C / C++ / MFC

Neville Franks31-Jan-03 10:24

Neville Franks

31-Jan-03 10:24

Have you checked that the IDC_xx values don't conflict with any standard Windows IDC_xx values. I assume the CWnd::OnSysCommand() is a Restore or Maximize command. Then you get OnSetFocus() which is fine. But why are you then getting OnCommand(). Have any of the controls been disabled.

Neville Franks, Author of ED for Windows. www.getsoft.com
Make money with our new Affilate program

Mike Eriksson2-Feb-03 23:16

Mike Eriksson

2-Feb-03 23:16

Hello Neville!

Thanks for your answer!

I recieved an answer from an another source and the problem seems to be with the checkboxes and tab order. If I have the check boxes as the first tabs, i.e. The IDC_AUTOMATIC is tab 1 and therefore default focused. It will generate an ON_COMMAND IDC_AUTOMATIC. If I change tab order and let an editbox get the first tab then everything is OK!

Strange Confused | :confused:

Mike Eriksson
Sonork 100.21825

Neville Franks3-Feb-03 9:27

Neville Franks

3-Feb-03 9:27

Mike Eriksson wrote:
I recieved an answer from an another source and the problem seems to be with the checkboxes and tab order. If I have the check boxes as the first tabs, i.e. The IDC_AUTOMATIC is tab 1 and therefore default focused. It will generate an ON_COMMAND IDC_AUTOMATIC. If I change tab order and let an editbox get the first tab then everything is OK!

This doesn't make any sense to me. You should see an ON_SETFOCUS but not an ON_COMMAND. It sounds like someone is sending an ON_COMMAND when they shouldn't be. Do you do any SendMessage( .., WM_COMMAND, .. ) in your code.

Also I'd suggest you use Spy++ or WinsSpector and look at precisely what messages the IDC_AUTOMATIC checkbox is getting.

Neville Franks, Author of ED for Windows. www.getsoft.com
Make money with our new Affilate program

Mike Eriksson3-Feb-03 22:08

Mike Eriksson

3-Feb-03 22:08

Neville Franks wrote:
This doesn't make any sense to me. You should see an ON_SETFOCUS but not an ON_COMMAND. It sounds like someone is sending an ON_COMMAND when they shouldn't be. Do you do any SendMessage( .., WM_COMMAND, .. ) in your code.

No, I don't use SendMessage(). All message handlers are created with ClassWizard.

Also I'd suggest you use Spy++ or WinsSpector and look at precisely what messages the IDC_AUTOMATIC checkbox is getting.

I will try this, although since the program now works it's nothing I will prioritize

Thanks for your answer, I appriaciate it very much!
Big Grin | :-D

Mike Eriksson
Sonork 100.21825

Changing registry key

Rage31-Jan-03 2:12

Rage

31-Jan-03 2:12

[This post follows one I've posted earlier]

I'd like to be able to read and write the following registry entry (NT System):

HKEY_CURRENT_USER\Control Panel\International\sDecimal

I do not know how to access it. If this can help, i know that this entry is actually in the hive %User%.dat where %User% is the current user logged on the PC. File is in Win\Profiles\%User%. But i do not know how to read this file, neither how to change it (since it is currently in use).

::SetRegistryKey(KEY) and correspondign functions Get/WriteProfileInt/String only allow to change the entries contained in HKEY_CURRENT_USER\Software\KEY ...

So ... any idea ?

~RaGE();

Rickard Andersson²⁰31-Jan-03 2:24

31-Jan-03 2:24

Rage wrote:
::SetRegistryKey(KEY) and correspondign functions Get/WriteProfileInt/String only allow to change the entries contained in HKEY_CURRENT_USER\Software\KEY ...

Yes, the Get/WriteProfileInt/String functions do that... but not:

RegOpenKeyEx(), RegQueryValueEx() and then RegSetValueEx() can do what you want!
Look up these in MSDN Library and you'll find more such functions.

