Updates:
Status: FixProposed Comment #11 on issue 1064 by marianopeck: There are still problem with keyboard shortcuts in Windows http://code.google.com/p/pharo/issues/detail?id=1064 from tn.. KeyboardEvent keyValue for 'ctrl + i' is evaluated to 9 as ASCII control code, and the value conflicts to Character tab value. I tried to correct source code for this problem. Below is the workaround for Pharo-1.1.1-OneClickCogVM on Windows XP. -------------------- InputEventSensor>>processEvent: processEvent: evt ... "Finally keyboard" type = EventTypeKeyboard ifTrue: [ (evt at: 4)=0 ifTrue: [ "-- add --" "Sswap ctrl/alt keys if neeeded" KeyDecodeTable at: {evt at: 3. evt at: 5} ifPresent: [:a | evt at: 3 put: a first; at: 6 put: a first; "-- add --" at: 5 put: a second]. ]. "-- add --" "Update state for polling calls" modifiers := evt at: 5. ^evt]. ... In order to work this code, Preference Control and Alt keys must be set to "Full matching". Because ASCII control characters (e.g. ^i) is mapped to character code with KeyDecodeTable, only at the time. About key event, see InputEventFetcher class comment. And it is interesting that to contrast implementations between Pharo and Squeak4.2. Each implementation is below. Pharo: InputEventSensor>>processEvent: Squeak4.2: EventSensor>>processEvent: Thanks, |
Comment #12 on issue 1064 by [hidden email]: There are still problem with keyboard shortcuts in Windows http://code.google.com/p/pharo/issues/detail?id=1064 Hi, I'm tn. 'ctrl + m' problem is Cog VM bug. Normally, when character key is pushed, keyboard events occur as follows on Squeak. 1. key press (down) 2. character (character code notification) 3. key release (up) They are produced by a primitive method in IutEventFetcher >> primGetNextEvent. But when ctrl + m is pushed, a character event doesn't occur. Because 'ctrl + m' value is 13. The value is same as ENTER key. Cog VM ignores WM_CHAR message(character event) when ENTER key is pushed. The one of solution for this problem is to add CTRL key state to the if-condition. -- sqWin32Window.c -- int recordKeyboardEvent(MSG *msg) { ... switch(msg->message) { ... case WM_CHAR: case WM_SYSCHAR: /* Note: VK_RETURN is recorded as virtual key ONLY */ if(keyCode == 13 && !ctrl) return 1; // add "&& !ctrl" pressCode = EventKeyChar; break; ... } ... Note: ENTER keyCode is 13. But when ctrl + ENTER is pushed, the keyCode is 10. So if the condition is revised, ctrl + m problem is corrected. but other behavior is not changed. If you want to prevent to generate character event for CTRL + ENTER, you can add following code. By this way, we can produce not only ctrl+m (0x0D)'s character event but also ctrl+j (0x0A)'s one. -- sqWin32Window.c -- // add this function static int isNotKeyDownCtrlReturn(MSG* msgp){ if(msgp->message == WM_KEYDOWN && msgp->wParam == VK_RETURN){ if(GetKeyState(VK_CONTROL) & 0x8000) { return FALSE; } } return TRUE; } ... int ioProcessEvents(void) { static MSG msg; ... if(isNotKeyDownCtrlReturn(&msg)){ // add TranslateMessage(&msg); } // add DispatchMessage(&msg); ... } ... int recordKeyboardEvent(MSG *msg) { ... if(pressCode == EventKeyDown && virtCode != 0 && isNotKeyDownCtrlReturn(lastMessage) ) { // add " && isNotKeyDownCtrlReturn(lastMessage) " /* generate extra character event */ ... } return 1; } Note: Below is a comment in the source code. /* note: several keys are not reported as character events; most noticably the mapped virtual keys. For those we generate extra character events here */ Those keys(e.g. arrow) should not be ascii code. ENTER key is not. but an extra character events is created for ENTER. I don't think this behavior is correct. As for ENTER key, an extra character event must not be generated and WM_CHAR should be processed. Please see msdn about TranslateMessage. An attached file is source, which has been corrected by above way. see also Comment 11. Thanks, Attachments: sqWin32Window.c 99 KB |
