C / C++ / MFC

24-Oct-07 6:01

Run-Time Type Information[^]

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

Multiple definitions

John R. Shaw23-Oct-07 15:49

23-Oct-07 15:49

The following represents standard C++ code, no Microsoft specifics involved.

// NOTE: The "Visual C++ 6.0" compiler could not destiguish between
// the following 'add' methods when they had the same name ('add_set').
void add_c_set(nfa_type& nfa, const string_set_type& c_set, bool and_flag) const;
void add_id_set(nfa_type& nfa, const ccid_set_type& id_set, bool and_flag) const;

The above problem occured, in long standing code, when I seperated some unrelated methods from a seperate template class, as they could stand on their own. Although I considered it a seperate issue, this must be related, the "Visual C++ 6.0" compiler started complaining about heap space (for symbols) when compiling for Unicode (not for ASCII). When compiling with "Visual C++ .Net" there are no problems.

The "string_set_type" is an STL set of STL strings (char or wchar_t) and "ccid_set_type" is a set of special identifiers. A "ccid" is basically a type derived from a template class that wraps an "unsigned char" value.

If you have any ideas, or have seen this problem before, please throw it out there, as I need a clue.

Thanks.

INTP
"Program testing can be used to show the presence of bugs, but never to show their absence."Edsger Dijkstra

Christian Graus23-Oct-07 16:07

Christian Graus

23-Oct-07 16:07

John R. Shaw wrote:
string_set_type

John R. Shaw wrote:
ccid_set_type

I think this use of macros is dumb and makes code unreadable. Can you provide the actual macros so we can see what we're dealing with ?

Christian Graus - Microsoft MVP - C++

"also I don't think "TranslateOneToTwoBillionOneHundredAndFortySevenMillionFourHundredAndEightyThreeThousandSixHundredAndFortySeven()" is a very good choice for a function name" - SpacixOne ( offering help to someone who really needed it )

John R. Shaw23-Oct-07 17:24

23-Oct-07 17:24

Christian Graus wrote:
Can you provide the actual macros...

Christian,

Why would you think I was using macros? That would be counter productive and bad practice in C++. The code is pure standard C++, and applies nothing but the best practices. Using “typedef” is common practice, as examining any STL container will show.

typedef std::set<string_type> string_set_type;
typedef std::set<ccid_type> ccid_set_type;

The template takes the same arguments as “basic_string” and “string_type is defined as:

typedef std::basic_string<_CharT_, _Traits_, _Alloc_> string_type;

I am using a C++ STL style of coding, because this code is intended to be familiar to anyone use to using the STL.

The code has been working for months, but it needed some minor improvements. The “add_set” methods worked perfectly until yesterday, when I separated those other methods from a separate template class that has nothing to do with this one.

INTP
"Program testing can be used to show the presence of bugs, but never to show their absence."Edsger Dijkstra

User 58385223-Oct-07 16:07

User 583852

23-Oct-07 16:07

Looks like the compiler cant tell the differnce between string_set_type and ccid_set_type.

Is it possible that when this code is compiled the compiler only has a forward declaration of one or both of these?

ie...

class ccid_set_type;
class string_set_type;

class Foo
{
void Add(const ccid_set_type&);
void Add(const string_set_type&);
};

Christian Graus23-Oct-07 16:29

Christian Graus

23-Oct-07 16:29

I don't think they are classes, I think they are #defines for stl containers.

John R. Shaw23-Oct-07 17:50

23-Oct-07 17:50

Josh Gray wrote:
Is it possible that when this code is compiled the compiler only has a forward declaration of one or both of these?

No, everything is declared where it is supposed to be. Until yesterday it compiled perfectly with both methods named “add_set”, as it has for months. As a matter of fact I had not touched this template until this unexplained phenomenon occurred.

It has to have something to do with making those other methods, I mentioned, stand alone, as that is when it started happening. It only happens with VC 6.0, as VC .NET has absolutely no problems with it, even at warning level 4.

