logand *-current-event-modifier-state-* +hyper-key+) 0) (= (logand current-modifier-state #$NSAlphaShiftKeyMask) 0)) - nil) + '(key-release-event :hyper)) ((and (= (logand *-current-event-modifier-state-* +hyper-key+) 0) (> (logand current-modifier-state #$NSAlphaShiftKeyMask) 0)) - t) - (t nil))) + '(key-press-event :hyper)) + (t '(key-release-event)))) ;; Need to make use of the Cocoa method for getting modifier state - this is independent of events @@ -764,21 +765,29 @@ (let ((key-name (lookup-keysym (send ns-string-characters-in :character-at-index 0)))) ;; If key-name is nil after all that, see if we can look up a mapping from those supported in ;; Cocoa... -;;; (when (null key-name) -;;; (setf key-name (get-key-name-from-cocoa-constants ns-string-characters-in))) -;;; (format *terminal-io* "Got key-name of: ~A~%" key-name) - key-name)))) + (cond + ((null key-name) + (let ((clim-key + (get-key-name-from-cocoa-constants + (send ns-string-characters-in :character-at-index 0)))) + clim-key)) + (t key-name)))))) ;;; From CLX/keysyms.lisp (defun numeric-keysym-to-character (keysym) - (and (<= 0 keysym 255) - (code-char keysym))) + (cond + ((= #x1b keysym) + (get-key-name-from-cocoa-constants keysym)) + ((and (<= 0 keysym 255)) + (code-char keysym)) + (t nil))) (defun keysym-to-character (keysym) (numeric-keysym-to-character (reverse-lookup-keysym keysym))) -(defconstant *beagle-key-constants* '(#$NSUpArrowFunctionKey :UP +(defconstant *beagle-key-constants* (list + #$NSUpArrowFunctionKey :UP #$NSDownArrowFunctionKey :DOWN #$NSLeftArrowFunctionKey :LEFT #$NSRightArrowFunctionKey :RIGHT @@ -849,10 +858,33 @@ #$NSRedoFunctionKey :REDO #$NSFindFunctionKey :FIND #$NSHelpFunctionKey :HELP - #$NSModeSwitchFunctionKey :MODE-SWITCH)) + #$NSModeSwitchFunctionKey :MODE-SWITCH + #x1b :ESCAPE)) ;;;(defun get-key-name-from-cocoa-constants (ns-in) ;;; (loop for target, key in *cocoa-key-constants* ;;; (do ;;; (when (send target :is-equal-to-string ns-in) ;;; key)))) + +(defvar *beagle-key-hash-table* + (make-hash-table :test #'eql)) + +(defvar *reverse-beagle-key-hash-table* + (make-hash-table :test #'eq)) + +(defun define-beagle-key (ns-key clim-key) + (pushnew clim-key (gethash ns-key *beagle-key-hash-table*)) + (setf (gethash clim-key *reverse-beagle-key-hash-table*) ns-key)) + +(defun lookup-beagle-key (ns-key) + (car (last (gethash ns-key *beagle-key-hash-table*)))) + +(defun reverse-lookup-beagle-key (clim-key) + (gethash clim-key *reverse-beagle-key-hash-table*)) + +(loop for key-binding on *beagle-key-constants* by #'cddr + do (define-beagle-key (car key-binding) (cadr key-binding))) + +(defun get-key-name-from-cocoa-constants (ns-in) + (lookup-beagle-key ns-in))

1 0

[mcclim-cvs] CVS update: mcclim/Backends/beagle/windowing/frame-manager.lisp mcclim/Backends/beagle/windowing/graft.lisp mcclim/Backends/beagle/windowing/mirror.lisp mcclim/Backends/beagle/windowing/port.lisp
by drose＠common-lisp.net 16 May '05

16 May '05

Update of /project/mcclim/cvsroot/mcclim/Backends/beagle/windowing In directory common-lisp.net:/tmp/cvs-serv13235/beagle/windowing Added Files: frame-manager.lisp graft.lisp mirror.lisp port.lisp Log Message: Restructured beagle code. Added support for patterns (tiled works, untiled is a hack, stencils untested (probably don't work)), some speedups (each medium caches an NSBezierPath now with which to draw), scrolling largish output histories is improved (only redraw mirror region rather than whole sheet). Tried to squash a few bugs, largely without success but one or two have succumbed. Added Glimpse, which is a weak window inspector and even weaker Peek application but it's largely unfinished at the moment. Date: Tue May 17 00:13:21 2005 Author: drose

1 0

[mcclim-cvs] CVS update: mcclim/Backends/beagle/tests/drawing-tests.lisp mcclim/Backends/beagle/tests/graft-tests.lisp mcclim/Backends/beagle/tests/hello.lisp
by drose＠common-lisp.net 16 May '05

16 May '05

Update of /project/mcclim/cvsroot/mcclim/Backends/beagle/tests In directory common-lisp.net:/tmp/cvs-serv13235/beagle/tests Added Files: drawing-tests.lisp graft-tests.lisp hello.lisp Log Message: Restructured beagle code. Added support for patterns (tiled works, untiled is a hack, stencils untested (probably don't work)), some speedups (each medium caches an NSBezierPath now with which to draw), scrolling largish output histories is improved (only redraw mirror region rather than whole sheet). Tried to squash a few bugs, largely without success but one or two have succumbed. Added Glimpse, which is a weak window inspector and even weaker Peek application but it's largely unfinished at the moment. Date: Tue May 17 00:13:20 2005 Author: drose

1 0

[mcclim-cvs] CVS update: mcclim/Backends/beagle/profile/README mcclim/Backends/beagle/profile/profile.lisp
by drose＠common-lisp.net 16 May '05

16 May '05

Update of /project/mcclim/cvsroot/mcclim/Backends/beagle/profile In directory common-lisp.net:/tmp/cvs-serv13235/beagle/profile Added Files: README profile.lisp Log Message: Restructured beagle code. Added support for patterns (tiled works, untiled is a hack, stencils untested (probably don't work)), some speedups (each medium caches an NSBezierPath now with which to draw), scrolling largish output histories is improved (only redraw mirror region rather than whole sheet). Tried to squash a few bugs, largely without success but one or two have succumbed. Added Glimpse, which is a weak window inspector and even weaker Peek application but it's largely unfinished at the moment. Date: Tue May 17 00:13:20 2005 Author: drose

1 0