I am experiencing a very weird behavior while running my Active-X control on “Windows 8” OS, and browser “Microsoft Internet-Explorer 10”.
The control:
My Active-X control is very stable product that my company is already distributing for several years between hundreds of happy customers. The control is an ATL custom-draw control, written in C++, and built with “Visual Studio 2008”.
The problem – the drawing disappears:
While I press the link that opens my control for the first time, I can see that my control is being painted gradually (as my data loading is sometimes slow), but eventually all the drawing disappears, leaving a blank white background color where the controls is located.
Using ‘Spy++’ I can see that my control is really there, but all its painting disappear.
If I move the browser or change its size the control will be re-painted, but otherwise it is simply not visible.
I tried to debug this problem and find out who is erasing my drawing.
I came up with the following technique:
Using WinDBG I added breakpoints at ‘user32.dll’ DLL on the following methods: user32!InvalidateRect, user32!InvalidateRgn, user32!FillRect,
I defined these methods to only show stack and continue with to run,
I ran the above problematic scenario,
When the problem happens I am trying to view who called ‘Invalidate’ or ‘FillRect’,
Since all this happens very quickly I am using the ‘Print Screen’ button to freeze the moment
After doing this several times I find that exactly at the time my control disappears the following stack calls ‘user32!FillRect’:
user32!FillRect
MSHTML!memcpy+0x9805
MSHTML!COmWindowProxy::SwitchMarkup+0x468
MSHTML!CMarkup::SetInteractiveInternal+0x428
MSHTML!CMarkup::RequestReadystateInteractive+0x98
MSHTML!CMarkup::BlockScriptExecutionHelper+0xde
MSHTML!CHtmPost::Exec+0x794
MSHTML!CHtmPost::Run+0x1c
MSHTML!PostManExecute+0x5f
MSHTML!PostManResume+0x7b
MSHTML!CHtmPost::OnDwnChanCallback+0x3a
MSHTML!CDwnChan::OnMethodCall+0x19
MSHTML!GlobalWndOnMethodCall+0x169
MSHTML!GlobalWndProc+0xd7
user32!InternalCallWinProc+0x23
user32!UserCallWinProcCheckWow+0x100
user32!DispatchMessageWorker+0x3ef
user32!DispatchMessageW+0x10
IEFRAME!CTabWindow::_TabWindowThreadProc+0x981
IEFRAME!LCIETab_ThreadProc+0x378
iertutil!CIsoWinMsg::PostQueuedMessagesToComponent+0x4b
IEShims!NS_CreateThread::DesktopIE_ThreadProc+0x66
KERNEL32!BaseThreadInitThunk+0xe
ntdll!__RtlUserThreadStart+0x72
ntdll!_RtlUserThreadStart+0x1b
This is consistent.
When examining this stack I found that this message belongs to a window of class Internet Explorer_Hidden, which is window of size zero (0x0), but with a WS_VISIBLE style. This window belongs to the browser.
Does anyone know this phenomenon?
Is it possible that the browser hidden window is causing this problem?
Could you recommend on other approach to try and hunt the stack that is erasing my drawing?
Many thanks for any hint!
Paz Offer
Conclusion:
The bug was in Microsoft Internet Explorer 10
I used Microsoft Support service to farther investigate this problem. Enclosed here are the findings:
Microsoft Support viewed the version of my IE (Internet Explorer) and suggested to first upgrade IE to its latest patch.
As it happens, the machine I was working on was not setup to be automatically updated by the Windows Update service for some time. As part of upgrading IE I had to first update windows using Windows Update service to its latest patches.
During the process of Windows Update the Microsoft Internet Explorer 10 was automatically updated to its latest patch.
Once IE updated the problem disappeared.
In addition, the ActiveX GUI load process became much faster.
The old version of Microsoft Internet Explorer 10, which caused the problem:
Version: 10.0.9200.16384
Update Versions: RTM (KB2718695)
The new version of Microsoft Internet Explorer 10, which solved the problem:
Version: 10.0.9200.16635
Update Versions: 10.0.7 (KB2846071)
Now all is working as expected.
Thanks, Pazo
Related
I am creating calculator VBO in blueprism object studio.
In application modeller I gave calc.exe path and able to launch calculator, but I am not getting identify option to spy on any element..
In Element, After clicking Launch buttion, identify is appearing for a split of sec and Launch button is coming back instead of Identify
PS- As mentioned in videos and links, I have launched calc via application modeller not directly via windows..
I am using windows 10, BluePrism V.5.0.11.0 versionenter image description here
Windows 10 calculator is complicated case, please consider using different application in your training. For example paint is much simpler to get started.
