I'm opening a control using MSAA Server (Miscrosoft Active Accessibility) in order to automate tests over it using Coded UI. I'm using MSAA because the code is for VS2008 so UI Automation is unavailable.
It works good except for the case when there are two control elements on the same Window. The problem is the UI Map Hierarchy generated. Its wrong, but for the case when you are dealing with one control, works ok.
The problem using the example with MSAA example, found on sdk:
MSAA Server Sample from Microsoft
So, when I map the window, I get the right map, but when I map a element inside the list, the element is right, but isn't mapped inside the list:
Coded UI Map from mapping the events and asserts
I have exactly the same problem in my app.
Thank you
I've fixed that using a static std::map<string,int> to get an identifier when a new object is created.
So when a new object is created, I query the static structure with the current window title (I know this is a risky operation, but it works) and I get a new identifier.
I return the identifier when the MSAA method: IAccessible::get_accValue is invoked
I add the Value search property of coded-ui to that property and it works as I expeceted.
Instead of adding and answer to the Windows Message: WM_GETOBJECT, I've overrided the accessible methods from class CWnd.
The problem is the name returned by get CWnd::get_accName( VARIANT varID, BSTR *pszName) is the same in all chindrens of the window, so I've modified this method a bit in my class to add some extra information to identify my element.
And it works.
Related
I am working on a Windows application using MFC and Visual Studio C++ 17.
I have multiple tabs in the application, one of which is currently implemented using CListBox and needs to be reimplemented using CCheckListBox.
CCheckListBox is a child class of CListBox.
I have a vector of unique CString's I want to display in the box.
To start simple, I tried making the tab with CListBox and by using CListBox::AddString(), this worked exactly as I wanted it to. It sorted the strings alphabetically, and automatically added a scroll bar on the side when the vector had too many CString's to display all at once in the list.
However, when swapping the CListBox variable to a CCheckListBox variable, I have come across the error when I press run along the lines of:
Debug Assertion Failed! ..... \mfc\winctrl3.cpp Line: 588
I found this GitHub link that has the winctrl3.cpp file and on line 588, there is a function called OnLButtonDblClk which is somewhat explained here.
The trouble is I am unsure of how to swap a variable from a parent class to a child class. Is it so different for Windows classes? In the courses and programs I have taken and written in the past, switching from a parent to child variable was straightforward but this is quite the opposite !
Any help is appreciated, thank you in advance ! :D
There is no assertion in line 588 of the source file in your link. That github upload dates back to 2014. Why not search the winctrl3.cpp source file in your own Visual Studio installation instead? In my own installation, it is in function PreSubclassWindow() and there is indeed an assertion there:
// CCheckListBoxes must be owner drawn
ASSERT(GetStyle() & (LBS_OWNERDRAWFIXED | LBS_OWNERDRAWVARIABLE));
That is the CCheckListBox must be owner-drawn. Find the CCheckListBox constructor code above, you will notice that it calls the CListBox (parent-class) constructor, applying either of the two owner-drawn styles. Instead, the CListBox constructor calls CWnd::Create() with the "LISTBOX" Windows Class-Name as a parameter. That is, it seems to be a MFC-implemented control, overriding the CListBox class, with an owner-drawn style. There is no mention of a native "Check-ListBox" class or style in the Win32 documentation either. Also the CCheckListBox documentation (in the link you posted) clearly states: CCheckListBox is only for owner-drawn controls because the list contains more than text strings.
Therefore:
If you are creating the control using a resource script, add an owner-drawn style (eg LBS_OWNERDRAWFIXED) to the control (or set the Owner-Drawn property in the property editor).
If you are creating the control programmatically make sure you are calling the CCheckListBox constructor, not the CListBox one.
Not sure what you mean about the scroll-bar (and why you mention it), doesn't the list-box display a scroll-bar automatically (WS_VSCROLL style - or property), do you have to do something on your own?
EDIT:
Forgot to mention, if you are creating the control using a resource script use the Class Wizard to add a member variable (of "Control" type, not "Value") to the dialog class. The new variable must be of type CCheckListBox (MFC class). This will subclass the control. If the old project already had a CListBox member variable change it to CCheckListBox.