Why it can not tell the difference between two types that have totally different signatures is very confusing. It has to be related to the compilers heap space, but it never gets this far if that error occurs.

error C2535: '…' : member function already defined or declared

INTP
"Program testing can be used to show the presence of bugs, but never to show their absence."Edsger Dijkstra

User 58385223-Oct-07 18:22

User 583852

23-Oct-07 18:22

This is a long shot but you could try using the typename keyword

void add_c_set(nfa_type& nfa, const typename string_set_type& c_set, bool and_flag) const;
void add_id_set(nfa_type& nfa, const typename ccid_set_type& id_set, bool and_flag) const;

John R. Shaw23-Oct-07 18:46

SuspendThread and StackWalk64

23-Oct-07 18:46

It was a long shot, but I tried it anyway. It had no effect, as expected. Wink | ;)

I am getting offline now, but tomorrow I am going to reverse the changes I made yesterday and proceed with the changes one step at a time using the VC 6 environment instead of VC .NET, so I can see precisely when the error first appears.

Thanks for trying. Smile | :)

INTP
"Program testing can be used to show the presence of bugs, but never to show their absence."Edsger Dijkstra

Baltoro23-Oct-07 14:08

Baltoro

23-Oct-07 14:08

If you know anything at all about either of these two functions, it's that SuspendThread[^] can be dangerous (primarily because it creates application deadlocks if not properly used). Read, Jochen Kalmbach's WebLog [^]
However, if you want to use the StackWalk64[^]function from the dbghelp.dll, you must call SuspendThread to get relavant and useful results.
So, what is the PROPER way to use SuspendThread? How could you possibly know of all the mechanisms and objects that are currently activated by a thread?

Re: SuspendThread and StackWalk64

cmk24-Oct-07 10:16

cmk

24-Oct-07 10:16

Baltoro wrote:
However, if you want to use the StackWalk64[^]function from the dbghelp.dll, you must call SuspendThread to get relavant and useful results.

Only if you want to walk the stack of a thread other than the current one.

This is fairly obvious, if you didn't suspend the other thread it's stack would likely change while you are walking it.

The only problem in suspending another thread is that, as mentioned, any other threads waiting on the the suspended thread are going to keep waiting. This isn't a problem unless their waits time-out, even this isn't a problem because it tests your codes response to a rare error condition.

...cmk

The idea that I can be presented with a problem, set out to logically solve it with the tools at hand, and wind up with a program that could not be legally used because someone else followed the same logical steps some years ago and filed for a patent on it is horrifying.
- John Carmack

Thanks for Reply

Baltoro25-Oct-07 6:11

Baltoro

25-Oct-07 6:11

...thanks, ...cmk
I was referring to a remote thread, even though it was unstated.
I asked this question over on the Microsoft Technical Forums and got approximately the same answer.
It isn't really a problem, I was just curious.

irc via socket

Adno23-Oct-07 12:29

Adno

23-Oct-07 12:29

using a socket I'm connecting to an irc server, when i do i get:

NOTICE AUTH :*** Processing connection to efnet.demon.co.uk

NOTICE AUTH :*** Looking up your hostname...
NOTICE AUTH :*** Checking Ident
NOTICE AUTH :*** Found your hostname
NOTICE AUTH :*** Got Ident response

this is all i get no welcome message, no matter what i send through this socket i don't get any response.
then after a while i get:
ERROR :Closing Link: 127.0.0.1 (Connection timed out)

thanks

Mark Salsbery23-Oct-07 13:42

23-Oct-07 13:42

What do you mean, you "get" that from the server? How are you receiving that text?

Are you sending the appropriate messages to connect to the server, as per the IRC protocol?

Mark

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

Adno23-Oct-07 14:59

Adno

23-Oct-07 14:59

What do you mean, you "get" that from the server? How are you receiving that text?
yes that is what i get from the server when i connect, via recv (s, buffer,bufferSize, 0);

Are you sending the appropriate messages to connect to the server, as per the IRC protocol?
not sure, according to this all you have to do is send commands like NICK name etc, once you are connected, but it doesn't seem to be working.
http://www.irchelp.org/irchelp/rfc/rfc.html

Mark Salsbery24-Oct-07 6:07

24-Oct-07 6:07

It's impossible to say without seeing any code, but things to look at are:

1) Parsing messages from the server - the protocol doesn't include message length
parameters so it's up to you to recv() all bytes whenever there's an FD_READ event
and parse the bytes into messages. Remember, on one recv(), you may only get a partial message.
You have to deal with that.

