🔒 Locked

WTF??!! NULL is not defined in MSVC60??!!

Started by exa_einstein Aug 4, 2004 at 8:15 AM 106 replies 21.9k views

Original Post

Aug 4, 2004 at 8:15 AM

I tried to compile a system of some nested classes and what happenned.. (small code dump:)



// in header file of the class

class gravityobjectsystem:public spaceobject
{
	spaceobject* soa;
	int nso;
public:
	gravityobjectsystem();
	virtual ~gravityobjectsystem();

///...........
// and then in .cpp file ...

gravityobjectsystem::gravityobjectsystem()  //constructor
{
	nso=0;
	soa=NULL;
}
//....

I thought that this sholdn't be a problem, but MSVC6.0 came with two errors: error C2065: 'NULL' : undeclared identifier //??????!!!!!! error C2440: '=' : cannot convert from 'int' to 'class spaceobject *' does somebody had any similar problem? I though NULL is valid zero pointer really EVERYWHERE? ISN'T??!! WTF?!!! helpme... (i don't want to #define my own NULL) exa_einstein

Namethatnobodyelsetook

Aug 4, 2004 at 8:19 AM

Just use 0 (zero). Seriously

exa_einstein

Aug 4, 2004 at 8:27 AM

I know this, but question here is a little more about that "Has the NULL disappeared?" bug in visualC? (I have really checked everything)

ok,ok. now I use soa=(spaceobject*)0; because soa doesn't look like normal integer with it...

Agony

Aug 4, 2004 at 8:45 AM

NULL is #defined or whatnot in various library headers. If you don't include any of these in your project, then NULL is just an undeclared identifier. If you really want you can always add

#ifndef NULL#define NULL 0#endif

to your code, although switching to simply using 0 might be just as good. Only a few rare systems using something other than 0 for NULL. I'd switch, but I'm too lazy at the moment. Actually, if it's a big enough project, I'd probably just make my own const variable that matched the naming convention of the rest of my constants, and was in the proper namespace, too, if I was using namespaces.

Enigma

Aug 4, 2004 at 8:55 AM

class Null{	public:		template <typename TYPE>		operator TYPE*() const		{			return 0;		}};const Null null;GravityObjectSystem::GravityObjectSystem(){	aDescriptiveName = 0;	anotherDescriptiveName = null;}

Simple, clean, unambiguous.

Enigma

Qw3r7yU10p!

Aug 4, 2004 at 8:58 AM

Quote:
Original post by Enigma
*** Source Snippet Removed ***
Simple, clean, unambiguous.

Enigma

nice

evolutional

Aug 4, 2004 at 9:00 AM

Actually, you'll have to define your own NULL - because it's normal definition place is in the windows.h headers (or one hanging off that).

Enigma: Sweet [wink]

merlin9x9

Aug 4, 2004 at 9:06 AM

If I'm not mistaken, stddef.h always defines it. Of course, as Namethatnobodyelsetook mentioned, 0 is the correct C++ equivalent.

KulSeran

Aug 4, 2004 at 9:41 AM

Then why is there the line:

#define NULL ((void *)0)

in windows.h?

vovansim

Aug 4, 2004 at 11:43 AM

Hi,

Quote:
Original post by exa_einstein
I thought that this sholdn't be a problem, but MSVC6.0 came with two errors:

error C2065: 'NULL' : undeclared identifier //??????!!!!!!
error C2440: '=' : cannot convert from 'int' to 'class spaceobject *'

does somebody had any similar problem? I though NULL is valid zero pointer really EVERYWHERE? ISN'T??!! WTF?!!!

You thought wrong. NULL is *not* defined everywhere. One of these days, try to compile the following code:

int main() {  void* ptr = NULL;  return 0;}

I just did, and GCC gave me the following:

test.c: In function `main':test.c:2: `NULL' undeclared (first use in this function)test.c:2: (Each undeclared identifier is reported only oncetest.c:2: for each function it appears in.)

That is because NULL is not a symbol defined by default, but is contained in stdlib.h. (And also happens to be defined in windows.h, and some other headers.)

Vovan

null_pointer

Aug 4, 2004 at 11:43 AM

Quote:
Simple, clean, unambiguous.

Correction: overcomplicated, clean, and potentially ambiguous.

1. Overcomplicated. In C++ you use 0. If you feel an irrational need to use the word NULL instead of the numeric constant zero, you just # define NULL 0 or include one of the standard headers that is supposed to do that for you. Defining a template class to emulate the numeric constant zero is ridiculous.