I searched on the forum and I don't find a solution for my issue. So I hope you can help me :)
I work on a personal project for SFC designing (Sequential functional chart) and I'm working with visual studio in SDI(I'm using MFC library). If you see the "design" of an SFC you can see the different elements needed to compose this. So you can find Step, Transition, and more. If I take a step for explaining my issue, after double click on the step a popup dialog is opened with the elements to define this step (actions on this step, and more). The issue is here, I can't see two or more step elements at the same time. I want to reuse the existing concept on other software, see this.
Step close
Step open
Dialog to add
My question is, how I can implement a dialog with my graphic element in mainframe (In this case, a step)? I don't know how I can insert a dialog with my element, I think I need to use CFormView, but I don't know.
This dialog needs to be resizable and reduce directly by the step graphic.
any idea?
Thank you in advance! Sorry for my English ..
Sorry, I think my request is not clear .. (Thank you for your answer)
The context, it's an SDI application with the Document/View architecture. Actually the view is derived from CScrollView.
So, in the document class, you have the different lists of components for make SFC (Steps, transitions and symbols ..). I'm working today on Step element.
The user inserts a new step, the step is draw on the view like this :
enter image description here
And now the user want change the events on this Step, for this after double click on the step the events editor is opened at right of step draw, like this :
enter image description here
For this, I've created a new dialog resource and create the class by wizard in CForwView derived class. In step attribute, you can find one instance of this derived class (The derived class of dialog).
But this doesn't work correctly, I think my method is bad. At the first try, I have sent the pointer of the current document to the "CFormView::Create" function for having the "Save" button active with the focus on the FormView. But after destroying the step, the instance of FormView is destroyed and the document with the instance of formview ...
No problem, you can use "Create" within CCreateContext a null pointer. But with the document or without I have a lot of problems (graphic design not correct in FormView, regularly (not systematic I have assert fail on Proc exchange (for differents reason)). But the "concept" is good, the editor follows the draw if I scroll, I can open or close the editor at any time and on any elements.
For the old capture, it's my SFC designer "model". My application is a complement to this application, so I want a similar design. I don't know how work my model application ..
On my application all is draw by MFC GDI, I don't use ActiveX or other tools.
So what is the correct way to implement one instance of editor by instance of step ?
For the implementation on this FormView I have :
- Make new dialog in ressource
- Create a derived class of CFormView with the created dialog
- Add one instance in attributes of Step element
- "OnDbClickOnStep" -> Call "Create" with the good position / size, pointer of mainview (in my case the CSrollView derived class)
- Done, FormView inserted in a mainview, I can edit my step events.
? Not done, I lost save button and other function linked to the document with the focus on a control in FormView. The app want a document with this view, how to override this ?
? Error in Proc exchange, for different reason...
You have an idea ?
You normally don't draw anything in the "main frame" (or the "MDI clild frames", in the case of a MDI application), this is done by the library, and imo sufficiently so. You display your data in a CView-derived class.
CView is the base class of all other MFC view classes. It's a simple graphical class - you need to paint it yourself in the OnDraw() member.
CScrollView is a descendant of CView, adding scrolling functionality (scroll-bars are automatically displayed if the scrollable area is bigger than the visible window area).
CFormView is a descendant of CScrollView, displaying a dialog resource-script, containing "controls" (edit-boxes, check-boxes, images, ActiveX etc).
As in your case the "controls" won't be initially known (except maybe for some few special cases) and rather programatically created, the resource script will most likely be empty, so using either CFormView or CScrollView will basically be the same. I would suggest starting with CFormView, and "downgrade" it to CScrollView if CFormView is not necessary or causes you troubles.
What are those "Step" items shown in the pics? ActiveX controls, child dialogs maybe? These work best as child controls on a dialog window. Are they already implemented, or they are just pics of some other software? Btw ActiveX is a way to define controls that can be used in other projects too, without having to include them in the project source.
I'm currently starting to learn how to use C++ Builder. However, I'm stuck on doing something basic, which is to open a window when I click on an element of the menu. I'm ok with the event management, but when I try to display it with the method Show(), it's written when compiling that "the method is not reachable" (I have it in french so I'm not sure about the exact translation). I've tried it different ways, also with the popup element, but I always get this message. Here is the short code that I use to display the window :
TFrame1 * NewPageFormer = new TFrame1(this);
NewPageFormer->Show();
delete NewPageFormer;
NewPageFormer = NULL;
Do you have any idea where the problem comes from?
