C / C++ / MFC

2-Jul-11 23:12

Hi. It looks like this: there is a worker thread, which just runs and there is a button, which when clicked, should send some data to this worker thread. And now i am thinking, how this can be done. First off, i am a little bit skeptical about pipes and global variables. More real-life example would be something like this:
1. There is a thread.
2. There is a dialog window which accepts files. When file is dropped on this dialog window, thread should be notified that new file has been dropped and receive an open handle to this file and a file name.
I was thinking about user mode APC but actually dont know if using this i can communicate with a thread or change its context, influence its variables, access its variables, modify them and use them - that would be great but the only thing i did so far with APC is terminating a thread.
Another scenario i can thinks of, is having some global structure, update it time to time, signal thread so it will know that our struct has been filled with some new info so the thread can grab it - but yet again, global variables - not so good.
Is there some "clear" and nice way of doing this?
Thanks

011011010110000101100011011010000110100101101110
0110010101110011

Rajesh R Subramanian3-Jul-11 3:57

Rajesh R Subramanian

3-Jul-11 3:57

Short answer: You'll use WaitForSingleObject[^] to wait on an event, critical section, or so (depending on your scenario). The event should be triggered (SetEvent[^]) when a file is dropped that needs processing.

csrss wrote:

Is there some "clear" and nice way of doing this?

If you're using MFC, you could use an UI thread[^] (the name is weird - an UI thread need not necessarily be associated with an UI). If you do it, you'll just post a message to the message pump of the thread and the thread will pick it off its queue and process it. This is an easy way, but it is an overkill if the thread is just supposed to do only one type of job (as opposed to doing different things based on the message posted to its queue).

"Real men drive manual transmission" - Rajesh.

csrss3-Jul-11 4:45

3-Jul-11 4:45

Thanks. Well, i need this for a simple test project - i did already something like this UI Thread, so here is nice and simple solution Smile | :)

1. Parameters

struct PARAMS
{
    int Val1;
    int Val2;
};

2. Thread

DWORD __stdcall Thread(LPVOID lpParam)
{
    MSG Message;
    while(true)
    {
	::GetMessage(&Message, NULL,  0, 0);
        switch(Message.message)
        {
            case SOMETHING:
            // .......... do something
            // Message.lParam will contain a pointer
            PARAMS *p = (PARAMS*)Message.lParam;
        }
    }
}

3. Notification

PARAMS *p = new PARAMS;
p->Val1 = ...
::PostThreadMessage(m_dwMyThread, SOMETHING, NULL, (LPARAM)p);

011011010110000101100011011010000110100101101110
0110010101110011

Rajesh R Subramanian3-Jul-11 5:46

Rajesh R Subramanian

3-Jul-11 5:46

csrss wrote:

DWORD __stdcall Thread(LPVOID lpParam)
{
    MSG Message;
    while(true)
    {
	::GetMessage(&Message, NULL,  0, 0);
        switch(Message.message)
        {
            case SOMETHING:
            // .......... do something
            // Message.lParam will contain a pointer
            PARAMS *p = (PARAMS*)Message.lParam;
        }
    }
}

The code is doing a GetMessage within an infinite loop, which doesn't look good to me.

I recommend that you have your thread code wait on an event instead, and set that event from outside the thread when the thread needs to get to work.

"Real men drive manual transmission" - Rajesh.

Albert Holguin3-Jul-11 7:32

Albert Holguin

3-Jul-11 7:32

I agree, this could end up using a lot of processor time doing nothing.

Mark Salsbery3-Jul-11 7:40

Mark Salsbery

3-Jul-11 7:40

GetMessage blocks until there's a message, so it's not bad like using PeekMessage right? Smile | :)

Mark Salsbery
Microsoft MVP - Visual C++
Java | [Coffee]

Rajesh R Subramanian3-Jul-11 8:47

Rajesh R Subramanian

3-Jul-11 8:47

It's not as bad, but it certainly looks clumsy to me. The op needs to just use one of the several ways of thread synchronisation. Smile | :)

Also, will there be a message queue associated with the thread? (note that the OP is calling it from within the worker thread).

"Real men drive manual transmission" - Rajesh.

csrss3-Jul-11 9:07

3-Jul-11 9:07

Hey. You guys are totally right, my way is quite dirty, but you know, sometimes you need some quick code to test things and thats the exact situation. World will never see this :P

011011010110000101100011011010000110100101101110
0110010101110011

Mark Salsbery3-Jul-11 7:44

Mark Salsbery

3-Jul-11 7:44

You don't really need to "pass data to a thread", as all threads in a process have access to the same data.

