|
Hello,
I have a semaphore controlled thread queue in my application. Now I have a problem to start my threads from the queue.
After call of _beginthread I get a thread-id not equal -1 and not equal 0, but my thread is not started (no trace message from thread, no break in breakpoint).
If I "help the application" with debugger "set next statement" to an other call of the _beginthread, I get an other thread-id (!= -1 && != 0), but thread is not started. In Taskmanager the threadcount is not inkremented.
Who can help me?
My application worked correcly under W2K. After compile the code with the .NET compiler the application works fine, besides on only one computer. But this computer is the machine, there must run the application 
Can it be a problem with not correct installed framework or libraries ?
An important remember is, that the threats will be started correcly after restart the application. After some thread calls (between 10 and 100) it doesnt work. All threads terminated via _endthread() and in a good time (1-300 seconds).
Thanks for any helping
STP
|
|
|
|
|
Post the declaration of the callback function. Post the call to _beginthreadex().
Kuphryn
|
|
|
|
|
Good morning kuphryn,
there are more than one threadfuncs. All of this have the same prototyp:
void ThreadFuncName (void *args);
The application have one timer for each thread to start, but no thread should be run, if there an other running thread. To solve this, I use an self written thread-calling-queue. If a timer call, the corresponding thread would be add to my queue. The queue use a simple list and each element is a struct like this:
typedef void (*ThreadFunc) (void *args);<br />
<br />
typedef struct _THREADLIST_STRUCT {<br />
ThreadFunc ThreadFkt;
long ThreadId;
_THREADLIST_STRUCT *next;
} THREADLIST_STRUCT;
And I have an other timer (1 second). He looks in me queue and is there a thread in and no other running thread, he starts the first element.
unsigned long t_id = 0;<br />
<br />
t_id = _beginthread (m_Root->ThreadFkt, 0, m_pParent);<br />
if (t_id != -1) {<br />
} else {<br />
/ do an error message<br />
}
Each access of the internal thread-list (AddThread, DelThread, StartThread, ...) is protected via a semaphore. Each thread remove himself from the queue at the end of thread and call _endthread().
As I already said, I works !!!, but after 10-100 thread-calls, it doesnt work again. The _beginthread () return a value != 0 and != -1 -> no error message. But the thread is'nt called. This effect exist only on one machine. The only one, there it should work in real life. This machine have:
- W2K with latest SP's and Updates
- complete installed Visual Studio .NET 2002 German (we have some problems with the compiler in the actual VS.NET !!)
- SQLServer ODBC driver
- pcAnywhere 10.5
Thanks
Stephan
|
|
|
|
|
Interesting.
I think the problem is m_Root points to an invalid memory address.
You do not need to call_endthread(). Let the thread end on its on. If you keep track of it via a handle, then close the handle.
Kuphryn
|
|
|
|
|
Why does the application jumps into the catch (B* ex) and not the catch (D* ex)?
The dynamic type of p is D and not B!
#include "stdafx.h"
class B
{};
class D : public B
{};
int _tmain(int argc, _TCHAR* argv[])
{
try
{
B* p = new D;
throw p;
}
catch (D* ex)
{
delete ex;
printf("catch (D* ex)\n");
}
catch (B* ex)
{
delete ex;
printf("catch (B* ex)\n");
}
return 0;
}
Regards,
Daniel.
--
FIND A JOB YOU LOVE, AND YOU'LL NEVER HAVE TO WORK A DAY OF YOUR LIFE. 
|
|
|
|
|
Object instance is of D but object type is B.
I'll write a suicide note on a hundred dollar bill - Dire Straits
|
|
|
|
|
Because p is not a D*. It is a B*, which happens to be pointing to an instance of D. Nonetheless, the type of the pointer has not changed.
In other words, it's catching B* because that's what you're throwing. If you want it to catch D*, then you should throw D*.
e.g.: throw static_cast<D*>(p);
It's worth noting that you cannot use dynamic_cast here, because B is not considered a polymorphic type. i.e. It has no virtual functions.
|
|
|
|
|
Hi !
Here is a problem I'm not sure what would be the right way to solve it. I have a class which represents an operation and this operation has a duration. I want to be able to read the duration of the operation with the function getDuration(). As computing the duration is a long process, I don't want to do it before it is asked for, and I don't want to recompute it each time it is asked for. The code below is basically how I implemented it :
class Operation
{
public:
Operation() { m_DurationCalculated=false;} ;
~Operation();
long getDuration()
{
computeDuration();
return m_Duration;
};
private:
long m_Duration;
bool m_DurationCalculated;
void computeDuration()
{
if (!m_DurationCalculated)
{
m_Duration=;
m_DurationCalculated=true;
}
}
};
It works fine except that I want now the function getDuration() to be const, because, from an external point of view, getDuration() doesn't change the operation.
The problem, when making getDuration() const, is that it calls computeDuration(), which, I guess, should be const as well, and as m_Duration and m_DurationCalculated might be changed in a const function, I guess I have to put the mutable keyword in front of these member variables.
This is where I'm not sure : is it the correct way to solve this ? It looks strange to me that computeDuration, which really is not a const function, should be const in order to be called by the getDuration function.
What is the correct way to solve my problem ?
Thanks for your help !