Comment #13 on issue 1064 by [hidden email]: There are still problem with keyboard shortcuts in Windows http://code.google.com/p/pharo/issues/detail?id=1064 <span style="color:red">test</span> |
Comment #14 on issue 1064 by [hidden email]: There are still problem with keyboard shortcuts in Windows http://code.google.com/p/pharo/issues/detail?id=1064 Hi, I'm tn. 'ctrl + m' problem is Cog VM bug. Normally, when character key is pushed, keyboard events occur as follows on Squeak. 1. key press (down) 2. character (character code notification) 3. key release (up) They are produced by a primitive method in IutEventFetcher >> primGetNextEvent. But when ctrl + m is pushed, a character event doesn't occur. Because 'ctrl + m' value is 13. The value is same as ENTER key. Cog VM ignores WM_CHAR message(character event) when ENTER key is pushed. The one of solution for this problem is to add CTRL key state to the if-condition. -- sqWin32Window.c -- int recordKeyboardEvent(MSG *msg) { ... switch(msg->message) { ... case WM_CHAR: case WM_SYSCHAR: /* Note: VK_RETURN is recorded as virtual key ONLY */ if(keyCode == 13 && !ctrl) return 1; // add "&& !ctrl" pressCode = EventKeyChar; break; ... } ... Note: ENTER keyCode is 13. But when ctrl + ENTER is pushed, the keyCode is 10. So if the condition is revised, ctrl + m problem is corrected. but other behavior is not changed. If you want to prevent to generate character event for CTRL + ENTER, you can add following code. By this way, we can produce not only ctrl+m (0x0D)'s character event but also ctrl+j (0x0A)'s one. -- sqWin32Window.c -- // add this function static int isNotKeyDownCtrlReturn(MSG* msgp){ if(msgp->message == WM_KEYDOWN && msgp->wParam == VK_RETURN){ if(GetKeyState(VK_CONTROL) & 0x8000) { return FALSE; } } return TRUE; } ... int ioProcessEvents(void) { static MSG msg; ... if(isNotKeyDownCtrlReturn(&msg)){ // add TranslateMessage(&msg); } // add DispatchMessage(&msg); ... } ... int recordKeyboardEvent(MSG *msg) { ... if(pressCode == EventKeyDown && virtCode != 0 && isNotKeyDownCtrlReturn(lastMessage) ) { // add " && isNotKeyDownCtrlReturn(lastMessage) " /* generate extra character event */ ... } return 1; } An attached file is source, which has been corrected by above way. Note: Below is a comment in the source code. /* note: several keys are not reported as character events; most noticably the mapped virtual keys. For those we generate extra character events here */ Those keys(e.g. arrow) should not be ascii code. ENTER key is not. but an extra character events is created for ENTER. I don't think this behavior is correct. As for ENTER key, an extra character event must not be generated, and WM_CHAR message should be processed. Please see MSDN about TranslateMessage. see also Comment 11. Thanks, Attachments: sqWin32Window.c 99 KB |
Hi, I fordward this thread to the VM mailing list since it looks more appropiated.
David, here "tn" is proposing a fix. I am not qualified to take a look. Hope it helps Mariano
---------- Forwarded message ---------- From: <[hidden email]> Date: Sat, Mar 5, 2011 at 9:23 PM Subject: Re: [Pharo-project] Issue 1064 in pharo: There are still problem with keyboard shortcuts in Windows To: [hidden email] Comment #14 on issue 1064 by [hidden email]: There are still problem with keyboard shortcuts in Windows
Hi, I'm tn. 'ctrl + m' problem is Cog VM bug. Normally, when character key is pushed, keyboard events occur as follows on Squeak. 1. key press (down) 2. character (character code notification) 3. key release (up) They are produced by a primitive method in IutEventFetcher >> primGetNextEvent. But when ctrl + m is pushed, a character event doesn't occur. Because 'ctrl + m' value is 13. The value is same as ENTER key. Cog VM ignores WM_CHAR message(character event) when ENTER key is pushed. The one of solution for this problem is to add CTRL key state to the if-condition. -- sqWin32Window.c -- int recordKeyboardEvent(MSG *msg) {  ...  