Bugs (no particular order): ---------------------- Prioritize this list! When copying components, be sure to preserve locked bit. FIXED Text direction on rotated fast text line is wrong. BETTER Length of fast text line is wrong. Length of fast text line is still too long (for attributes) since I don't take SHOW_BOTH_VALUE into account Relative paths are not displayed properly in the component add dialog box Page Revert may or may not have a memory leak, investigate this. Try opening a file called untitled_1.sch, doesn't work since you have an untitled_1 (the default) already open. FIXED Change the color of an attached attribute, and then copy it (color changes to yellow oops). translate 0 does not seem to work right with nets in the file... (nets should not be in the picture anyway, but either way...) Zoom in a lot and try to select something with the select box. Not easy to do. GONE stretch seems to pick random objects to stretch if you don't select something (cannot replicate) Swedish characters don't work. In fact, any Multi key character acts funky (or core dump) FIXED still have some core problems with net extension and selections Thick lines aren't imaged (write png) correctly Zoom with pan doesn't work 100% when you are zoomed out all the way. FIXED Net consolidate will not work right (core dump) if you select a net and then try to extend that net.. You need to remove that net which is being deleted from the selection list, before you delete it. Misspell a keyword value in the rc file and gschem cores.. BETTER Almost got that 180 postscript printing bug fixed... not perfect yet Mouse pan crashed my X server once FIXED slow redraw (due to o_conn_disconnect_update being there), added o_redraw_all_fast to be used in places where it's necessary to redraw quickly (like in the panning) slow drawbounding box (caused by way too many get_complex_bound calls) GONE Select some text while a symbol is at 0,0 and you will see something being erased at 0,0. I haven't got a clue as to what is being erased. Actually I do, it's the selection list's copy of the text. it seems. I bet before I translate text object I erase it. hmmmm If a symbol isn't found, it's removed from the schematic.... Is this the behavior we want? Mirror still has some bugs in it when dealing with the new conn system (mirror will create extra midpoint connections which are temporary only) If you put bogus numbers (ie not 0 or 1) in for the visibility and/or show_name_value field, you get *totally bogus results! Multiple instances of gschem clobber each other's log file, solve this how? Document how tswindow is being used in the include file. (it's also used by snap size), have snapsize get it's own window selection2 related bugs: FIXED lock and unlock need a seperate color for locking.. FIXED go through o_misc and get them all working (lock, unlock, embed, unembed) FIXED o_mirror doesn't work FIXED color change doesn't work... FIXED o_edit temporarly #ifed out FIXED i_callback_attrib_attach. needs to loop over selection list... FIXED copy attribute, stays an attribute... ? FIXED i_callback_attrib_detach is also broken FIXED edit text and it gets an invalid color FIXED toggle vis totally broken FIXED selection of invisible FIXED Copy an object and attribute, then unselect each using CTRL and then click away to unselect both and attribute becomes invalid color (caused specifically by: select object, deselect object, text was already selected) (not related to copy at all...) FIXED net conslidate is broken on objects which have attributes and selected, totally broken (dumps core), fixed with a hack (unselect all before you attempt). Still todo in selection2: FIXED All slot= stuff does not work yet does it cause problems that the translate dialog box doesn't grab the focus for the app? there's a bug in arc visiblity clipping, just zoom around close in using DB*-1.sym (the corners) GONE Mirror something and then change slots... selection is wrong GONE Move a single net. You will see a stray visual cue at 0,0 Setting slot=something other than 1 in a symbol doesn't work (ie slot isn't updated when you place the component). Copy something with slot != 1 and it doesn't draw right when it's placed Start a component place and then change to a different page. You get a whole bunch of cmt2 NO! warnings GONE? Core dump on 64 bit architectures? 