Rickard Andersson@Suza Computing
C# and C++ programmer from SWEDEN!
UIN: 50302279
E-Mail: nikado@pc.nu
Speciality: I love C#, ASP.NET and C++!

Rage31-Jan-03 5:06

Rage

31-Jan-03 5:06

Thanks a lot. Incredible I've never came across these functions, since I'Ve been searching for registry, key, ... the whole day through MSDN. Roll eyes | :rolleyes:

~RaGE();

Rage31-Jan-03 5:53

Rage

31-Jan-03 5:53

Well, actually, i cannot open this key.

here's my code : what's wrong ?

PHKEY kHandle=new HKEY;
LPCTSTR Key="Control Panel\\International";
LPTSTR b;
PLONG d;

long int yop=RegOpenKeyEx(HKEY_CURRENT_USER,Key,0,KEY_QUERY_VALUE,kHandle);
if (yop==ERROR_SUCCESS)
{
     RegQueryValue(*kHandle,"sDecimal",b,d);
}

yop comes up with 998.
Could you provide me with an example ? The key was

HKEY_CURRENT_USER\Control Panel\International\sDecimal

Thank you very much

~RaGE();

xxhimanshu31-Jan-03 19:12

xxhimanshu

31-Jan-03 19:12

hi,
this is done like this..I think you are doing right..
except.. in last argument pass address of kHandle..like this and use it like..
RegOpenKeyEx(HKEY_CURRENT_USER, _T("software"), 0, KEY_WRITE|KEY_READ,
&kHandle);
I hope this time it works..

cheers

Himanshu

Client/Server advice please!

Abbas_Riazi1-Feb-03 3:36

Abbas_Riazi

1-Feb-03 3:36

Take a look at my article: Registry List Control[^].
You can find many useful codes and idea!
A. Riazi

Rickard Andersson²⁰31-Jan-03 1:52

31-Jan-03 1:52

I'm here again asking for advice for my server that I'm developing.
I'm using blocking sockets and of course it's multithreaded.

I put all my accepted clients in a std::list object and when a client send a message to all clients (or just one), should I always check the list if there is any disconnected clients and then remove them from the list before I send the messages to all clients in the list?

Should I have a timer or something that updating the list every n seconds or would that slow down the system?

Short say: How often or when should I update my std::list?

Rickard Andersson@Suza Computing
C# and C++ programmer from SWEDEN!
UIN: 50302279
E-Mail: nikado@pc.nu
Speciality: I love C#, ASP.NET and C++!

valikac31-Jan-03 5:51

valikac

31-Jan-03 5:51

I highly recommend that you redesign the server using a robust non-blocking I/O such as Overlapped I/O and Completion Port I/O. Under non-blocking I/O, the OS will notify the server the state of the sockets.

In terms of solution for this problem, there is no elegant solution. One solution is to process all sockets and determine their states. The time interval depends on the type of server.

Kuphryn

Rickard Andersson²⁰31-Jan-03 6:47

31-Jan-03 6:47

kuphryn wrote:
non-blocking I/O

I don't understand non-blocking sockets but I want to.
There is one problem, I have no idea where to read about non-blocking sockets!
So if you could give me a name of a good book or a link to a good and big tutorial, would that be nice! Smile | :)

Rickard Andersson@Suza Computing
C# and C++ programmer from SWEDEN!
UIN: 50302279
E-Mail: nikado@pc.nu
Speciality: I love C#, ASP.NET and C++!

Rohit Sinha31-Jan-03 7:19

Rohit Sinha

31-Jan-03 7:19

There are several articles on overlapped I/O and I/O completion ports on CP. Look in the internet and networks section. Non blocking sockets, while good, will not tell you whether a client is connected or not (unless you do something else to your code, but then you could do something similar to your existing code too).

Overlapped I/O is simple really. Instead of thinking in terms of threads, clients or sockets, you need to think in terms of events (read, write, send, recieve, etc). Once you get the hang of it, you will find it is more appropriate in many situations. On the other hand, just because non blocking sockets exist, does not mean you should use them in every situation. It takes much longer to implement them, and the gain is not always worth it. Only if your server is always busy, will you see an advantage.