If you really need to get that going, then you may need to launch application first, using one object, and afterwards use second object to attach and interact with it.
can be two reasons
1. Either your application is not launched properly.
2. Sometimes we face problems with some versions of applications.
The best way to fix your problem is
Edit your settings remove the application path from the navigate stage.
Launch the application separately
Open Navigate stage > Action > Attach
Provide the window title within quotes (example: "Calculator")
Click OK then run the object
Now if you go to application modeller
You will be able to see the identify button under the element.
Hope this will help you.
I resolved this by unchecking "Disable invasive techniques (hooking)".
The Windows 10 calculator runs with process name win32calc which is also located under c:\windows\system32.
So when you launch the application please make you identify the applications process name and its location
I resolved this by circumventing to windows 7 calculator instead of a windows 10 calculator on windows 10.
Link to download win 7 calculator for win 10-
https://winaero.com/download.php?view.1795
This will be installed in the same place as your default calculator.
Inside-
C:\Windows\System32
The name of the app in my case was "calc1.exe" to avoid collisions with the default calculator.
I used this as my base reference app for spying and it worked.
Also note, the hover to highlight button seems to be a bit buggy, try a bit hit and trail to get you required button to be highlighted and mark using "Ctrl+LeftClick"
Reading the book "The Spec UI framework". Trying to implement the part described in chapter "Taking over the entire screen".
After executing the code suggested in the book:
WindowExample new openWorldWithSpec
it seems to be impossible to return to previous state. Tried to delete the new WorldMorph in the inspector. Also tried with halos, as the book suggests but those buttons that are available in the halo menu don't allow it to be closed.
Is it an intended behaviour for this (to be executed to prepare an end-user environment and disable programming UI) or am I missing something?
Working in Pharo 5.0, Mac OS X version.
That is indeed intended behaviour, as part of making applications that do not allow access to the development environment anymore. But you can take a look at the code for PharoLauncher to see how you can enable a developer mode
You can actually get back to normal by:
Alt-Ctrl-Clicking on the Morph (macOS combo may be different, this is for Windows).
Clicking on the little wrench and ask for inspect.
In the evaluation pane, do:
self delete
World menu is available again.
Open a browser.
Done.
If one disables halos in settings, this is a tad harder.
I have written a VB application in VS2010. When I try to record its operation for a demo video with Windows Media Encoder x64 or Apowersoft free screen recorder, I get a recording of everything on the desktop except my application. It's like the application is invisible to the recorder. It has worked in the past (2 weeks ago). It must be something I have done but I'm baffled. Is there a "feature" in VB that can prevent you from recording the Form?
BTW, a still screen capture works fine and sub forms I bring up are captured. Only the main form is invisible to the recorder.
Thank you for your help and time.
Not a VS or VB problem after all. What I found was that windows 7 was switching me from Aero Theme to Windows 7 Basic. The message that showed up was that I was low on display memory. (A quick pop-up in the bottom right corner that I was ignored during the recording). However, I have 4.75 G of display memory?
BTW on a video conf call it produced the same invisibility "feature" until I found the Theme switch.
I haven't actually fixed anything but now know to watch for the tell tale screen blink as it switches mode. I pause the recording, and switch it back before restarting the recording. Got through the recordings I needed for now. I was using MS Media Encoder. I have now replaced it with Adobe Premier elements but haven't tried it yet to see if I still have the problem. Need to get the "how to" videos out the door first.
Thanks again. I have read many of your Answers to others in the past that have helped me solve numerous other VB and VS problems.
Michael
That's weird since it's recording forms from the application and not other forms (main form) from the same application.
Make sure you run the recorder as administrator and not the application.
I'm developing a website with ASP MVC 3 and Razor in VS2010 SP1.
When I'm running the site from VS, if the view was open in the IDE, I can edit a view in runtime, and press F5 to show the changes, but if I open the view after run the site, then the view is locked (no message, simply I can write in the VS Editor)
When I stop debuggin the solution, the views that are locked remains locked, and I have to close the file and open id again in order to modify the code.
This behaviour is uncomfortable. Can anyone tell me why sometimes I can write in the VS2010 editor and other times I can't?
Other Notes:
The TFS is not the problem.
It happens in all the VS 2012 SP1 of my office.
It only happens in vb.net projects, not in C# proyects.
It didn't happens in VS2012.
I've disabled all the extensions.
Thanks.
I have experienced this as well.
I've found that just reopening the file, while in debug, works most of the time. The nearest thing I can figure is that you're not actually editing the file but actually viewing a debug version of the file.
