C / C++ / MFC

GetMessage Hook and App Crashing

24-May-04 8:31

Go here and search for the PSD format.

"The pointy end goes in the other man." - Antonio Banderas (Zorro, 1998)

menu next to images

Member 466743724-May-04 6:54

Member 4667437

24-May-04 6:54

is there a way to place a drop menu nexto and image

User 1278224-May-04 6:50

User 12782

24-May-04 6:50

I have an MFC application that creates a WH_GETMESSAGE with a callback function in an external non-MFC DLL. It specifically monitors mouse and keyboard events. Event notifications are sent to my app with PostMessage from inside the DLL callback function. This in itself works perfectly. However, when I try to close the app, the app will crash or start to take up 100% cpu usage. If I right click the app in the task bar and choose close, I get 100% cpu usage, or if I then press 'C' for the 'Close' menu option, the app crashes. The X button will crash, double clicking the icon in the titlebar will exit without error, as will Alt-F4, and choosing Close from the system menu will crash.

It doesn't matter if I uninstall the hook in OnClose or not, because the DLL will cleanup on DLL_PROCESS_DETACH. I get the same result anyway.

I'm very confused, I have no idea what the problem is, its not consistant at all. I do know that if in the DLL I comment out the PostMessage lines in the event callback, there are no crashes or hangs at all. So I don't know if my app or dll is to blaim. Before the GetMesssage hook I had separate mouse and keyboard hooks, which didn't cause the CPU hang, but would occassionally crash.

Any suggestions?

Thanks!
-Aaron

modified 12-Jul-20 21:01pm.

jmkhael24-May-04 7:04

jmkhael

24-May-04 7:04

Maybe the dll shouldn't notify the application in case the application was closing? (infinite loop?)
Try unloading the dll from the application first

Papa

while (TRUE)
Papa.WillLove ( Bebe ) ;

User 1278224-May-04 7:57

User 12782

24-May-04 7:57

The problem is I want the DLL running the entire time the app is, without some sort of stop button being requried.

I was wondering if there may be an infinite loop, and I think you're right, that that was the problem. What I did is add a conditional to my hook callback that would only postmessage if the orginator of the message was not my window. And that works great! I don't really need the messages to be posted while they're in my app anyway.

Thanks!

modified 12-Jul-20 21:01pm.

Ryan Binns24-May-04 18:37

Ryan Binns

24-May-04 18:37

Aaron Stubbendieck wrote:
I have an MFC application that creates a WH_GETMESSAGE with a callback function in an external non-MFC DLL. It specifically monitors mouse and keyboard events. Event notifications are sent to my app with PostMessage from inside the DLL callback function.

Uh oh. That there is an infinite loop waiting to happen. Your app calls GetMessage(), which triggers your hook, which posts a message to your app, which calls GetMessage() to read it, which triggers your hook, which posts a message to your app...

See where that's gonna end up going? Smile | :)

You'll have to find some other way of notifying your app - look at shared memory or some other inter-process communication; messages won't work in this case.

I can't really explain why some cases didn't cause the crash, although it may be due to the "feature" of WH_GETMESSAGE hooks where messages retrieved via GetMessage() are processed by the hook, but messages retrieved with PeekMessage() are not. I never really understood why this was the case, but it's possible PeekMessage() may be being used for some of those cases.

Ryan

"Punctuality is only a virtue for those who aren't smart enough to think of good excuses for being late" John Nichol "Point Of Impact"

converting TCHAR to LPVOID

User 1278224-May-04 18:50

User 12782

24-May-04 18:50

Yeah, I eventually worked around that by not posting the message if it was from my app, that's as clean of a solution I could come up with. The odd thing is, that except when closing the app, there was no problem.

Thanks for the help on this thread and the other.

modified 12-Jul-20 21:01pm.

pnpfriend24-May-04 6:25

24-May-04 6:25

how can I convert TCHAR to LPVOID
here the example what I have,

<br />
CString str = "HELLO"<br />
TCHAR k[512];<br />
wcscpy(k,(LPCTSTR)str);<br />
LPVOID lpmsg = k;<br />

could run without any problem, but lpmsg only contain only one character, which is 'H'
how can I store CString content to LPVOID or TCHAR to LPVOID.