1) open stack_1.sch from the examples dir 2) box zoom somewhere in the schematic 3) drag the vertical scroll bar up to the top of its range KLUDGED Really odd behavior with display :'s... type some :'s and then ??????? add a component or net or pin and they disappear? huh? (FIXED by adding gdk_draw_point in o_line, not a good solution though) Also don't draw lines with zero length in general Observed some really odd characters being written to the log files (between Read local gschemrc and Saved ... GONE Rotate object... slot is wrong Rotate object with a slot number then copy it immediately. slot attribute not being selected. View Limits is broken for a single horizontal line (at least you can't see the line because it's below the drawing area) GONE When placing an invisible attribute, you get the left over image that needs to be erased. I'm not eraseing or dealing with the slot= attribute if you change the symbol and the attribute is visible. It won't get properly drawn. (need to do a redraw to see the visible results) GONE In o_select and o_select_many, when you try to update the tail to be proper, select_many doesn't work. I wonder what else is effected Might have fixed some of this, but it's still highly suspect. Got a "Ahh you blew it.. tried to delete something which does exist" error. need to watch for that and replicated it GONE? Got an unexplained core dump, dealing with trying to disconnect a bogus net (already freed?). I haven't been able to replicate this at all. GONE Figure out and fix why text when selected and deselected (and placed) is sometimes offset (visually). You have to do a redraw to get one image (non-offset) GONE When doing "open symbols...", the newly loaded symbols are not zoomed limits Bug in arcs and large large world-size... Don't draw right. I think this was caused by incorrect aspect ratio Minor bug dealing with zooming out of corners (the zoom factors going out are different then going in) GONE? Saw some odd selection behavior, ie selecting something when you were not inside it's bounding box. I think this was caused by not having snap on. OKAY? Selection boxes and hitting the control key does not behave as it should (ie selection box toggles existing objects) OKAY? Found some wierdness with the symbol translate (due to snap being off) OKAY? Unexplained core dump with related to object_lastplace being bogus. cannot replicate this. (haven't seen this at all again) Regular objects which are locked and then saved to a file will not be locked anymore... locking only seems to work on components. I'm not sure if tty vs both vs window_log stuff is working? low priority though Select an clib from the comp list and scroll down the component pane and then select another clib, other pane not starting from top OKAY? Another segfault. This one totally unexplained. Was seeing some odd behavor with the hotkeys (hotkeys not doing what they should). Core dump happened the moment I press the right mouse button (the popup menu) which makes me suspect a gtk problem. This might be related to the odd accelerator problem I was seeing (now accelerators are removed from the right mouse button popup) Unfortunately core not generated since gtk wasn't compiled to generate code dump files. Need to recompile gtk to do so. GONE A segfault will occur if you attach NON text items to other items and then delete the whole bunch and then try to select something. The cause is unknown but a possible fix is known. You should not be allowed to attach NON text items as attributes. (it segfault doesn't always happen, can't replicate it, but it's there somewhere) OKAY Apparently, an undocumented feature (or maybe it is documented), but if you press a key while you select a item off of a menu (popup or pulldown) then you can change the hot key? Hmmm... that's pretty interesting. Some sort of resize error when you open a file and the maximize Related to above: load up some schematics, maximize window, flip between pages... grid drawn wrong. (probably because in the resize handler, you don't go through and set width/height of window in all pages) You still get xor droppings during a place and expose event happens You still get window droppings due to interactions with: - status updating (causing expose events which are being ignored - middle button shortcut updates (same reason as above) GONE What happens if you attach an attribute to two places... nothing, but you should disallow that. GONE Still have dangling cue blip in lower corner... now seeing it on copys/moves, related to selected drawing conns. CHANGED flag -- make sure it's set in all places. BETTER Do some sanity on input files (check for zero's...) Need to fix +4 stuff in get_line_bounding etc... : - get consistant name - and perhaps make +4 zoom_factor aware OKAY You can't put ='s in text strings (unless it's the seperator for name=value in attributes) (is this a bug or just a restriction?) (either way, document it). MOSTLY Go through and be sure you use the #DEFINES created by configure DONE If a move/copy/compdraw is under an exposed area and an expose event happens then you get xor droppings There is still a bug.. draw a net and then another one which would cause a midpoint connection... only after you do a redraw do you get the change many ways to fix this... but all required either 1) a redraw after every major action, or 2) or a smarter "recalc" routine Check the issues related to visibility and if you can select it and if it's an attribute Pressing key during place causes droppings? Hitting keys and during an action (x_events.c) Select something, then hit 'n' instead of 'm' and you get left overs from the net command (same thing as mentioned above. inside action and hitting a key leaves ghosts) GONE? place two complex. - Connect net inbetween. - Connect another net to that new net Move all... see what