In the mean time, while you are reading up on non blocking sockets, you could try this:

Keep a timer for all your sockets. Every receive resets the timer. If the timer is fired, a function will check to see if the client is disconnected. If it is, the client will be removed from the list. If not, the timer will be reset.

This will only be advisable if there aren't too many clients at the same time. Otherwise a lot of resources might be used up just for the timers. If there are too many clients connected, you may consider overlapped I/O.

Regards,

Rohit Sinha

Moak1-Feb-03 3:04

1-Feb-03 3:04

Hey Rickard, I do it similar in my chat:

If a client socket gets a disconnect event, it shuts down and then marks itself as dead. One regular server timer will check for dead sockets (in a std::vector or std::list) and remove them from the list. I dont like if objects delete themself on their own, which is possible in C++ of course (but you have to do some locking etc to be thread safe).

A regular timer sounds like a waste... but it doesnt eat much time and for my chat I already needed regular checks, for things like: is client timeout, is client flooding, client needs a welcome etc. ATM I do this checks every second.

So, when the server becomes a message "Rickard greets all" for sending to all clients, it goes through the list and sends it to every client that is not dead (and properly logged in). All dead sockets become removed in regular intervalls and won't polute the socket list for long. You might want to maintain more then one internal list, e.g. one for 'incoming and not logged in crap sockets' and one for 'logged in chat client sockets'. Smile | :)

Hope it helps... hej då, Moak

Rickard Andersson²⁰1-Feb-03 4:29

1-Feb-03 4:29

Moak wrote:
So, when the server becomes a message "Rickard greets all" for sending to all clients, it goes through the list and sends it to every client that is not dead (and properly logged in). All dead sockets become removed in regular intervalls and won't polute the socket list for long.

Okay, let's say you do start sending to all clients in the list and then suddenly a socket gets closed , the regular interval will delete it after sending because it suddenly became dead (if you know what I mean, perhaps not), that means I will do a send() call to a dead socket. Is that bad? I mean it shouldn't be... doesn't it just return an error code or something if I send to a dead socket?

Rickard Andersson@Suza Computing
C# and C++ programmer from SWEDEN!
UIN: 50302279
E-Mail: nikado@pc.nu
Speciality: I love C#, ASP.NET and C++!

Moak1-Feb-03 4:49

1-Feb-03 4:49

Rickard Andersson wrote:
start sending to all clients in the list and then suddenly a socket gets closed

Well... that is strictly NOT allowed to happen (but it will of course). What you do to prevent such cases does hardly depend on your socket concept.

If you run a nonblocking multi threaded server concept you need to hazzle with locking on your own. E.g. wait in client thread until server thread signs "okay, socket list is available for write attempts" (beware of deadlocks). I mean, two threads writing simultanously on a STL container is a 'no no'. Just sending to an unconnected/dead socket will cause IMHO an uncritical error return value (which you can ignore).

If you have a nonblocking asychronous concept within one thread context you're lucky as it is. Because all client/server events are allready "synchronized" and come one after another. I do it this way in my chat. Regarding your question: in this socket concept one socket event says "receiving data" and you decide to send your stuff to all clients.... the next socket event says "socket disconnected" and you mark socket as dead for later remove. That's it here.

Hope it helps a bit, Moak Smile | :)

Rickard Andersson²⁰1-Feb-03 7:30

1-Feb-03 7:30

Moak wrote:
If you run a nonblocking multi threaded server concept you need to hazzle with locking on your own. E.g. wait in client thread until server thread signs "okay, socket list is available for write attempts" (beware of deadlocks). I mean, two threads writing simultanously on a STL container is a 'no no'. Just sending to an unconnected/dead socket will cause IMHO an uncritical error return value (which you can ignore).

My first design was almost like this except from my design was with blocking sockets.
But now I trying to use that one you mention below...