Thankyou

jmkhael24-May-04 6:46

jmkhael

24-May-04 6:46

Exactly like you did

If you want to see all the chars cast it back to a TCHAR*

Papa

while (TRUE)
Papa.WillLove ( Bebe ) ;

pnpfriend24-May-04 6:57

24-May-04 6:57

I dont' want to see all the characters back
I need to send this message to another application.
I have two applications, let's named them app1 and app2. They used pipe to communicate. So
app1 needs to send the message to app2 and here what app1 did,

<br />
CString msg = "sample text";<br />
LPVOID lpvMessage;<br />
TCHAR chBuf[512];	<br />
wcscpy(chBuf,(LPCTSTR)msg);<br />
lpvMessage = chbuf;<br />
int k = strlen((const char*)lpvMessage);<br />
<br />
fSuccess = WriteFile( hPipe,lpvMessage,strlen((const char*)lpvMessage) + 1,	&cbWritten,NULL); <br />

app2 gets the message sent by app1 but only the first character.
So what I need is app1 has CString conent and need to send that string to app2.
How can I send that string out to app2?

by the way, I'm using UNICODE

thank you

jmkhael24-May-04 7:00

jmkhael

24-May-04 7:00

How much is strlen((const char*)lpvMessage)??

and how are you reading the data in app2?

Papa

while (TRUE)
Papa.WillLove ( Bebe ) ;

pnpfriend24-May-04 8:26

24-May-04 8:26

strlen((const char*)lpvMessage) return 1

app2 reads as followed,
fSuccess = ReadFile( hPipe, chBuf,512,&cbRead, NULL);

chBuf only contains the first character of what app1 sent.

thank you

Dominik Reichl24-May-04 7:11

Dominik Reichl

24-May-04 7:11

A tip:

wcscpy(chBuf,(LPCTSTR)msg);
+
int k = strlen((const char*)lpvMessage);

Krrrrrsh...

Would be the first time ANSI-routines would work for Unicode strings Wink | ;)

Instead of using strlen, use _tcslen. Also, you shouldn't use wcscpy, but _tcscpy.

For the WriteFile function, use the length returned by _tcslen and multiply this size with sizeof(TCHAR).

Best regards,
Dominik

_outp(0x64, 0xAD);
and
__asm mov al, 0xAD __asm out 0x64, al
do the same... but what do they do??
(doesn't work on NT)

pnpfriend24-May-04 8:32

24-May-04 8:32

Yes.. k now get more than 1 length..
No.. app2 can't still read it.. it's all junk app2 got.
I don't know whether app1 sent the right information or it is just the converting?

<br />
LPVOID lpvMessage;<br />
TCHAR chbuf[512];<br />
_tcscpy(chbuf, (LPCTSTR)msg);<br />
lpvMessage = chbuf;<br />
int k =_tcslen((TCHAR*)lpvMessage);<br />

app1

<br />
fSuccess = WriteFile( hPipe,lpvMessage,_tcslen((TCHAR*)lpvMessage) *sizeof(TCHAR)+ 1,	&cbWritten,NULL); <br />

app2

<br />
	fSuccess = ReadFile( hPipe, chBuf,512,&cbRead, NULL);<br />

David Crow24-May-04 8:39

24-May-04 8:39

I suspect using LPVOID in app1 is the cause of the problem. Even though the second parameter of WriteFile() is a LPCVOID, this would be acceptible:

TCHAR chBuf[512];
_tcscpy(chBuf, msg);
WriteFile(hPipe, chBuf, _tcslen(chBuf) * sizeof(TCHAR), &cbWritten, NULL);

"The pointy end goes in the other man." - Antonio Banderas (Zorro, 1998)

pnpfriend24-May-04 9:05

24-May-04 9:05

No.. app2 still get the first letter o ftf whatever pap1 sent
is app2 might aslo has the problem? app2 doesnt' use unicode as app1.

<br />
fSuccess = ReadFile( hPipe, chBuf,512,&cbRead, NULL);<br />

thank you

David Crow24-May-04 9:49

24-May-04 9:49

pnpfriend wrote:
app2 still get the first letter...

How are you verifying this? In the Watch window, look at:

chBuf[0]
chBuf[1]
chBuf[2]
chBuf[3]
chBuf[4]

Or you might could try chBuf,su since it is a UNICODE string.

"The pointy end goes in the other man." - Antonio Banderas (Zorro, 1998)

pnpfriend24-May-04 10:49

24-May-04 10:49

I found out that
chBuf[0]=H,
chBuf[1]=0
chBuf[2]=E
chBuf[3]=0
chBuf[4]=L
chBuf[5]=0
chBuf[6]=L
chBuf[7]=0
chBuf[8]=O

why the characters are store in every other buffer?
Is it because of UNICODE? why UNICODE string does that?

thank you

David Crow24-May-04 10:52

24-May-04 10:52

pnpfriend wrote:
why the characters are store in every other buffer?
Is it because of UNICODE? why UNICODE string does that?

That's what UNICODE is. Perhaps you need an introduction.

"The pointy end goes in the other man." - Antonio Banderas (Zorro, 1998)

pnpfriend24-May-04 9:19