happens? ( be sure components are at the end of each of the nets ) It's caused by order dependant searching and if the old and new overlap you will get a round connection Things get really slow with huge object copies... Perhaps we don't want to be doing the o_conn_recalc_all on actions. Major bug with add component with outline/bounding, if you zoom inbetween an add you get this HUGE monstrocity! (transparent zooms not implemented yet) MUCH Arcs are a hack by default BETTER This has gotten better (a lot better). FIXED? Unexplained grid locking when you draw an arc? must have something to do with the way width+x-x and height+y-y rounds off? Huge bounding box on arcs... uggg For now, arcs have been removed from the boundingbox calc. So if you just move an arc, you can't see the bounding box. Odd bug related to doing a objtosym (sym cmd), in that you don't get an exact move to zero unless you zoom one (or more?) XBUG? Saw some serious root-window corruption and drawing_area corruption because 1) of an X server bug in XF86S3V or 2) I'm not clipping lines (line clipping should be done by gschem) ----------------------------------------------------------------------------- Finished stuff: FIXED Zoomin far enough and lines/nets start acting really weird!!!!!!! (by properly clipping the line) FIXED pins colors are being written out as -1 FIXED Gschem core dumps when I hit the special Windows key... ha, it doesn't like microsoft keys? :) DONE THERE IS A SIGNIFICANT slowdown in this version (19981013 before another release) delete really (with o_conn_disconnect_all with big schematics) FIXED Attaching an attribute doesn't set CHANGED flag FIXED Right now you can attached a text item as an attribute in two places. Disallow that FIXED Right now you can attach non text items as attributes. Disallow that. FIXED MAJOR PROBLEM!!!!!!!!!!!!!!!!!!!!! gschem is broken with this broken Got a major annoyance... order in copying objects to get attributes to attach is a problem. How do I fix that? by copying text last FIXED If you try to unembed a symbol and it cannot be found in any search path then things get slightly screwed up. See a_basic.c o_unembed FIXED Found another core dump bug, if you send a DELETE event to the window (from the wm) then on exit gschem will core. This is okay, since I don't deal with the DELETE event yet... OKAY Order in the basename_list is not based on alphabetics (it's based NOW on the creation date/inode of the file) (make the order alphabetic) BETTER Get consistant behavior for SHIFT/CONTROL in the various modes FIXED AGGGG... stuff showing up in basename_list in the wrong order :( DONE Can't print )'s and spacing between chars on paper is much larger than on the screen (due to postscript using )'s : filter input) DONE limits on complex_filename, drawing_filename, text_string etc... REMOVE them (the limits that is) Size of drawing_filename and friends need to be dynamic.... drawing_filename text_string complex_filename basename clib FIXED Droppings with thick pins and move/copies/deletes? DONE Multiconnection midpoints, should draw a circle, but doesn't for now. Put warnings in documentation (ie when you draw many nets to a single point you should get a circle... instead you get nothing... DONE Cleanup grid/snapping so that it's more flexable so that you could have half grid snaps... (MAJOR ANNOYANCE) GONE! Related bug to above, drawing lines, first point must lock and then the second can be free. doesn't work vice versa! (related to above) OKAY A small bug dealing with arcs and the symbol command. The x,y location will get negative values (which may be okay actually) FIXED Roundoff errors creaping into move/copy commands ( some objects would be offset too much or too little relative to the other ones ) Fixed mostly, but still need to re-examine screen to world mapping functions. This still needs to be dealt with properly. I think this is fixed now. Did move in world coords instead of screen coords FIXED Unexplained core dump when opening another window (when compiled with -lccmalloc -ldl) related to: Odd behavior when quiting multiple windows and then opening another schematic in existing windows (can't find a font?) Was caused by freeing common rc variables (ie each window didn't have a proper copy of the string) GONE! The difference between a save (saveas) and open window is not enough. This has cause me to nuke a schematic more than once already! (this was fixed by removing this behavior) FIXED Opening up a second window (or more) causes the rc file to be read again and that adds the new paths to the current ones.. ugg! change that! DONE There is still some sort of odd behavior with the page manager losing its mind. Not good. not sure if it's a gtk problem or mine. Problem has been identified as bogus parameters to gtk_clist_moveto. DONE Need to check if file exists (or check the fp) when you open files. FIXED The dialog boxes (in x_dialog) don't look all the same FIXED Zoom in a bit, and try selecting something... and then something else (many times) and you will see a bug (objects not being selected when they