undefined reference to std list during linking

Hi,

I have a header file:

Code:

---


#ifndef _TX_SERVER_H
#define _TX_SERVER_H

#include<tx-types.h>
#include<nssi_xdr.h>
#include<list>
#include "nssi_server.h"

#define NUMCLIENTS 1

extern tx_info current_push_tx;
extern nssi_service xfer_svc;
extern nssi_remote_pid my_procid;

extern std::list<client_response*> push_respond_list;
extern std::list<client_response*> pull_respond_list;
extern std::list<tx_info*> tx_history_list;

void init_server();
void push_transaction(
                      const unsigned long request_id,
                      const nssi_remote_pid *caller,
                      const push_tx_data *tx,
                      const nssi_rma *data_addr,
                      const nssi_rma *res_addr
                      );
void process_transaction(
                        tx_info * info, tx_status STATUS
                        );
#endif

---

I am able to create the .o files with no problem. but when I can't link... for example, I get the following error...

tx-server.cpp:(.text+0x1f): undefined reference to `tx_history_list'

I get this for each one of the lists above.

I have a file tx-server.cpp which includes tx-server.h

Here are the calls I make for tx_history_list

Code:

---

for(list<tx_info*>::iterator it = tx_history_list.begin(); it != tx_history_list.end(); it++)

---

and in another function

Code:

---


tx_history_list.push_front(info);

---

in tx-server.cpp, I do have using namespace std

Is there something I'm missing during the linking? Something specific to which I have to link?

Currently, I am only linking to in-house libraries...

Code:

---

g++ -Wall -L $HOME/lib -L /ascldap/users/xxxx/lib/ -lcommandlineprocessor -lnx_server -lnx_client -lnx_support -lezxml -o tx-test tx-server.o tx-test.o

---

I'm extremely new to C/C++, so this is a bit confusing for me.

Thanks for the patience!!

Do you actually make a tx_history_list? If so, where?

In the header file I just did this:

extern std::list<tx_info*> tx_history_list;


I'm guessing that means I just declared it, but did not instantiate it?

I kind of followed this: http://www.yolinux.com/TUTORIALS/Lin...+STL.html#LIST

Code:

---

list<int> L;
 L.push_back(0);              // Insert a new element at the end
L.push_front(0);            // Insert a new element at the beginning
 L.insert(++L.begin(),2);    // Insert "2" before position of first argu

---

I'm guessing since I declared the list in the header file, I have to instantiate it differently? Not sure of the syntax to do that...

Basically yes. That line means "I claim that such an object exists somewhere, but It's Not My Job to make one." Which is all well and good, and it's what you have to have in a header file, since header files go everywhere. But at the end of the day, somebody has to make that object.

I see. I was following that example. I just see them doing list<int> L, then immediately start calling functions on L. I guess, what's the syntax then to instantiate a list that I declared in the header file...

Just what you said.

Code:

---

std::list<tx_info*> tx_history_list;  //hey look it's a list

---

But under no circumstances can you put this into a header file. It must be in a code file.
It may sound crazy, but you might want to look into a book so that you know what all the words mean (like "extern").

if I remove extern, I get this:

tx-server.cpp:21: error: redefinition of ‘std::list<tx_info*, std::allocator<tx_info*> > tx_history_list’

So for the love of all that is good, stop being an idiot. Which part of "you can't put this in a header file" is hard to understand?

considering I said you declare it in a headerfile and instantiate it in the .cpp file, then proceeded to say "the code is the same", which was not corrected...

With Java you can do things like that.. declare them in some parts, and instantiate them in others. That's what I guess is confusing here.

Quote:

---

Originally Posted by dayalsoap

considering I said you declare it in a headerfile and instantiate it in the .cpp file, then proceeded to say "the code is the same", which was not corrected...

---

Which post are you referring to? None of the posts that I can see in this thread have those words in them, or anything synonymous that I can find.

When you see examples on the web that do different things in the header files and the code files, and then it is pointed out here that you do different things in the header files and the code files, and you yourself even claim to know the difference between declaring and instantiating, and then you don't do different things in the header files and the code files, then what exactly do you expect?

me: I'm guessing since I declared the list in the header file, I have to instantiate it differently? Not sure of the syntax to do that...

you: Basically yes.

So you've said yes that I can declare it in the header file.

me: I guess, what's the syntax then to instantiate a list that I declared in the header file...

You: Just what you said.

me: I'm guessing since I declared the list in the header file, I have to instantiate it differently? Not sure of the syntax to do that...

you: Basically yes.

So you've said yes that I can declare it in the header file.

me: I guess, what's the syntax then to instantiate a list that I declared in the header file...

You: Just what you said.

So I agreed that you had to instantiate it differently. In no way did I say you can insta it in the header file. If you know anything about computers, then you yourself know (with or without me telling you) that you can't instantiate it in the header file.

And then I agreed that the syntax to instantiate the list was exactly as you had posted in your post #3 and #5 -- which as you know was never in your program, for reasons known only to you -- and reemphasized the fact that you can't instantiate it in the header file.

"So I agreed that you had to instantiate it differently"

Then why did you say yes to "I declare it in the header file, I have to instantiate it differently"

When you say Yes, you are agreeing with two points. And I later discuss "declare in the header file" and you say "just what you said".

