Compiler Help Needed

[Darkman_X000]

Ok guys, what is the deal!? I am getting the following errors when I compile:

OUTPUT
MPLINK 3.40, Linker
Copyright (c) 2003 Microchip Technology Inc.
Error - symbol 'axelWidth' has multiple definitions.
Errors : 1

BUILD FAILED: Sat Feb 07 11:04:01 2004

This is really annoying because it does not tell me where the other definitions are. In C++, and in this compiler, if I define the same thing twice explicitly, it tells me where the previous definition is. Here is a sample of my code:

UserUtilities.h
#ifndef USERUTILITIES_H
#define USERUTILITIES_H

#include "bool.h"

float axelWidth=.725;
float wheelRadius=.105;
int clicksPerRevolution=16;

bool newValue(int *staticVar, int dynamicVar);
float thetaDegrees(int c);
float thetaRadians(int c);
float Forward(int c);

#endif

UserUtilities.c
#ifndef USERUTILITIES_C
#define USERUTILITIES_C

#include "bool.h"
#include "UserUtilities.h"

bool newValue(int *staticVar, int dynamicVar)
{

if (*staticVar==dynamicVar)
{
*staticVar=dynamicVar;
return true;
}
else
{
return false;
}
}

float thetaDegrees(int clicks)
{
extern float axelWidth;
extern float wheelRadius;
clicks*=22.5;
clicks=clicks*wheelRadius;
clicks/=(float)axelWidth;
return clicks; //22.5*(float)wheelRadius*clicks/(float)axelWidth;
}

float thetaRadians(int clicks)
{
extern float axelWidth;
extern float wheelRadius;
clicks*=3.1415926;
clicks=clicks*wheelRadius;
clicks/=8*axelWidth;
return clicks;
}

#endif


Note: these are the only files that contain the definitions of these functions and varibles. There are NO duplicate names or any of that.
WHAT IS GOING ON!?!?!?!?!?!?!?!?!?
Clearly, including the file multiple times is prevented by the preprocessor #ifndef, #define, #endif statements, but it is acting like it just disregards those...

[Mike Betts]

Slow down and look at what you are doing.

You defined axelWidth in userutilities.h and then included userutilities.h in userutilities.c... This means that axelWidth is defined as a float.

Now you redefine axelWidth as an external variable in userutilities.c meaning that some OTHER module has defined axelWidth. This is a double declaration and is only the tip of the iceberg... I see a bunch. i.e. you include bool.h multiple times...

Also, take a few minutes and see various threads about float variables and implicit initializations... You are going down a very dangerous road.

Regards,

[Mike Soukup]

Note: these are the only files that contain the definitions of these functions and varibles. There are NO duplicate names or any of that.
WHAT IS GOING ON!?!?!?!?!?!?!?!?!?
Clearly, including the file multiple times is prevented by the preprocessor #ifndef, #define, #endif statements, but it is acting like it just disregards those...
The problem is actually with the line "extern float axleWidth". Since you're including the header that has the declaration of "axleWidth" there is no need to re-declare it inside your function. 'extern' is only used when you have a global variable defined in another file.

From the names of your variables & the way you use them, it looks like those are values that will never change while your code is being executed, ie a constant. If so, there is no need to use a variable to store the values, use a #define instead:

#define axelWidth .725
#define wheelRadius .105
#define clicksPerRevolution 16

Post more questions if you have them.

[Darkman_X000]

Ok, I just replaced what I had before with:
#define axelWidth .725
#define wheelRadius .105
#define clicksPerRevolution 16
and I have deleted the externs. Now the compiler has no trouble with "axelWidth". Instead it says:

MPLINK 3.40, Linker
Copyright (c) 2003 Microchip Technology Inc.
Error - symbol 'newValue' has multiple definitions.
Errors : 1

BUILD FAILED: Sat Feb 07 12:05:03 2004

I wrote the function prototype in UserUtilities.h and the actual function definition in UserUtilities.c. There should be no conflicts here because prototypes ARE allowed.

Each time I fix a problem with one function or variable name in the set of files, it has a problem with the next one down the line.

Please consider that because bool.h has the necessary preprocessor statements, including it multiple times should not be a problem... right? Isn't that the primary use of #ifndef, #define, and #endif structures in header files?

I also tried to delete the #include "userutilities.h" line from UserUtilities.c after making the above changes, but then it said that axelWidth and wheelRadius were not defined. Can I use extern with a variable that was #defined?