Jerome
|
|
|
|
|
If you dont want to change computeDuration to constant function, you can follow this trick:
<br />
.........<br />
long getDuration() const{ <br />
((Operation*)this)->computeDuration();<br />
return m_Duration;<br />
}<br />
........<br />
|
|
|
|
|
Actually, I thought about this trick, but I would like to make sure that it is a correct way to handle this problem and I also wanted to know what would be the recommended way.
Jerome
|
|
|
|
|
Casting off the const-ness is a nasty hack. It's also undefined if the object you're casting was originally allocated as a const object.
A cleaner way to do it is to declare computeDuration as a const function, and declare m_Duration and m_DurationCalculated to be mutable (i.e. they can be modified by const functions).
e.g.
void computeDuration() const
{
// safe to modify mutable variables
}
mutable long m_Duration;
mutable bool m_DurationCalculated
|
|
|
|
|
Thanks for your answer.
I have another question regarding the constness of 'computeDuration()' : if I make it const, each function called in computeDuration would also have to be const which, in my real application would be at least 4 or 5 more functions. These functions are clearly modifying the object, but would have to be const. This seems strange to me. Is it the correct way to do it ?
Thanks !
Jerome
|
|
|
|
|
Hmm. Sounds to me like you've got a design problem. If computeDuration() is calling functions which actually modify the object, then no, it should not be declared const. That means computeDuration() is indirectly modifying the object by calling those other functions. But by the same token, getDuration() is modifying the object by calling computeDuration(), and shiuld also not be declared as const.
Anything which changes the fundamental state of the object, directly or indirectly, should not be const. If a call to getDuration() is going to change anything fundamental, it shouldn't be const at all.
If it needs to be const for some specific reason, then I'd say you probably need to restructure the code and somehow separate the call from its dependencies on the other non-const functions. Otherwise, you'll have to take the other route and cast off the const-ness of the object, but I think most programmers will agree that it's poor style to do that. (Because if the call actually changes anything, it shouldn't be declared that it doesn't.) And additionally, it's undefined behavior if this happens on an object originally allocated as const.
|
|
|
|
|
Ofcouse, you should choose the way that you think it's the best.
|
|
|
|
|
|
|
Hai,
I need to know how to insert an JPG image in the Picture control in the MFC of VC++.
If Anybody knows Kindly reply me as soon as possibly...
bsujin
|
|
|
|
|
First using IPictureDisp or GDI+ library load the Jpeg Image.
and using the Exposed Bitmap Handle by above display picture in picture Control!
"Opinions are neither right nor wrong. I cannot change your opinion. I can, however, change what influences your opinion." - David Crow
cheers,
Alok Gupta
|
|
|
|
|
Hi all;
I have followed the instructions in MSDN for creating a bitmap button on a dialog box:
that is, I have drawn my bitmaps and named them accordingly with suffixes "U", "D", "F" and "X".
Then I have drawn my button on the dialog box, checking "Owner draw" in properties and with caption the name of the bitmaps without the above suffixes.
I have named a variable of type CBitmapButton in the dialog class and called the AutoLoad in 'OnInitDialog()' function as follows:
VERIFY(okBtn.AutoLoad(IDOK, this));
And I get an "ASSERT" error at this line:
ASSERT(FromHandlePermanent(hWndNew) == NULL);
What's the problem and what can I do?
Can somebody please help me??
_Lostris.
|
|
|
|
|
CBitmapButton can be cantankerous. You either need to call AutoLoad() before calling OnInitDialog(), or call LoadBitmaps() after calling OnInitDialog().
Have you tried Joe's button class?
"The pointy end goes in the other man." - Antonio Banderas (Zorro, 1998)
|
|
|
|
|
I've got a situation where we are running two redundant data servers where one should take over when the other fails. Our problem is in keeping the two servers synchronized when there may a thousand updates per second, and the order in which the updates arrive is very important and messages arriving in different orders to the two servers will cause problems. It is not a standard database (it's proprietary format), and everything is stored in memory. Basically, any data loss during the failover should be kept to a minimum. Some is allowed, but there are no guidelines on this. Less is better of course...
Obviously this problem has been overcome for large databases, but my search of google for algorithms has so far been fruitless. Does anyone have any information on synchronization algorithms that may help?
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"
|
|
|
|
|
Friends i want to display a 48*48 icon on the dialog window. For this purpose i first created the desired icon, Then i placed a picture control on dialog and given it, the path to my icon. But the problem is that, the picture control shrinks my icon and unable to display it in the original size of 48*48. Can anyone tell me abt the problem here?
Imtiaz
|
|
|
|
|
specify to the ressource editor that the icon is 48x48 instead of 32x32.
do this in the combobox up to the area you edit your icon in VC++. if this doesn't apear in the choices, click the button next to it.
TOXCCT >>> GEII power
|
|
|
|
|
I have taken a CView class and ripped it apart head to toe keeping only those few things I need for dynamic creation pretty much.
Why would IMPLEMENT_DYNCREATE cause a memmory leak???
The debugger is indicating a leak and points to this line
#ifndef CREATE_STATIC
IMPLEMENT_DYNCREATE(CMyOwnViewCls, CWnd)
#endif
What a bummer 
Any ideas???
How do I print my voice mail?
|
|
|
|
|
Did you close your app prematurely?
This macro defines the function that creates the window dynamically - something like return new CMyOwnViewCls. If you don't allow the app to exit normally, the window won't be deleted and the debugger will complain that this line allocated memory that wasn't freed.
Remember that everything in the macro expands to a single line, so everything in the macro is on the same line.
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"
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
Casting tricks are not the way to solve design issues.