No where, until after that, did you say "you can't declare objects in a header file like this"

Which is also confusing, because this person declares his list in his headerfile

Declaring a list in a header file - C++

"If you know anything about computers, then you yourself know (with or without me telling you) that you can't declare it in the header file."

C++ != computers.

I have no idea how you're reading what you're reading out of what I wrote, but somehow we're not communicating, so we'll let that go.

The link you posted does not have a person declaring a list in a header file, so that's remarkably irrelevant. (It does have a person creating a class that contains a list inside it -- and look! that person is only declaring a class and not instantiating it! even better/worse, depending on your perspective.)

Quote:

---

Originally Posted by dayalsoap

"So I agreed that you had to instantiate it differently"

Then why did you say yes to "I declare it in the header file, I have to instantiate it differently"

When you say Yes, you are agreeing with two points. And I later discuss "declare in the header file" and you say "just what you said".

No where, until after that, did you say "you can't declare objects in a header file like this"

Which is also confusing, because this person declares his list in his headerfile

Declaring a list in a header file - C++

"If you know anything about computers, then you yourself know (with or without me telling you) that you can't declare it in the header file."

C++ != computers.

---

What's confusing to me then is why the person has this in his header file

Code:

---

private:

List<int> intergerList1;

---

To me, that looks almost identical to what I was doing inside my headerfile, without the private:

And yet, it isn't. It isn't, because it inside the declaration of a class. Do you know what a class is in C++? If not, please go read what it is. Here might be a good start. A class declaration in C++ is a new declaration of a type like, well, isn't it a class declaration in Java? (I've been trying to burn Java out of my brain, and I've pretty well succeeded.) You are saying "I am defining a new type of object which will contains these objects and these methods" -- but you are not actually making any objects, you are just saying how they will be made when they are needed.

Quote:

---

Originally Posted by tabstop

And yet, it isn't. It isn't, because it inside the declaration of a class. Do you know what a class is in C++? If not, please go read what it is. Here might be a good start. A class declaration in C++ is a new declaration of a type like, well, isn't it a class declaration in Java? (I've been trying to burn Java out of my brain, and I've pretty well succeeded.) You are saying "I am defining a new type of object which will contains these objects and these methods" -- but you are not actually making any objects, you are just saying how they will be made when they are needed.

---

It's not in a header file? Pretty sure he posted it in a .h file.

Your earlier statement said you can't declare things inside header files. You did not add any qualifiers to say "except when it's inside of a class". Unless .h files are not exclusively header files?

Perhaps it's worth understanding the reasons for the rules tabstop mentions.

Your program must allocate space for a list exactly once. To do this requires that they be instantiated in exactly one object file (.o), which is compiled from each source file(.c). However, every identifier that is used must be previously declared in the post-processes source.

With a variable such as a list, this means that you must declare it using exten in a header file, and instantiate it in exactly one source file. With a class, no space is allocated, so there declaration goes in the header, and it might contain lists that use similar syntax to a list definition, but no list with space is actually instantiated.

I think I'm starting to understand. So, without using classes, there isn't really a way to declare it in the header file, but instantiate it in the .cpp file?

Quote:

---

Originally Posted by King Mir

Perhaps it's worth understanding the reasons for the rules tabstop mentions.

Your program must allocate space for a list exactly once. To do this requires that they be instantiated in exactly one object file (.o), which is compiled from each source file(.c). However, every identifier that is used must be previously declared in the post-processes source.

With a variable such as a list, this means that you must declare it using exten in a header file, and instantiate it in exactly one source file. With a class, no space is allocated, so there declaration goes in the header, and it might contain lists that use similar syntax to a list definition, but no list with space is actually instantiated.

---

Quote:

---

Originally Posted by dayalsoap

So, without using classes, there isn't really a way to declare it in the header file, but instantiate it in the .cpp file?

---

Err... there is: use extern to declare - but not define - the file scope variable, as in your initial example.

The point is to avoid breaking the one definition rule: if you only ever include your header file in a single source file, then you can away with defining a variable at file scope in a header. But header files are typically included in more than one source file. Or, you could define the file scope variable in an anonymous namespace, or as static, so that it would be specific to each translation unit (source file + headers) and hence not break the one definition rule when the header is included in multiple source files (but this is clearly not feasible if you want to access the same global variable from multiple translation units).

Quote:

---

Originally Posted by laserlight

Err... there is: use extern to declare - but not define - the file scope variable, as in your initial example.

The point is to avoid breaking the one definition rule: if you only ever include your header file in a single source file, then you can away with defining a variable at file scope in a header. But header files are typically included in more than one source file. Or, you could define the file scope variable in an anonymous namespace, or as static, so that it would be specific to each translation unit (source file + headers) and hence not break the one definition rule when the header is included in multiple source files (but this is clearly not feasible if you want to access the same global variable from multiple translation units).

---

I think i finally get it now, especially in regards to including the header in multiple files. Thank you so much for your patients. I'm a math person, so I'm not really familiar with this coding stuff. I'm doing this at an Internship at a national lab, and they have a lot of math and analysis libraries.