should be). (only when zoomed in...) This is caused by the event state machine going into sbox mode when it should not! Fix this by converting the threshold into something that is dependent on the current zoom factor FIXED There is a condition that cause gschem to exit (unwanted), Edit some text and then unselect the text that is being edited. This bug surfaced when we removed the dialog box grabs. Other dialog box might be affected as well... (fixed by adding grab back to some dialog boxes) FIXED Suddenly some sort of an event happened and drawing window area was totally wrong (too small) (caused by status label being too long) ( Don't have too long status labels ) GONE For every drawing_filename and strcpy (sprintf) you need to check the size (now you need to worry about page_filename) GONE Once got a "(null)"... how? Haven't seen it again when I was drawing text. String was set to null. Should be fixed, but don't know. GONE Be sure to go back and fix the huge bounding boxes on (16,32pt) text FIXED The selection name in saveas file selectors (and/or others) is being nuked. (problem with gtk?) OKAY Set a default value for show_name_value FIXED Small erases happen once in a while when connecting three nets to the same point FIXED Also, you cannot print any filename (the path is the problem) that has a . in it. Fixed by replacing the .sch with .ps or just adding it FIXED Still a minor textsize vs pan vs zoom bug. zoom in all the way, the pan to the edge repeatedly then zoom out... text size doesn't look right (fix by getting flexable font system see below ) FIXED Drawing a box incorrectly (ie right then left) cause invalid boxes to be drawn and puts invalid boxes into the database. FIXED Printing arcs is broken! FIXED Uggg... open multiple windows... add ntext and then close them... core dumpy (caused by stupid programmer, freeing memory which was already freed accidentally) GONE Some xor droppings being left behind when moving text? X server bug? This was seen only on my old box and nowhere else DONE Change MOVE so it's sequence is: 1) press m, 2) click to start, 3) press to place. FIXED Odd coredump see CORE.LOG, have to see if the happens again with coredumping on. (wasn't reseting left,right,top,bottom for heads) FIXED Hitting delete and saveas window. and then hitting cancel. FIXED Misc pixels at the bottom of drawing area, cause by doing an open and then win_height gets set to 719 (supposed to be 720). The right height will be set if a configure event happens.... FIXED Don't get a configure event if you hit key while the window is coming up. DONE Inconcistancy with click and move and click-hold-draw of nets and others. GONE Semaphore the interface shared memory FIXED Place something when something is selected! BUG, original selection doesn't get unselected FIXED Cause of MAJOR instablity on SGI and others: Change type of type and p_type to be int or do something about the -1 Type is now an int GONE Try quitting program before gschem_intr starts.. you get: _X11TransSocketUNIXConnect: Can't connect: errno = 4 GONE Still occasionally see a screwup on the update of an interface DONE You need to put a linewidth fix in o_pin so that when you draw a pin it gets finish (redraw) with the right width (don't need this in o_net because of the o_conn_recalc stuff) FIXED Still have some doubts over which object should be unredraw ( the selection or the actual object ) (right now it's the actual object) FIXED Placing a component on another object causes the old comp to be selected. This is due to the way we handle the select mode ( on the release ). There should be another mode which is null which causes a switch to SELECT after the release. FIXED Still have some draws when you are doing mass deletes.... coredumps See o_conn_disconnect_all in s_basic (s_delete) OKAY Visibility of text works, but don't hit toggle vis during an action OKAY Fixed a small bug in o_select, verify it all still works FIXED Round connection droppings. Demonstrate: when you have many nets some with round connetions. Delete them all. Now, when you do this some of the nets where the round connections were are deleted (and the associated link in the other round connected net). So when you erase the round net, you get droppings, because the net doesn't know that it had round connections. Solution: put in some code in draw: which: 1) if override color == background 2) connection_1 or connection_2 are not zero 3) erase the corresponding one THE FIX is in o_net.c in the draw... just as above This fix might cause side effects if we start copy/moving the connection points FIXED Another round connection problem, draw a net. Then draw a net which would cause a round connection point to occur. You will see that second net is on top of first connection point. How do you solve this? (by having a redraw all connection points after doing a redraw) GONE Still occasionally see interface die or lock up... ???? FIXED when you draw some text and then you resize the window, the text isn't the same length as it was. (Fixed by drawing each char individually and varring this space depending