switch(msg->message) {   ...   case WM_CHAR:   case WM_SYSCHAR:    /* Note: VK_RETURN is recorded as virtual key ONLY */    if(keyCode == 13  && !ctrl) return 1;       // add "&& !ctrl"    pressCode = EventKeyChar;    break;   ...  }  ... Note: ENTER keyCode is 13. But when ctrl + ENTER is pushed, the keyCode is 10. So if the condition is revised, ctrl + m problem is corrected. but other behavior is not changed. If you want to prevent to generate character event for CTRL + ENTER, you can add following code. By this way, we can produce not only ctrl+m (0x0D)'s character event but also ctrl+j (0x0A)'s one. -- sqWin32Window.c -- // add this function static int isNotKeyDownCtrlReturn(MSG* msgp){   if(msgp->message == WM_KEYDOWN && msgp->wParam == VK_RETURN){     if(GetKeyState(VK_CONTROL) & 0x8000) {       return FALSE;     }   }   return TRUE; } ... int ioProcessEvents(void) { static MSG msg;  ...    if(isNotKeyDownCtrlReturn(&msg)){  // add     TranslateMessage(&msg);    }                  // add    DispatchMessage(&msg);  ... } ... int recordKeyboardEvent(MSG *msg) {  ...  if(pressCode == EventKeyDown && virtCode != 0 && isNotKeyDownCtrlReturn(lastMessage) ) {                          // add " && isNotKeyDownCtrlReturn(lastMessage) "   /* generate extra character event */   ...  }  return 1; }
An attached file is source, which has been corrected by above way.
Note: Below is a comment in the source code.
and WM_CHAR message should be processed. /* note: several keys are not reported as character events;   most noticably the mapped virtual keys. For those we   generate extra character events here */ Those keys(e.g. arrow) should not be ascii code. ENTER key is not. but an extra character events is created for ENTER. I don't think this behavior is correct. As for ENTER key, an extra character event must not be generated, Please see MSDN about TranslateMessage. see also Comment 11. Thanks, Attachments:     sqWin32Window.c  99 KB |
Thanks Mariano, I added this to http://bugs.squeak.org/view.php?id=7611
so the issue will be tracked. The standard VM and the Cog VM share identical support code for this function, so I am assuming that the issue affects all Windows VMs (please correct me if this is wrong). Andreas, Eliot - this would be a one-line fix in the support code if you approve, and should be applied to both the trunk and the cog branches. Thanks, Dave. On Sun, Mar 06, 2011 at 04:12:24PM +0100, Mariano Martinez Peck wrote: > > Hi, I fordward this thread to the VM mailing list since it looks more > appropiated. > David, here "tn" is proposing a fix. I am not qualified to take a look. > Hope it helps > > Mariano > > ---------- Forwarded message ---------- > From: <[hidden email]> > Date: Sat, Mar 5, 2011 at 9:23 PM > Subject: Re: [Pharo-project] Issue 1064 in pharo: There are still problem > with keyboard shortcuts in Windows > To: [hidden email] > > > > Comment #14 on issue 1064 by [hidden email]: There are still problem > with keyboard shortcuts in Windows > > http://code.google.com/p/pharo/issues/detail?id=1064 > > Hi, I'm tn. > > 'ctrl + m' problem is Cog VM bug. > > Normally, when character key is pushed, keyboard events occur as follows on > Squeak. > > 1. key press (down) > 2. character (character code notification) > 3. key release (up) > > They are produced by a primitive method in IutEventFetcher >> > primGetNextEvent. > But when ctrl + m is pushed, a character event doesn't occur. > Because 'ctrl + m' value is 13. The value is same as ENTER key. > Cog VM ignores WM_CHAR message(character event) when ENTER key is pushed. > > The one of solution for this problem is to add CTRL key state to the > if-condition. > > -- sqWin32Window.c -- > > int recordKeyboardEvent(MSG *msg) { > ... > > switch(msg->message) { > ... > > case WM_CHAR: > case WM_SYSCHAR: > /* Note: VK_RETURN is recorded as virtual key ONLY */ > if(keyCode == 13 && !ctrl) return 1; // add "&& !ctrl" > pressCode = EventKeyChar; > break; > ... > } > ... > > Note: ENTER keyCode is 13. But when ctrl + ENTER is pushed, the keyCode is > 10. > So if the condition is revised, ctrl + m problem is corrected. > but other behavior is not changed. > > > If you want to prevent to generate character event for CTRL + ENTER, > you can add following code. By this way, we can produce > not only ctrl+m (0x0D)'s character event but also ctrl+j (0x0A)'s one. > > -- sqWin32Window.c -- > > // add this function > static int isNotKeyDownCtrlReturn(MSG* msgp){ > if(msgp->message == WM_KEYDOWN && msgp->wParam == VK_RETURN){ > if(GetKeyState(VK_CONTROL) & 0x8000) { > return FALSE; > } > } > return TRUE; > } > > ... > > int ioProcessEvents(void) > { static MSG msg; > ... > if(isNotKeyDownCtrlReturn(&msg)){ // add > TranslateMessage(&msg); > } // add > DispatchMessage(&msg); > ... > } > > ... > > int recordKeyboardEvent(MSG *msg) { > ... > > if(pressCode == EventKeyDown && virtCode != 0 && > isNotKeyDownCtrlReturn(lastMessage) ) { > // add " && > isNotKeyDownCtrlReturn(lastMessage) " > /* generate extra character event */ > ... > } > return 1; > } > > An attached file is source, which has been corrected by above way. > > > Note: Below is a comment in the source code. > > /* note: several keys are not reported as character events; > most noticably the mapped virtual keys. For those we > generate extra character events here */ > > Those keys(e.g. arrow) should not be ascii code. > ENTER key is not. but an extra character events is created for ENTER. > I don't think this behavior is correct. > As for ENTER key, an extra character event must not be generated, > and WM_CHAR message should be processed. > Please see MSDN about TranslateMessage. > > > see also Comment 11. > > Thanks, > > Attachments: > sqWin32Window.c 99 KB |
In reply to this post by pharo
Comment #15 on issue 1064 by [hidden email]: There are still problem with keyboard shortcuts in Windows http://code.google.com/p/pharo/issues/detail?id=1064 The workaround on Comment 11 is for only Windows. I do not know well how the workaround works on other platforms. Probably some shortcuts may not work. |
Comment #16 on issue 1064 by [hidden email]: There are still problem with keyboard shortcuts in Windows http://code.google.com/p/pharo/issues/detail?id=1064 'ctrl+.' doesn't also work. So I'm proposing a new solution. The issue of shortcut key such as ctrl+i is solved by fixing VM, without repairing an image file of Pharo 1.1.1. The workaround at Comment 11 is replaced to this solution. (1)Problem Thear are two problems. First, key character event for Ctrl-key combination which can not encode to ASCII control character is not generated. i.e. we can use only ASCII control character as Ctrl-key combo. But some Ctrl-key combo which is not ASCII control character are used in Smalltalk. In current Windows VM implementation, a key character event on Smalltalk is created based on WM_CHAR Windows messege. Windos does not generate WM_CHAR message for Ctrl-key entry which can not encode to ASCII control character. Second, when ctrl-key combo is pressed, ASCII control character is set to keyboard event imformation, but the key's character code is not set. Concretely, when '^A' is pressed, ASCII code '^A' is set to keyboard event, ASCII code 'A' is not set. The second problem is probably on only Windows VM, is not on other platforms. (2)Solution Create key character events for Ctrl-key combinations when WM_KEYCHAR is fetched, not when WM_CHAR is fetched. And then, by using ToAscii(or ToUnicode) Win32API function, convert virtual key code to ASCII code and set key's character to keyboard event object. I think this approach is can be used to Squeak VM. (3)Implementation I corrected "sqWin32Window.c", please see also the attached file. **** line 1060 in mapVirtualKey **** /* case VK_RETURN: return 13; */ **** line 1188 - 1236 new function **** void recordCtrlKeyCharEvent(MSG* msg, sqKeyboardEvent* evt) { BYTE