|
Lion
2014-04-14 12:21
reporter
|
|
|
|
|
Push |
|
|
|
This is by design.
One of the buttons gets Cancel = True.
This is determined by the buttons you use.
However it is rather inconsistent with MessageDlg behaviour.
I will discuss this with the other devels to see if and how to fix it.
|
|
|
|
gtk2-qdlg.diff (947 bytes)
Index: lcl/interfaces/gtk2/gtk2lclintf.inc
===================================================================
--- lcl/interfaces/gtk2/gtk2lclintf.inc (revision 46316)
+++ lcl/interfaces/gtk2/gtk2lclintf.inc (working copy)
@@ -987,7 +987,7 @@
Btns := GTK_BUTTONS_NONE;
DefaultId := 0;
- CancelId := -1;
+ CancelId := idButtonCancel; //-1;
for X := 0 to Buttons.Count - 1 do
begin
DlgBtn:=Buttons[X];
@@ -997,10 +997,11 @@
if (Buttons.CancelButton=DlgBtn)
or ((Buttons.CancelButton=nil) and DlgBtn.Cancel)
then begin
- CancelID := X;
- ADialogResult := DlgBtn.ModalResult;
+ //CancelID := X;
+ //Always respond with mrCancel to VK_ESCAPE, so ignore CancelButton.
+ //This is consistent with PromptDlg behaviour
+ ADialogResult := mrCancel;
end;
-
if (ADialogResult = mrNone) and
(Buttons[X].ModalResult in [mrCancel, mrAbort, mrIgnore, mrNo, mrNoToAll])
then
|
|
|
|
qt-qdlg.diff (937 bytes)
Index: lcl/interfaces/qt/qtlclintf.inc
===================================================================
--- lcl/interfaces/qt/qtlclintf.inc (revision 46316)
+++ lcl/interfaces/qt/qtlclintf.inc (working copy)
@@ -176,7 +176,9 @@
if (BtnID >= Low(IdButtonToQtStandardButton)) and
(BtnID <= High(IdButtonToQtStandardButton)) and
(IdButtonToQtStandardButton[BtnID] <> QMessageBoxNoButton) then
- QtMessageBox.AddButton(Caption, IdButtonToQtStandardButton[BtnID], ModalResult, Default, Cancel)
+ //Setting AEscapeBtn parameter to False ensures the result will be mrCancel allways when
+ //user presses VK_ESCAPE, this is consistent with PropmtUser behaviour
+ QtMessageBox.AddButton(Caption, IdButtonToQtStandardButton[BtnID], ModalResult, Default, False {was: Cancel})
else
QtMessageBox.AddButton(Caption, ModalResult, Default, Cancel);
end;
|
|
|
|
win32-qdlg.diff (664 bytes)
Index: lcl/interfaces/win32/win32lclintf.inc
===================================================================
--- lcl/interfaces/win32/win32lclintf.inc (revision 46391)
+++ lcl/interfaces/win32/win32lclintf.inc (working copy)
@@ -387,8 +387,11 @@
try
Result := IDCANCEL;
TaskDialogIndirect(@TaskConfig, @Result, nil, nil);
+ //To be consistent with PromptUser, do not convert IDCANCEL value
+ {
if (Result = IDCANCEL) and Assigned(Buttons.CancelButton) then
Result := Buttons.CancelButton.ModalResult;
+ }
finally
RestoreApplicationState(State);
for i := 0 to TaskConfig.cButtons - 1 do
|
|
|
|
Attached possible fixes (so that it behaves like MessageDlg) for win32, GTK2, QT. |
|
|
|
Bart, thanks for finding the old issue report! I missed it - yes, it's a duplicate.
-----
Now to the patches:
I checked yours and mine and I think that my patch is better because in my patch the widgetset code still takes CancelButton into account. If somebody wants to create his own QuestionDlg with a custom CancelButton, he can still do it with my patch. No chance to do it with your patch because your patch ignores the CancelButton on the widgetset-level.
Do you agree?
(+ for consistency, CancelButton support should be added to Carbon as well.) |
|
|
|
Since there were no objections and nobody made any other proposal how to solve it, I am resolving it with my patch :)
|
|