This normally happens, for instance, a javascript method causes an exception within the source of that page. It then loads the active version of that page instead of the one you're editing. Though I've had it happen without any errors generated so I'm fairly certain that's not the whole issue. It might be worth posting an issue on http://aspnet.codeplex.com/workitem/list/basic It hasn't happened often enough for me to narrow it down but it sounds like it happens for you more often.
After more than a year suffering this problem, a solution was found.
This MS HotFix for VS2010SP1 solved the issue.
http://connect.microsoft.com/VisualStudio/Downloads/DownloadDetails.aspx?DownloadID=40811
I have a virtual machine (VMware) with Mercury Quick Test Professional 9.2 installed. I have a script to test an application, written in VB.NET using the Infragistics library.
If I access this virtual machine using my laptop (using Remote Desktop), everything works fine, the script completes without a problem. My laptop runs XP, with Windows Classic theme.
If I access this virtual machine using another machine (using Remote Desktop), the script starts fine, but stops halfway through, without no error message from QTP, nothing. This machine runs XP, with Windows Classic theme.
One difference between the two setups is the size of the screen, the laptop is 1920x1280, other machine 1280x1024.
The step where the script stops involves checking a checkbox within an UltraWinGrid. The checkbox itself is displayed, is on the screen in both cases.
Has anyone had this problem before, or have any idea why the behaviour is different between the two machines?
Thanks.
OK. I've found the problem. In fact, the script was failing silently because that's what the person who wrote the script told it to do. It couldn't validate something which was off screen, so the script failed.
The problem was the QTP definition of 'off screen'. I have two screens attached to my laptop, the screen for the laptop itself (1920x1200) and another screen (1280x1024). I connect to the VM for QTP using remote desktop, and it uses the settings of the screen for the laptop. This means that when I launch my QTP script, and move it to the other screen, it doesn't fit, so the screen is no longer maximized, and the object is partially off screen, so can't be found.
The fix is simple: in the Remote Desktop, use the Display tab, and set the size of the screen to a size to 1280x1024, and QTP doesn't have any more problems.
Voilà.
If you are not using Expert Mode, and / or are allowing QTP to do most of the work to create your repository objects, then yes it is referencing everything by pixels.
I create all of my repository objects by hand, viewing the source (in the case of automated web-application testing) and using the Object Spy for assistance where needed. I make a point to not have any positioning information as part of my object definition, for the very reason you are running in to.
For the parts of my web-app that interacted with Windows (opening a file to upload, etc.) the Object Spy was essential for the trial and error necessary to create a unique identifier for creating the repository object. But it can be done.
Ex1: File Browse Dialog
text = "Choose file"
nativeclass = #32770 (apparently some Windows VooDoo for a file open dialog?)
Ex2: Filename textbox in Browse dialog:
nativeclass = "Edit"
attached text "File &name:" (more Windows VooDoo? It woudn't work for me without the "&")
Ex3: Open Button in the dialog:
text = "&Open"
object class = "Button"
Good Luck!
Point of clarification: You mentioned that QTP stops with no error message. Does that also mean that the test results log file also has no error message? If the log has any information, that may be helpful in diagnosing the problem. Could you share the lines of code at the point where the script fails?
Also, remote desktop will resize the desktop on the remote machine. Although QTP scripts are not inherently coordinate based, individual statements can be coordinate-based relative to an object. The resolution could be an issue in that regard. For example, imagine you had a line like Button.Click(5, 150) recorded on a higher resolution machine. But if you attempted to play it back on a lower resolution machine, and the 150 is out of bounds of the object on the lower resolution, it could cause an issue.
QTP does not use screen coordinates except as a last resort, if the objects are identified as high level objects (SwfTable in this case) you should be OK, if however QTP doesn't recognise the object it falls back to WinObject and screen coordinates.
If you're using Infragistics then you should know that they extend QTP's support with their TestAdvantage product which will probably solve your issue.
Edit:
#MatthieuF said:
In fact, we use the Infragistics plugin for QTP, and we still have the problem
Can you give me an example of a line that fails?
A few things:
You should be able to debug on the VM easily - just wait for it to stop, go into your object repository, and see if it can identify the object. If no then use object spy to figure out what properties are different between the OSes. If there is a difference then you can always set that property to a regular expression and have it check for both possibilities.
Assuming that isn't the issue we've run into problems using remote desktop with QTP if the Remote window is closed or minimized. For us, it was an issue where the clipboard can not be changed when an RDP window isn't visible, but there could be other surprises when using QTP that way.