C / C++ / MFC

CAsyncSocket Client/server

9-Apr-04 3:50

Anonymous wrote:
What does this mean.

What does what mean? Are you saying that the option to set the stack size should be elsewhere?

Anonymous wrote:
What should the values be?

1MB is the default, and is satisfactory is most instances.

Anonymous wrote:
I also saw something like commit stack? What's this?

Per the docs:

Provides a space for you to specify the amount, in bytes, of physical memory to allocate from the reserve memory.

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

VBZ9-Apr-04 1:21

VBZ

9-Apr-04 1:21

hi,
I am in a gr8 trouble. I am making a client server application using ASync sockets with oracle as backend. Every thing is working fine but the problem is with closing the connection after it's use completes. i.e. when for the first time client connects to the server he is able to log in and send the data but after that is any client tries to log in then server gives the Assssertion failure error in line 177 winsock on ignoring the error the client works fine but if this happens server cannot log in the clients automatically. if there is a way to ignore this failuer or what i feel abt error is that once the socket is closed and after that if some body tries to give it gives a invaled socket error this was found out when i tried to debug the error. tell me what i can do to solve the problem. can this be due to the single m_sConnection socket i have used that when closed closes everything but there is another socket cvariacle ie m_sListenSocket that i have iused to accept connections ie m_sListensocket.accept(m_sconnectionsocket).
help me to understnd it

2G limit when creating threads

Anonymous9-Apr-04 1:17

Anonymous

9-Apr-04 1:17

Has anyone come across this before. Let say you create two threads with the create thread function. You specify 1 gig (I know you would never do this) for the stack parameter for each thread. Now you try to create a third thread. The stack value is let's say 10 MB. The function will not work. It won't create the third thread. This problem seems to happen on different combinations. Let's say you make 200 threads with a stack value of 100 MB each, then you add one more thread; this thread will not create.

Re: 2G limit when creating threads

David Crow9-Apr-04 3:53

Re: 2G limit when creating threads

9-Apr-04 3:53

Anonymous wrote:
Has anyone come across this before.

No, and I can't imagine anyone would under most circumstances.

You are encroaching on the 2GB limit. If you were using an OS that supported it, the /3GB switch would possibly get you access to a larger stack.

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

Gary R. Wheeler9-Apr-04 4:50

Gary R. Wheeler

9-Apr-04 4:50

This is a basic Windows constraint, in that they 'reserve' virtual addresses for the requested stack space for each thread, and they constrain the total amount of reserved space so that it fits within the 2 Gb process memory limit.

Taking a guess, you are implementing a multithreaded algorithm that is heavily recursive, with lots of local data at each recursive call (therefore the need for the huge stack). This is something I've seen in the artificial intelligence courses I took (a long time ago, I'll admit Blush | :O

).

The constraint here is that specifying a stack size S constrains the number of threads T, such that (S * T) < 2 Gb. The key to getting around this constaint is to move the data from the stack to the heap.

Here's an example. Suppose our original recursive function looks like this:

struct S {
    char BigBlobOfData[100000];
};
void F()
{
    S s;
    // do stuff with 's'
    //...
    // now call F recursively
    F();
}

Every recursive call deeper you go uses an 'S'-sized chunk of stack. Therefore, if you need to recurse N levels, you need a stack depth of at least (N * S), and this gets applied to your constraint for the number of threads. If you used the heap instead:

void F()
{
    S *s = new S;
    // do stuff with 's'
    //...
    // now call F recursively
    F();
}

the stack required for each recursive call is only the size of the pointer. Therefore, for the same depth of recursion, you require far less stack space. This lets you increase the number of threads.

The big caveat to all this is: you've only got so much memory available. Whether it's allocated on the stack or the heap, this approach doesn't address the fact that your algorithm is memory-intensive. It's just given you a little more flexibility.

Software Zen: delete this;

Service can't receive message from another service

tank10259-Apr-04 1:16

tank1025

9-Apr-04 1:16

Hi all,
I'm writing a WinNT service. My service prog will have to receive a user defined(or registered) message from another service based platform.but the based platform service is designed to send message to application(i think it has bind the application's desktop with SetThreadDesktop function). my service can not receive any message from it.
can anybody help me?what should i do to receive the message?

