LabWindows/CVI app terminates under Win XP when file select popup is open, but everything works fine on Win 98

[awd]

David,

I replied to Wendy's post yesterday, and as a result it appears just below her reply to me (i.e. several messsages back from this one).  However, I also request that I get an e-mail if someone replies to my messages, but that does not happen, I have to keep coming back to look for myself.  Perhaps this is because I have my subscriptions to forums set up to mail me once a week?  Unfortunately I have looked back at Monday's batch, and this thread does not appear, but it does say that there were 63 messages that did not get sent to me, and I should therefore check manually.  I would have thought that it would be useful for correspondents to get e-mails back immediately, rather than waiting for a weekly batch - is this possible Wendy?

Thanks

Adrian

[Wendy L]

Hello awd,

So when you create a CVI distribution kit in CVI 6.0, it also distributes that version of the runtime engine.  The runtime engine is actually included in a merge module stored in your redist folder in the CVI install directory.  That is why you do not explicitly see a 6.0 version of cvirte.dll in your system32 directory, but instead you see the latest version of the DLL in that directory.  So when you run your application from CVI 6.0 and you have a more updated version of CVI on your machine, it is going to use the more recent version of the run time engine, but if you distribute the application, it's going to deploy the version of the runtime engine that is in your merge module.  Note that this behavior is version dependent due to changes in our distribution kit model, but the description above is true for CVI 6.0.

Now for the other questions in this thread ...

[Wendy L]

Hi David G and awd,

I was just able to reproduce the incorrect behavior on my machine!  As such, I filed this error with our CVI R&D team for further investigation.   Thanks for reporting this behavior, and for your detailed documentation on how to reproduce it. 

Thanks again and sorry for the inconvenience.

[Wendy L]

Hello awd,

So you can only have one subscription setting for thread/messages and boards.  As in, you will not be able to get a weekly digest on the board you are subscribed to and then get immediate notification for the thread/messages you are subscripted to.  Therefore, if you set your Subscription Email Delivery to Weekly Digest, you will get one email listing the first 50 replies to your board subscriptions.  Also, if you scroll to the bottom of that email, you should have notifications for your particular thread/message subscriptions.

The reason why you have a note in that email stating that 63 messages did not get sent to you is because only the first 50 messages for your board subscriptions are displayed in the email, and then you are notified of the number of remaining new messages that were posted on the forums.

I hope that helps clarify things.  Thanks again!

[David G]

Wendy,

Sounds good, thanks for your help.  Is there anyway you could post on here when a patch is available? And do you think I can expect that in the next week or two, or is this something that may take awhile to make it through your release process?  In the latter case I would be willing to try a beta version of the patch if that was possible for you to forward to me.

And I'm also kind of curious, once the patch is available, if it will also be included by CVI in any distribution kits created?  As awd mentioned, it seems the distribution kit may not contain the same runtime dll as what is in the Win/system32 folder?

Thanks,
David

[Wendy L]

Hello David,

I won't be able to nail down when a fix for the situation will be available but I have added your contact information to the error log so you should be contacted when we have updates on the issue.  Whether or not the fix will be included in future distribution kits is going to be dependent on if the fix is bundled into the runtime engine and also dependent on which version of CVI you have.  If the fix is strictly isolated to the runtime engine (which may or may not be the case), independent of which version of CVI you create the distribution kit in, you can download and install the latest runtime engine onto the machine you are going to deploy to and your application will automatically use this updated runtime engine.

Thanks.

[Erwin Timmerman]

Dear David and awg,

I have had problems with XP as well, in fact still do, with some fast computers.

Could you both maybe look at this thread and try out the very small test program I've attached there, to see if it crashes on your computer as well? And if the computers it crashes on is consistent with the computers your programs crash on?

Here's another thread with someone having a problem with menus freezing in XP, he solved ith with messing with the plug and play manager (look in the thread for details). maybe this will work for you too?

As for now, I deliver the computer with our test equipment, and the computer has W2k on it, because I can't get my CVI programs to work reliably with XP. Whether or not it is the fault of M$ for delivering a crappy OS or of NI for delivering a compiler with previously hidden bugs that XP now reveals is for myself irrelevant. It doesn't work, that's the bottomline...

Another thing I asked in another thread but never got an answer to is the following:

I have found that on a computer with a network, and a network disk which is connected but broken (you know, the small red cross under the drive character in the Windows explorer), the Multifileselectpopup function hangs when it is being called. The hour glass appears and before the popup window appears the program locks up, until the disk becomes available again.

I'm not sure this it a CVI issue, because Windows is searching for a disk it can't find, but it would be graceful if after something like 10 seconds the search was abandoned. What exactly is happening here, and how can it be avoided?

Which is also related to the file select popup window, so maybe it's apporopriate to post it here again.

I hope you guys get a better result from NI with your problem than I did with mine... so far. I didn't get an update that the problem will or will not be solved yet... any comments on it from NI would be apreciated...

Regards,

Erwin Timmerman

[awd]

Erwin,

I have tried your test program and it fails on some machines and is fine on others, but there is no consistency.  The machine where I discovered my problem runs your program indefinitely (P4 2,4), another that has my problem fails with yours after many tens of seconds (P4 2.4), my dual Xeon 2.4 runs yours indefinitely and was fine with my program, and an almost virgin XP installation on a PIII 450 fails, but was OK with my program.

I tried the experiment with disabling Plug and Play services on a laptop but that did not solve my problem.  I am reluctant to try this on any other machines in view of the warning that it could make the system unstable.

Adrian

[Wendy L]

Hello David and awg,

So after some digging, we were able to isolate this to a Windows XP issue.  Actually there is a long thread on a Google forum discussing this behavior.

In order to verify that you can see this behavior independent of CVI:
1. Open up Windows Notepad (Start >> Run >> Notepad).
2. Navigate to File >> Open and hover over a file on the desktop (to display the file information tooltip).
3. Cancel out of that File Open dialog.
4. Navigate to File >> Open again and hover over the same file on the desktop.  The file information tooltip should display again ... but instead, Notepad crashes. 

I've verified this behavior on multiple machines.  Luckily, we have a way to work around this issue in CVI.  If you are seeing this behavior on your machine, call the following function at the beginning of your application:

CA_InitActiveXThreadStyleForCurrentThread (0, COINIT_APARTMENTTHREADED);

This should correct the problem.  Let me know if you are able to reproduce this problem using MS Notepad, and if the following workaround addresses the issue.

Thanks again.

[LuisG]

Just to add one more data point to what Wendy already posted. This behavior seems to be introduced when installing Acrobat Reader 7.0. If you have an older version of Acrobat Reader the crash doesn't seem to be reproducible. This might explain why it was so hard to come up with a pattern for machine configurations in which this does and does not happen.

Luis