2) Make sure you're forming your client-to-server messages correctly. ASCII only - no Unicode.
Use proper whitespace. Check for server replies.

Mark

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

Adno24-Oct-07 9:03

Adno

24-Oct-07 9:03

connect function experimenting with:

int Client::Connect(HWND hwnd)<br />
{<br />
	// Initialize Winsock version 2.2<br />
	char message[200];<br />
	cUser user; // struct user<br />
	sprintf(user.nick,"myNic");<br />
	sprintf(user.ident,"IDspoonMan");<br />
	sprintf(user.email,"vertexar@yahoo.com");<br />
	sprintf(message,"USER %s %s: %s %s  \n\r", user.nick , user.email , user.ident , user.ident );<br />
	<br />
	Port = 6667;<br />
    WSAStartup(MAKEWORD(1,1), &wsaData);<br />
	LPHOSTENT hostEntry;<br />
    //store information about the server<br />
    hostEntry = gethostbyname("efnet.demon.co.uk");<br />
	if(!hostEntry)<br />
	{<br />
		::MessageBox(NULL,"Failed gethostbyname()","Error",0);<br />
		//WSACleanup();<br />
	}<br />
   <br />
   // Create a new socket to make a client connection.<br />
    s = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP);<br />
	ServerAddr.sin_family = AF_INET;<br />
	ServerAddr.sin_port = htons(Port); <br />
<br />
	//ServerAddr.sin_addr.s_addr = inet_addr("127.0.0.1");<br />
	ServerAddr.sin_addr = *((LPIN_ADDR)*hostEntry->h_addr_list);<br />
<br />
  	// Mark our socket non-blocking, and ask for asynch notifications.<br />
	if (WSAAsyncSelect(s, hwnd, customClientMessage,FD_READ | FD_WRITE | FD_CONNECT | FD_CLOSE) == SOCKET_ERROR) <br />
	{<br />
		::MessageBox(NULL,"TSWCCould not set to non-blocking! ", "ERROR..", MB_OK);<br />
		//REPORT_PROBLEM(WSAGetLastErrorMessage("WSAAsyncSelect failed."));<br />
		return false;<br />
	}<br />
	int CON_ERROR = connect(s, (SOCKADDR *) &ServerAddr, sizeof(ServerAddr));<br />
	//send(s, message, strlen(message), 0);<br />
	//sprintf(message,"NICK %s\n\r", user.nick);<br />
	//send(s, message, strlen(message), 0);<br />
	Sleep(10);<br />
	send(s,"NICK mynic \n\r",15,0);<br />
	send(s,"USER mynic 0 * myname \n\r",27,0);<br />
	int errorCode = WSAGetLastError();<br />
	if(!CON_ERROR)<br />
		return true;<br />
	return false;<br />
}

using telnet i tried this and it works:
o efnet.demon.co.uk 6667
NICK mynic
USER mynic 0 * myname

i dont see why is not working when i send it through my socket.

Mark Salsbery24-Oct-07 10:59

24-Oct-07 10:59

You're not using asynchronous sockets correctly.

You should be checking if connect() is successful. If it's not, you need
to check the error code - if it's WSAEWOULDBLOCK, then you need to wait until the
socket is connected before writing data to it. You'll know this when you get the
FD_CONNECT notification, at which time you need to check for errors again, to make sure
the connection succeeded. THEN you can write data to the socket.

"Sleep(10)" is really bad design. It could take hundreds or thousands of milliseconds to
connect.
Maybe you'll get lucky and connect() will succeed instantly (unlikely when connecting to a
WAN address) - you don't check for that though.

Mark

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