Re: Service can't receive message from another service

2249179-Apr-04 3:55

224917

9-Apr-04 3:55

did you set SERVICE_INTERACTIVE_PROCESS
for interacibg with desktop ...

greatest thing is to do wot others think you cant
suhredayan@omniquad.com

messenger :suhredayan@hotmail.com

Re: Service can't receive message from another service

tank10259-Apr-04 18:27

tank1025

9-Apr-04 18:27

yes,i had set SERVICE_INTERACTIVE_PROCESS,but can not receive.

CToolTipCtrl::RelayEvent

pie9-Apr-04 1:06

pie

9-Apr-04 1:06

Hello!

I've read that when working with the CToolTipCtrl, you have to either relay the mouse messages to it, or do some kind of subclassing. I've chosen to relay the messages, and I thought that the logical thing was to do this through OnMouseMove, and I did something like:

void X::OnMouseMove(...)
{
MyToolTip.RelayEvent(const_cast<msg*>(GetCurrentMessage()));
}

I used the const_cast because otherwise the compiler complained about GetCurrentMessage being const!

Well... Didn't work! Now I have found out that I must relay the messages in PreTranslateMessage, and that works nicely... But why can't i relay in OnMouseMove ?

Thanks in advance Smile | :)

Just a correction

pie9-Apr-04 1:08

pie

9-Apr-04 1:08

Damn html-tags!

void X::onmousemove(...)
{
MyToolTip.RelayEvent(const_cast<MSG*>(GetCurrentMessage()));
}

Re: CToolTipCtrl::RelayEvent

Sharif Ahmad7-Jan-15 2:18

Sharif Ahmad

7-Jan-15 2:18

Do it in

C++

X::PreTranslateMessage()

Hope it will work.

modified 7-Jan-15 8:19am.

How to run a function every minute?

StarMeteor8-Apr-04 22:53

StarMeteor

8-Apr-04 22:53

Dear All,

I am working on my project (in MFC) which require me to run certain function every minute...I would like to know how I can do so except using a loop to check if the time is reached. I would prefer a "signal and slot" like method as there are some other functions need to run so I can use loop method to do so Cry | :((

~_~

Thanks~

GflPower8-Apr-04 22:55

GflPower

8-Apr-04 22:55

OnTimer()

d

toxcct8-Apr-04 23:03

8-Apr-04 23:03

pliiiiiiiiiiiiiiiiizz

read the MSDN before asking such questions !!!!

OnTimer() of course !!!!!

D'Oh! | :doh:

TOXCCT >>> GEII power

2249178-Apr-04 23:30

224917

8-Apr-04 23:30

thats okey toxcct

they shld check codeproject b4 msdn.
tats our aim... Big Grin | :-D

greatest thing is to do wot others think you cant
suhredayan@omniquad.com

messenger :suhredayan@hotmail.com

SiddharthAtw9-Apr-04 1:09

SiddharthAtw

9-Apr-04 1:09

Yes. this is should be the first choice, even before MSDN. Rather microsoft should trigger it with F1. Laugh | :laugh:

2249179-Apr-04 3:33

224917

9-Apr-04 3:33

Thats sounds nice, to have F1 shortcut to codeproject.;)

greatest thing is to do wot others think you cant
suhredayan@omniquad.com

messenger :suhredayan@hotmail.com

toxcct9-Apr-04 4:06

9-Apr-04 4:06

download the CodeProject ToolBar so... Big Grin | :-D

TOXCCT >>> GEII power

ThatsAlok9-Apr-04 3:06

ThatsAlok

9-Apr-04 3:06

USE THIS
#define ID_timer 1000
use add window handler to deliver window message WM_TIMER
OnTimer

SetTimer(ID_timer,1000*60,NULL);

to end timer

KillTimer(ID_timer);

Alok Gupta
visit me at http://alok.stormpages.com

toxcct9-Apr-04 3:39

9-Apr-04 3:39

hey Mr Alok,

you mustn't call KillTimer() otherwise the timer associated function will be call once whereas it has to call and call each 60 second indefinitely...

besides that, you post repeats the ones ever posted, thank you very much ! Has somebody told here, let us sleep Sleepy | :zzz:

TOXCCT >>> GEII power

David Crow9-Apr-04 3:59

9-Apr-04 3:59

What are you saying? When a timer is no longer needed, of course KillTimer() should be called. Even though a few WM_TIMER messages may still be in the queue, the message-handling function can react accordingly.

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

toxcct9-Apr-04 4:05

9-Apr-04 4:05

yes, but Alok seems to calls KillTimer() at the end of OnTimer() Frown | :(

Maybe i'm wrong, but that's not clear...

TOXCCT >>> GEII power

David Crow9-Apr-04 5:10