keystate[256]; WORD wd; WORD asciiKeyChar = 0; WORD asciiCtrlChar = 0xFFu; sqKeyboardEvent* extra; WPARAM virtualKeyCode = msg->wParam; assert(GetKeyState(VK_CONTROL) & 0x8000); /* Exit when only control key is pressed. */ if(virtualKeyCode == VK_CONTROL) { return; } GetKeyboardState( keystate ); /* get ASCII control character */ switch(ToAscii( virtualKeyCode, HIWORD(msg->lParam), keystate, &wd, 0 )) { case 0: break; case 1: asciiCtrlChar = wd; break; default: return; } /* cancel ctrl key state */ keystate[VK_CONTROL] = 0; /* get key character */ if (ToAscii( virtualKeyCode, HIWORD(msg->lParam), keystate, &asciiKeyChar, 0 ) != 1) { return; } /* Exit when Ctrl + ENTER is pushed, because the combo is duplicate to ^N or ^M. */ if (asciiKeyChar == 13) { return; } /* generate extra character event */ extra = (sqKeyboardEvent*)sqNextEventPut(); *extra = *evt; extra->pressCode = EventKeyChar; extra->charCode = (asciiCtrlChar != 0xFFu ? asciiCtrlChar : asciiKeyChar) & 0xFF; extra->utf32Code = asciiKeyChar & 0xFF; } **** line 1276, 1304 - 1315 in recordKeyboardEvent **** /* if(keyCode == 13) return 1; */ ... if(pressCode == EventKeyDown) { if (virtCode != 0) { /* generate extra character event */ sqKeyboardEvent *extra = (sqKeyboardEvent*)sqNextEventPut(); *extra = *evt; extra->pressCode = EventKeyChar; } else if (ctrl) { /* note: when virtCode is not zero, ctrl+key combination (e.g. ctrl+arrow) are not created by recordCtrlKeyCharEvent. */ recordCtrlKeyCharEvent(lastMessage, evt); } } **** line 1586 - 1595 new function **** BOOL notCtrlDown(MSG* msg) { switch(msg->message) { case WM_KEYDOWN: case WM_SYSKEYDOWN: if (GetKeyState(VK_CONTROL) & 0x8000) { return FALSE; } } return TRUE; } **** line 1615- 1620 in ioProcessEvents **** /* Don't create WM_CHAR and WM_SYSCHAR message, when ctrl key is pressed. */ if (notCtrlDown(&msg)) { TranslateMessage(&msg); } DispatchMessage(&msg); } (4)Correct Pharo image file To work some Ctrl+key combo whose key is non-alphabet, image file should be corrected. e.g. below is a sample of correction for 'ctrl+.' on Pharo. handleEvent: evt "Store the event in the queue if there's any" | type cmdBit ctrlBit | "# change #" type := evt at: 1. type = EventTypeKeyboard ifTrue: [ cmdBit := 0. "# add #" ctrlBit := 0. "# add #" Preferences swapControlAndAltKeys ifTrue: [ctrlBit := (evt at: 5) bitAnd: 2] ifFalse:[cmdBit := (evt at: 5) bitAnd: 8]. "# add #" Preferences duplicateAllControlAndAltKeysSetting ifTrue: [ctrlBit := (evt at: 5) bitAnd: 2]. "# add #" cmdBit := (cmdBit bitShift:8) bitOr:(ctrlBit bitShift: 10). "# add #" "Check if the event is a user interrupt" ((evt at: 4) = 0 and: [((evt at: 3) bitOr: cmdBit) "# change #" = interruptKey]) ifTrue: [ Display deferUpdates: false. SoundService default shutDown. self handleUserInterrupt]. ^self ]. Thanks tn Attachments: sqWin32Window.c 100 KB |
Comment #17 on issue 1064 by [hidden email]: There are still problem with keyboard shortcuts in Windows http://code.google.com/p/pharo/issues/detail?id=1064 This attached file is a original "sqWin32Window.c" for comment 16. Thanks tn Attachments: sqWin32Window.c.original 99.0 KB |
Comment #18 on issue 1064 by [hidden email]: There are still problem with keyboard shortcuts in Windows http://code.google.com/p/pharo/issues/detail?id=1064 > Create key character events for Ctrl-key combinations when WM_KEYCHAR is > fetched, ... Sorry for my typo. It is WM_KEYDOWN, not WM_KEYCHAR. Thanks tn |
Comment #19 on issue 1064 by [hidden email]: There are still problem with keyboard shortcuts in Windows http://code.google.com/p/pharo/issues/detail?id=1064 This issue should be moved to the COG tracker |
Updates:
Status: Closed Comment #20 on issue 1064 by [hidden email]: There are still problem with keyboard shortcuts in Windows http://code.google.com/p/pharo/issues/detail?id=1064 Moved to the Cog tacker: http://code.google.com/p/cog/issues/detail?id=26 |
Free forum by Nabble | Edit this page |