Here is the updated version of my files:

UserUtilities.h
#ifndef USERUTILITIES_H
#define USERUTILITIES_H

#include "bool.h"

#define axelWidth .725
#define wheelRadius .105
#define clicksPerRevolution 16

/*float axelWidth = .725;
float wheelRadius = .105;
int clicksPerRevolution = 16;*/

bool newValue(int *staticVar, int dynamicVar);
float thetaDegrees(int c);
float thetaRadians(int c);
float Forward(int c);
void move(float distance);
void turnLeft(float degrees);
void turnRight(float degrees);

#endif

UserUtilities.c
#ifndef USERUTILITIES_C
#define USERUTILITIES_C

#include "bool.h"
#include "UserUtilities.h"

bool newValue(int *staticVar, int dynamicVar)
{

if (*staticVar==dynamicVar)
{
*staticVar=dynamicVar;
return true;
}
else
{
return false;
}
}

float thetaDegrees(int clicks)
{
clicks*=22.5;
clicks=clicks*wheelRadius;
clicks/=(float)axelWidth;
return clicks; //22.5*(float)wheelRadius*clicks/(float)axelWidth;
}

float thetaRadians(int clicks)
{
clicks*=3.1415926;
clicks=clicks*wheelRadius;
clicks/=8*axelWidth;
return clicks;
}

#endif

[Darkman_X000]

Take a few minutes and see various threads about float variables and implicit initializations... You are going down a very dangerous road.

So, after the prescribed research, I still do not understand the "dangerous road" of floats and implicit initializations. Please enlighten me.

Thanx in advance.

[Random Dude]

So, after the prescribed research, I still do not understand the "dangerous road" of floats and implicit initializations. Please enlighten me.

Thanx in advance.


Well, the issue with floats is that this micro does not have a Floating Point Unit. So, while on your PC the FPU would quicky preform any math involving floats, that all has to be emulated in software on this micro. That is bad becuase that emulation is fairly complex and could potentially slow down your program considerably, especially if you use alot of floating point numbers.

The concerns about the implicit initializations, is basically that normally on microprocessors, the line

int i = 5;

may not automatically initalize i = 5. Though that is really a moot point in this case since IFI has a routine that does initalize the variables for us. (in ifi_startup.c for those who haven't seen it.)


-----------------------
No, you can't use extern with #define. #define isn't actually a variable, what it means, is before the code is complied (in the pre-processor), the any time the first item after the #define is found, it is replaced with the second. So:

#define PI 3.1415926538
int C, R = 3;
C= 2*PI*R;


Becomes (after the pre-processor, but before the actuall compiler)

int C, R = 3;
C= 2*3.1415926538*R;



So if you just put the

#define axelWidth .725

in one header file, that you include everywhere, you can use axelWidth with no other effort.


------------------------------

[Jeff McCune]

As an example, when using a microcontroller, using this:

#define PI (22 / 7)

is *considerably* faster than using 3.14159... and it gives a fairly decent approximation.

Also, to prevent re-inclusion of header files, people usually do something like this:

myheader.h
#ifndef MYHEADER_H
#define MYHEADER_H

.. Entire header file ...

#endif
/* ifndef MYHEADER_H */

This prevents weird recursion problems from having a header file included twice at the top of a C file.

It's sloppy, and I don't recommend doing this, because if you *need* to do this, you're already down the dangerous road they've been speaking of and you ultimately need to fix that or you'll just have a *monstrous* headache at competition when you're rapidly changing code.

[Jeff McCune]

Also, you can completley eliminate floating point from your code.

#define axelWidth 0.725

becomes:

#define AXEL_WIDTH_NUM 29
#define AXEL_WIDTH_DEN 40

Constants that are #defined are usually all caps by convention.

Also, keep in mind that when using integers, order of operations isn't the same as it is in your math class. Well, it is, but the order of operations will affect the result of the equation.
// Consider:
#define NUM 50
#define DEN 100
int foo = NUM / DEN * joy1_y
int bar = NUM * joy1_y / DEN

foo will ALWAYS be 0. Always. this is becuase NUM divided by DEN in integer math will be 0, and 0 multiplied by anything is zero. bar will contain what you think it should based on the equation.

Basically, stick to integer math, and multiply to the biggest number you can before starting to divide using integers to prevent always multoplying by zero.

[josh_johnson]

Basically, stick to integer math, and multiply to the biggest number you can before starting to divide using integers to prevent always multoplying by zero.

This can be a dangerous aproach, as often the largest value is too large for the variable type that you are putting it in, but the final result would fit just fine. Just make sure that any intermediate values are not going to be too large before doing this.

[Astronouth7303]

... From the names of your variables & the way you use them, it looks like those are values that will never change while your code is being executed, ie a constant. If so, there is no need to use a variable to store the values, use a #define instead:

#define axelWidth .725
#define wheelRadius .105
#define clicksPerRevolution 16
...