I typically use a semaphore to signal the worker thread when there's a job to do. The "jobs" are queued in a collection of some sort, and whenever a job is queued, the semaphore's count is incremented, setting its state to signaled. The worker thread waits on the semaphore and pops a job off the queue (don't forget to synchronize access to the collection with a critical section!) and processes it any time the semaphore is signaled.

That's for a single worker thread...If I need a pool of worker threads, I'll use an I/O completion port instead....it does all the above and saves writing alot of thread plumbing code.

Mark Salsbery
Microsoft MVP - Visual C++
Java | [Coffee]

csrss3-Jul-11 9:03

3-Jul-11 9:03

Hey Mark. I have never actually dealt with semaphores before, mainly events in a regular fashion as well as mutexes and now IOCP Big Grin | :-D

Currently just needed some quick and dirty solution for this test project of mine, because it will land in a recycle bin as soon as i'll figure a couple of things out - just needed to smuggle pair of pointers here and there though Wink | ;)

Anyways, thanks for a hints - will probably need that in near future.

011011010110000101100011011010000110100101101110
0110010101110011

LiangGuangLin3-Jul-11 18:01

LiangGuangLin

3-Jul-11 18:01

looking...

NMAKE Error

SNI2-Jul-11 5:10

SNI

2-Jul-11 5:10

Hi,

I got a project which is written in VC++ but while compiling its giving me following error in MS-VC 6.0.
"Microsoft (R) Program Maintenance Utility Version 6.00.8168.0
Copyright (C) Microsoft Corp 1988-1998. All rights reserved.
No configuration specified. Defaulting to MyProject - Win32 Debug.
NMAKE : fatal error U1073: don't know how to make '"e:\win32app\microsoft visual studio\vc98\include\basetsd.h"'
Stop.
Error executing NMAKE."

Please let me know which version of MS-VC6.0 will support this. OR do this project have 3rd party libraries. How to supress this NMAKE errors and complie the code successfully.

SNI

Richard MacCutchan2-Jul-11 6:43

Richard MacCutchan

2-Jul-11 6:43

I don't think you can 'make' a header file; check your project settings.

The best things in life are not things.

David Crow2-Jul-11 16:53

David Crow

2-Jul-11 16:53

SNI wrote:
NMAKE : fatal error U1073: don't know how to make '"e:\win32app\microsoft visual studio\vc98\include\basetsd.h"'
Stop.
Error executing NMAKE."

Please let me know which version of MS-VC6.0 will support this.

Have you verified that the file exists?

"One man's wage rise is another man's price increase." - Harold Wilson

"Fireproof doesn't mean the fire will never come. It means when the fire comes that you will be able to withstand it." - Michael Simmons

"Some people are making such thorough preparation for rainy days that they aren't enjoying today's sunshine." - William Feather

SNI3-Jul-11 5:41

SNI

3-Jul-11 5:41

Thanks for your reply. I was able to resolve this problem but now I am facing following error
".\Release\MyApp.exe : fatal error LNK1120: 2 unresolved externals
NMAKE : fatal error U1077: 'link.exe' : return code '0x460'
Stop.
Error executing NMAKE.
"
Do I need to change NMAKE file for this?

SNI

Re: NMAKE Error [modified]

User 74293383-Jul-11 5:47

User 7429338

3-Jul-11 5:47

Unresolved externals means you're missing source files or libraries. To fix this, you should add the missing files to the command link of the linker (link.exe).

modified 13-Sep-18 21:01pm.

SNI11-Jul-11 1:49

SNI

11-Jul-11 1:49

I tried to identify what is the problem for giving this kind of error and I found that one of function declared object of dialog class as follows
Dlg_Details *pDet = NULL;
and then it creates instance
pDet = new Details_Dlg();
and here it is giving following err
"error LNK2001: unresolved external symbol "public: __thiscall Dlg_Details::Dlg_Details(class CWnd *)"
(??0Dlg_Details@@QAE@PAVCWnd@@@Z)
.\Release\myapp.exe : fatal error LNK1120: 1 unresolved externals"

when i commented out above line of code (creating instance using new) then appln compiles and link properly.

I tried to go to the detail and found that the constructor of above class is declare as
"Dlg_Details(CWnd* pParent = NULL)" in header file and its implementation as in .cpp file is as follows.
"CDlg_Card_Details::CDlg_Card_Details(CWnd* pParent /*=NULL*/)
: CDialog(Dlg_Details::IDD, pParent)
{
//{{AFX_DATA_INIT(Dlg_Details)
....
....
variables are initialized
}

and I tried to changing its declaration in header file as
Dlg_Details(CWnd* pParent)
but still the problem persist. Can you suggest where this code going wrong during linking.

Regds
SNI

Regds
SNI

SNI

modified on Tuesday, July 12, 2011 8:20 AM