on size of window) Height is still off though, and the lenght is still not *exactly* accurate, but this might be good enough (for now). OKAY? Minor bug, but resize events occure twice... put in flag so that it occures only once. FIXED Major pin/connection bug. Draw symbol. connect to another symbol I THINK (through just a pin) delete one of them.. then copy/move other. ? core dumpy... :( OKAY Odd bug, when you select a complex, the selected image doesn't I THINK always lineup. Has something to do with non snap objects and ? o_*_copy and o_*_translate (has to do with inaccuracies in the display at zoom_factor 0) FIXED Fix makefile so that it handles dependencies correctly FIXED Deal with pins touching pins... FIXED Deal with connections to the middle of an net OKAY? If you select a net, and then connect something to that selected net, it becomes unselected. (is this okay?) OKAY Connection points are NOT selected, minor asthetic bug, code gets messy if you try to select them, because un selection becomes hell OKAY Still have some minor net conns (the line) not being redrawn if you connect to the middle of a net. Have to go through code and really figure out what is happening where? What's happening is: how can the program know that you just drew a line with a possible midpoint connection? Well solution for now is the o_recalc_conn key DONE Move origin to lower left instead of upper left DONE Make aspect ratio of both window and coord system 4/3 DONE Still have some net connections not being update. Need a button for update nets KLUDGED Deleting object in particular a net, doesn't redraw connection points I THINK I CAN CLEANLY FIX THIS, BUT it's going to take a new s_delete???? OKAY Drawing a net, doesn't show the first connection point until you select the second. Is this the behavior we want? Yes, because you don't know what kind of connection you'll get FIXED: selecting object, and then zoom, doesn't redraw selected object FIXED: zooming too much core dump FIXED: possible, odd way net draw works with button down or up NOW, nets are drawn when the n key is hit. FIXED: When you move a complex, you get a total redraw! don't do that FIXED: Related to above, doing delete of a complex does a redraw! no! FIXED: Pan doesn't always give perfectly square aspects MOSTLY: Well there is still an aspect bug going to the far right/top bottom corner (still a bug with text size though) GONE Nasty constants in a_pan.c FIXED pan bug with interface. Select pan on the interface window, and then pick a point which will center an object. That object will be selected. (Fixed by removing the SELECT case in button_release, not needed and was causing this bug) DONE Change the add component so that you can either drag outline or boundingbox, should be straightforward, just change o_drawbounding to be more general GONE? There is still a select, copy with bounding box on (you get xor leftovers) FIXED Fixit so that aspect ratio of display/window doesn't matter?? Is this even possible? (It is possible, and it does matter!) DONE! Add hotkey for insert component (i perhaps? yes) DONE! Text dynamic storage, but it is still limited to 80 chars DONE! Text size selection (was solved through intr) DONE! Change arcs, so that degrees are stored, instead of deg*64 DONE! lack of coupling between zoom_factor and left,top,right,bottom MOSTLY (still need some work with font sizes and zoom_factor coupling ) NOT Zoom different amounts depending on the inital values for right, left DONE etc.. (Irrelavent) FIXED Contrained move/copy works for the most part, but try hitting the key AND and then going in and out of it... leaves ghosting... DONE The solution to this is to 1) disallow mode changing in the middle of an action or 2) erase old ghost depending on if you are entering contrained or exiting it. (you'll need a flag which is set depending on if your in it or not) This problem isn't with constrained but with copy/move and actions (Hitting any key will cause ghosting) I think? (key inside action...) DONE! Go through o_basic and move o_save(read)_attrib to o_attrib.c DONE Streamline all the X globals and stuff in setup NOTED Some sort of bug with text/zooming in all the way and outline moving Only with biggest font (only on my home machine, new linuxes works fine) GONE If you are typing text and before you hit enter you change sizes, the current text should reflect this text size FIXED Fonts and aspect ratio are not related, so they don't always look MOSTLY the same. GONE? Look for not eraseing bounding box with boundingbox draw outline mode GONE! Related to above try switching between outline and bounding while inside a move/copy, outline stuff is left behind! OKAY You need to redraw an object if you move/translate/ copy it, so I THINK that the bounding box is update... do we want to change that? (the way to fix this is recalc bounding in the translate functions) OKAY Minor mystery why attributes inside complex don't change colors (which is good), oh because they are read from disk! I think!? YES! FIXED Odd behavour of buttons (move/copy), wasn't written for multiselect