I, personally, have had problems with #declare in place of const, and only use it for alias, Example:

//Some constants
const char Varience = 10;
const char True = 255;
const char bLAH = 152;

//Some aliases
#declare Left_OI p1_y
#declare Left_Motor pwm01
#declare DyNoMiTe relay8


As long as you keep it in headers and include them, you should be ok.

[deltacoder1020]

first of all, it's "#define", not "#declare"

second, the reason you are having problems is because mcc18 automatically casts any number from 0 to 255 as an unsigned char, which can screw up math. instead, what you'd want is this:

#define Varience ((char) 10)

[Random Dude]

As an example, when using a microcontroller, using this:

#define PI (22 / 7)

is *considerably* faster than using 3.14159... and it gives a fairly decent approximation.


Umm.. But (22/7) evaluates to (3).

[Astronouth7303]

Only in integer math (pi rounded down is three).

Ps- My bad on the #declare!

[deltacoder1020]

Umm.. But (22/7) evaluates to (3).

technically, 3 is a "decent approximation" for pi... but heck, if you're gonna do that, just do what some crazy state legislature tried to do, and

#define PI 3

;)

[Random Dude]

Only in integer math (pi rounded down is three).

Ps- My bad on the #declare!


Yes, exactly (22/7) is integer math (22.0/7) would be floating point. [Or ((float)22/7)]

[KevinB]

... just do what some crazy state legislature tried to do ...
Although Alabama's state legislature is pretty darn stupid somtimes ... the story that floated around the internet about us trying to do that .... is fake. :)

[deltacoder1020]

Although Alabama's state legislature is pretty darn stupid somtimes ... the story that floated around the internet about us trying to do that .... is fake. :)

i'm not talking about alabama.

[Jeff McCune]

Umm.. But (22/7) evaluates to (3).
My #define shouldn't include ( )'s. Then, if you use PI in an equation, it'll work just fine even though it's integer math.

#define PI 22 / 7
#define DIAMETER 10

circ = DIAMATER * PI; /* circ = 10 * 22 / 7 */

[Darkman_X000]

Ok, thanks for the info guys, but my problem still isn't fixed. I still get the error about having multiple definitions, even though I don't. I have included the preprocessor directives #ifndef, #define, #endif to prevent such nonsense from bothering me, but to no avail. I have written several functions that I NEED to make my robot work right, and it is extremely frustrating to have the compiler return errors that are unfounded.
:mad:
Please help!

My code is presented on the previous page, and I don't want to re-post it because it is long... Please check it out and tell me why the function "newValue" keeps coming up as having "multiple definitions"!

:confused:

Thanx in advance!!!!!!

[deltacoder1020]

notice that the error you are getting is a link error - this means that the multiple definitions are in different files - are you sure you didn't accidentally name something (even a variable) newValue somewhere else?

[Mark McLeod]

Ok, thanks for the info guys, but my problem still isn't fixed. I still get the error about having multiple definitions, even though I don't. I have included the preprocessor directives #ifndef, #define, #endif to prevent such nonsense from bothering me, but to no avail. I have written several functions that I NEED to make my robot work right, and it is extremely frustrating to have the compiler return errors that are unfounded.
:mad:
Please help!

My code is presented on the previous page, and I don't want to re-post it because it is long... Please check it out and tell me why the function "newValue" keeps coming up as having "multiple definitions"!

:confused:

Thanx in advance!!!!!!

The code you posted here compiles without a problem.
You need to post or email one of us your whole project as a zip file.
Strip out the project and .o files before you Winzip it if you need to keep the size down.

[Astronouth7303]

I don't know. It looks like the only defs are the prototype and the actual code. I would say try putting the prototype in the same file as the sub.

[Greg Ross]

The code you posted here compiles without a problem.
You need to post or email one of us your whole project as a zip file.
Strip out the project and .o files before you Winzip it if you need to keep the size down.
I second the motion. There must be something going behind the scenes of what you've shown us.

[Darkman_X000]

All right guys, it seems to be working now. I just deleted all #include "userutilities.c" directives and replaced them with #include "userutilities.h" directive where necessary.

So... now I am no longer getting the "multiple definition" errors that were driving me insane. Thanks for your help!!! :D

[Mark McLeod]

All right guys, it seems to be working now. I just deleted all #include "userutilities.c" directives and replaced them with #include "userutilities.h" directive where necessary.

So... now I am no longer getting the "multiple definition" errors that were driving me insane. Thanks for your help!!! :DThat makes sense.

Now that you're no longer being driven insane, how are you getting there?;)

Good Luck!