Thank you
Try with:
TForm1 * NewPageFormer = new TForm1(this);
NewPageFormer->Show();
What you should Show() is a TForm (e.g. take a look at How do I open a new form with a button, using C++ Builder?).
Frames are combinations of components placed on a form-like object, which are considered a cohesive whole.
A frame (TFrame), like a form, is a container for other components. It uses the same ownership mechanism as forms for automatic instantiation and destruction of the components on it, and the same parent-child relationships for synchronization of component properties.
However a frame is more like a customized component than a form, so you cannot directly call the Show() method of a frame.
I have a question about an ActiveX Control I am developing. I have hashed my way through most problems but I am stuck at a real road block. This ActiveX is being imorted into other software so I have to have a good implementation of the Property Pages. I have managed to get some persistant properties working with one issue. When I make a change in the property page it is updated and persists as long as the application that is development app that is using it is open but it reverts back when I reload the app. I have narrowed this down to the fact that the development app doesnt realize the ActiveX has changed and therfore doesnt save. If i make a unrelated change and save the program all is good and the values persist as expected. I have tried everything and cant seem t get the application that imports my ActiveX to realize when it has changed via the property page. I am wondering if anyone has some work arounds for this type of problem. It seems to me if I could force the DoPropExchange() it would work but I dont know how to call this explicitly.
Thanks in advance
Matt
Thanks for the input. I did have the SetModifiedFlag() but it wasnt working. After poking at it I solved that problem but now I am redirecting my question. I had removed the property above in question from the idl files dispinterface in order to prevent the application that loads the activex from displaying the property in its "Connection List"(3rd party application specific"). It appears that the app loads all the interface into this list but this is not good as some properties should be persistant and only modifiable through the proppages. I tried all the flags like hidden and local but still were displayed in the list. When I removed it from the dispinterface it was how I wanted it but would not signal the IsModified. My new question is in there another way to define properties for an instance like this or is it possible to have a secondary interface(I have not tested if the application would see this interface because I am not sure how I would go about defining this inteface)or to have a property only between the proppage and control. Or is there another way to signal the dirty. The OnMemberVariable of the control was properly executing I know from testing it just seems SetModifiedFlag() doesnt do anything if the property is not in the dispinterface
Thanks Again
You need to mark your control as "modified", so that its host could detect it and re-save persistent properties. In this case IPersistXxx::IsDirty implemented by your control would indicate dirty state.
MFC based control has COleControl::SetModifiedFlag for this purpose:
Call this function whenever a change occurs that would affect your control's persistent state. For example, if the value of a persistent property changes, call this function with bModified TRUE.
Update: To hide a property from property browser you can use nonbrowsable attribute.
Use the [nonbrowsable] attribute to tag an interface or dispinterface member that should not be displayed in a properties browser.
I have a Window object which contains only a grid. I want to use Gtk::Builder to get a pointer to the grid, and then use some Gtk::Box's Gtk::Box->pack_end() to add the grid to it many times (with manipulated contents each time).
Though each time that pack_end() is called I get:
gtk_box_pack: assertion 'gtk_widget_get_parent (child) == NULL' failed in my terminal and nothing gets added to the box.
What should I do?
Thanks
* EDIT:
Goal:
I want entries of a DB table to be put into a fancy widget for each record, though all the records being shown vertically one after the other. I thought I can create the fancy widget as a window in Glade and use Gtk::Builder to get a pointer to it. So in the fancy's Glade file I have a window containing a grid that has my custom appearance. I get the above error when I try to add the pointer to the fancy *grid*, to the visible window's Box. I hope I'm clear.
Here's the solution to gtk_box_pack: assertion 'gtk_widget_get_parent (child) == NULL' failed:
All that needs to be done at the first place is that you should draw the widgets WITHOUT a window, so when loaded with builder, it won't have a parent and thus the assersion succeeds.
Though here's another point: When I add the first instance of the grid to the Box, the second one results in the same error again. After a couple of trials and errors I realized that in each interation you should use Gtk::Builder::create_from_file() to create a new parent-less instance of the grid to be able to use, and this way it works.
There has to be a great difference in performance, in case number of records is gonna be big, but Gtk::Widget's copy constructor is private and direct copying is not possible, and since it wasn't my main obsession I didn't insist on resolving this "performance" issue.