2. Clean. For what its worth, you did a good job formatting it, so I'll give it a 8.5. Technically, you should have used T for TYPE since it's shorter and just as obvious since it is well known.

3. Potentially ambiguous. You unnecessarily introduced several additional names into whatever namespace you would define this class and its instance, and when other people attempt to define their own versions of "null", you have created the potential for naming conflicts. No one will get such conflicts using 0 or the standard NULL macro.

However, it is interesting since I never would have thought of it and nostalgic since it brings to mind an old topic.

evolutional

Aug 4, 2004 at 11:45 AM

Quote:
Original post by KulSeran
Then why is there the line:

#define NULL ((void *)0)

in windows.h?

It's in windef.h which is included by windows.h

vovansim

Aug 4, 2004 at 11:50 AM

Hi,

Quote:
Original post by null_pointer
2. Clean. For what its worth, you did a good job formatting it, so I'll give it a 8.5. Technically, you should have used T for TYPE since it's shorter and just as obvious since it is well known.

LOL, you didn't take points off for the fact that he used hard tabs for indentation instead of spaces? :P

Seriously, I think the C-style NULL pre-proc symbol works just as well in C++. Don't really see a reason to wrap it in an object of its own. :)

Vovan

SiCrane

Aug 4, 2004 at 11:52 AM

In C++, the standard requires NULL to be defined if you include the clocale, cstddef, cstdio, cstdlib, cstring, ctime and cwchar headers. If you don't include any of these headers, you have no guarantee that NULL will be defined. (Of course other headers may include these headers themselves, so you may get NULL defined even if you don't include them directly, but you can't rely on such implementation specific behaviour.)

Dredge-Master

Aug 4, 2004 at 12:48 PM

a fun thing to do is after sticking all of your includes in, add the following

#ifdef NULL#undef NULL#endif#define NULL 1

vovansim

Aug 4, 2004 at 1:19 PM

Hi,

Quote:
Original post by Anonymous Poster
I never understood the hatred of hard tabs. Whats wrong with everyone being able to configure their editors to display the indentation however they feel comfortable with? Not to mention it saves just a tiny bit of space.

Well, the space argument is silly. If we want to save space, why don't we just type all the code on one line. Dropping newlines would also save a bit of space. ;)

As far as configuring your editor, well, not all editors can be configured. :) What with people who use notepad, or it's glorified brother - vi?

I guess my own "fear" of tabs comes from my experience in college classes. Every Java project there had to go cleanly through checkstyle, and that program complains about tabs. :)

But really, I was just trying to be funny, hence the :P. ;) I thought it was a little odd that null_pointer commented on the use of TYPE instead of T, and neglected to mention hard tabs instead of spaces. :)

Vovan

null_pointer

Aug 4, 2004 at 1:57 PM

Quote:
I thought it was a little odd that null_pointer commented on the use of TYPE instead of T, and neglected to mention hard tabs instead of spaces. :)

Assigning zero to a pointer is a C++ convention used to express a null pointer, but some of us think that we have to spell out the word N-U-L-L in all of our null pointer expressions for some reason. That's what I was objecting to; not the use of null pointers. :)

As for the tabs, well, the approach is clearly superior to spaces. Some people prefer 2, 3, or 4 spaces. I think that making the source code less flexible just to cope with silly editor deficiencies would be a very bad long term solution.

(And while we're ranting about "infinitely trivial" things, what the heck happened to all the smiley graphics in the posts? Is that only for paying members now? Or mayhap mine actions hath offended the mods?)

Sneftel

Aug 4, 2004 at 2:01 PM

The smiley codes were replaced with codes you're less likely to type when you don't mean 'em; see the FAQ for details. We were all fed up with things like std::pair showing up as std:[razz]air.

null_pointer

Aug 4, 2004 at 2:11 PM

The forum FAQs are actually being used? [cool]

BTW the [attention] icon appears to be corrupted in Firefox 9.2. It's got a wierd whitish border around it that makes it hard to figure out that it's a yellow exclamation mark. Um...that is what it is, right? [looksaround]

Xai

Aug 4, 2004 at 2:24 PM

is there anyone on this earth who still thinks that "hard tabs" are a formatting mistake and "spaces" are good white space usage??? (I know there are, but I'm here to say once and for all that a proper formatting convention never ever needs space based alignment and works properly no matter how large a tab is on the viewing system - hence the reason tabs are preferable).

Here's the One Rule:

Every source line is lined up to either -1, 0, or +1 tabs of the line above it.

Here's the One Exception:

A line may be aligned more than one tab less than the line above it if and only if the preceeding line was logically nested multiple levels and did not use braces surrounding the blocks

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.