FIXED BUG! Select something, and then draw a multiselect box (don't select anything in this box) and you will see that first object won't be unselected FIXED Coredump when you do an attach with nothing selected? GONE There is a small redraw bug in net/copy/cancel which is hard to reproduce, but basically object_selected wasn't being redrawn OKAY Another redraw bug related to hitting escape, if you have a redraw in that section then you get two redraws when you inside an action Be sure all cancel action buttons look like the right button one, because I think that's the most current one DONE! Having a too large text size causes a core dump, fix this by putting in a check during read/write and force value DONE! Corefile is still generated if you have lots of objects, connected it an odd fashion and then you do a delete. basically in the conn disconnect you get a dangling pointer. Solution is to go through and disconnect every object which points to the current object! Verify this! DONE! Go through and make all colors configurable DONE! Quirky multiple select cycle (okay I think) DONE! Not too happy with the feedback when you are placing a component Two lines are okay, but bounding box would be much much better! The way to do this is to create a dummy object (read in file) and use this as the object when you draw the bounding box (of course this is exactly what was done) MOSTLY Am I freeing everything when I leave? memory leak? FIXED There is still a small small leak somewhere (yup) NOTDONE go back and add gdk_flush wherever there was a XFlush Don't do this.. gdk_flush is not needed (and should be used). DONE When you copy/move something create a new list and copy those items into this list, so that you have something selected at the end. OKAY Attributes are going outside of a complex. Fine, but you need to do something about loads/saves so that some values which are inside a complex get put outside etc... DONE! Deleting all objects and then adding needs to call setup_struct(). FIXED Due to the new new_sel_head stuff, some buttons from interface are broken, see hack in x_events.c (next_state stuff) FIXED I think there is a bug related to this "it happened in deletes" NO Definately a data corruption bug - Select complex - copy complex - delete original - create copy of copy - unselect - redraw (copy is gone) Something with select and delete complex? broken behavior in s_delete, commented out "out for now" FIXED Possible bug in complex_draw and o_redraw Just copy and delete copy and delete DONE! Revamp structure by doing doubly linked list OKAY Not really a bug, but erasing the bounding box with an arc (since it is soo large) effects other parts of the drawing ( Fix this by making bounding box on arc smaller. OKAY Do we want to save complexs in a file or just the filename (like it is now) Yeah we want to keep it this way, gets way to complex OKAY Attributes, do we want a special structure or a regular one - Advantages of specialized, less data, simpler data struct - Disadvantages of specialized, whole new routines, code waste ( Attributes are text which special links on objects which link them ) DONE: Remove the font_inflation stuff DONE: Farthest left most point must be at the far left possible coord Solved with o_complex_to_symbol() GONE? When you do a delete (using keys), delete seems to want to erase an nonexistant bounding box (the cyan dashed thing). DONE! Fix names in i_keypressed.c FIXED using quit button isn't very clean for schem. stupid pause FIXED Maybe if you hit a action key (move/copy) and nothing is selected it can behave as if you were to hit the button... hmmmm GONE Work on detaching interface more... currently the program assumes that the interface is there... (this is getting there with the SIGCHLD handling ) -interface issues- ALL OF THESE ARE GONE with the intergration of the two RESOLVED If something is selected when you hit copy, schem will unselect it? Basically there are two modes: if an object is selected, then the next button press starts the action, else if not then first you go into select mode and then the action starts. GONE Add the ability to replace the file selection window when you hit the place command again (if the file selection window as already there) OKAY Ah, hitting okay when you don't select something.. ugg, in component select DONE outline of component for place, along with boundingbox GONE There is some sort of scrollbar problem with xforms? remove precision ( no more scroll bar for text size ) ( also no more xforms ) DONE Fix room for status (more more more!) DONE Fix right increment on text size DONE Component button needs ... DONE Fix grid color button DONE Fix default status line GONE Check for proper parameters, otherwise it will core DONE Possible race if both interface and schem try to send a command Yes.. solution is a semaphore DONE Free the semaphore when you are done? DONE somewhere put the hot key assignements (help? on buttons?) DONE Add outline toggle mode DONE Add attribute attach/detach stuff OKAY Missing GDK_CONFIGURE from event types wanted list? ----------------------