Moak wrote:
If you have a nonblocking asychronous concept within one thread context you're lucky as it is. Because all client/server events are allready "synchronized" and come one after another.

Yes, I use that concept for now. It's much more powerful and STABLE than the blocking multi threaded concept! Dead | X|

Moak wrote:
the next socket event says "socket disconnected" and you mark socket as dead for later remove.

Aaah! My regular interval removes it from the list, I just mark it dead!! Smile | :)

woohoo!!

Thank you so so much!
You've teached me more than anybody has! Smile | :)

I mean it, you're really know this topic!

Rickard Andersson@Suza Computing
C# and C++ programmer from SWEDEN!
UIN: 50302279
E-Mail: nikado@pc.nu
Speciality: I love C#, ASP.NET and C++!

Moak1-Feb-03 7:59

1-Feb-03 7:59

Rickard Andersson wrote:
You've teached me more than anybody has! I mean it, you're really know this topic!

Ah cool, it works and I could help!! prolly I know some stuff b/c I program a chat as well and so do run into same questions. Just ask if there is more...

Greets, Moak

PS: Oh and please don't make your chat better than mine! *kidding* Laugh | :laugh:

Rickard Andersson²⁰1-Feb-03 10:58

1-Feb-03 10:58

Moak wrote:
PS: Oh and please don't make your chat better than mine!

Big Grin | :-D

Mowahaha!
I will make a better chat than anyone! A better than ICQ and MSN!!!! Better then ALL chat systems in the whole world!!!

Rickard, stop dreaming.... Smile | :)

Rickard Andersson@Suza Computing
C# and C++ programmer from SWEDEN!
UIN: 50302279
E-Mail: nikado@pc.nu
Speciality: I love C#, ASP.NET and C++!

Rickard Andersson²⁰1-Feb-03 7:53

1-Feb-03 7:53

Hey Moak, can you explain how FD_WRITE works!?

I do like this:

case WM_SOCKET:

            // Determine whether an error occurred on the
            // socket by using the WSAGETSELECTERROR() macro

            if (WSAGETSELECTERROR(lParam))
            {
                // Display the error and close the socket
                closesocket(wParam);
                break;
            }

            // Determine what event occurred on the
            // socket

            switch(WSAGETSELECTEVENT(lParam))
            {
                case FD_ACCEPT:

                    // Accept an incoming connection
                    Accept = accept(wParam, NULL, NULL);

                    // Prepare accepted socket for read,
                    // write, and close notification

                    WSAAsyncSelect(Accept, hWnd, WM_SOCKET,
                        FD_READ | FD_WRITE | FD_CLOSE);
                    break;

                case FD_READ:

                    MessageBox(NULL, "FD_READ", "FD_READ",MB_OK);
                    break;

                case FD_WRITE:
		MessageBox(NULL, "FD_WRITE", "FD_WRITE",MB_OK);
                    break;

                case FD_CLOSE:

                    MessageBox(NULL, "FD_CLOSE", "CLOSE",MB_OK);
                    closesocket(wParam);
                    break;
            }
            break;

What I have problems with is that I don't really understand the use of FD_WRITE.
When does my WndProc get this message???

Rickard Andersson@Suza Computing
C# and C++ programmer from SWEDEN!
UIN: 50302279
E-Mail: nikado@pc.nu
Speciality: I love C#, ASP.NET and C++!

Moak1-Feb-03 8:03

1-Feb-03 8:03

Rickard Andersson wrote:
What I have problems with is that I don't really understand the use of FD_WRITE.

Hm... when you are allowed to send more data (ATM you don't need to handle with it since you do not send large amount of data).

Doesn't answer your question, right? Well, it's a FAQ so I'm lazy and suggest to read some winsock references. See link in last posts or e.g. here http://www.hal-pc.org/~johnnie2/winsock.html[^]

[EDIT]PS: Have a look into Sockcore.cpp if you want to stick with Winsock's asynchronous mode[/EDIT]

Rickard Andersson²⁰1-Feb-03 10:53