Unable to use sin and cos

[Yesideez]

Attached is the full code as I've got it now.

All I want to do is open a screen and window and draw some lines around a circle but having immense difficulty in getting two maths functions to work.

[thomas]

Would probably be easier to #include <math.h> and use ANSI C math functions. You might need to add -lmieee or -lmsoft to the vbcc command line in this case.

The main mistake I see is j <= 10000 and sin(j). The sin function takes the angle in radians. Which means a number between 0 and 2*pi. To get radians from degrees you can use the formular rad = deg * pi / 180. pi is 3.14159265358979323846.

Also you let the compiler convert between int and float and vice versa. But mathtrans.library uses Motorola float format while the compiler most likely uses standard float format. I don't know the difference, but there are functions in mathtrans.library to convert between these two formats, so I assume that the difference matters. As mentioned above, you should use the compiler's math functions from math.h. Then you can be sure that it uses the right float format.

Furthermore you can make your live much easier if you omit all the OpenLibrary and CloseLibrary calls and instead add -lauto to the vbcc command line.

[Yesideez]

btw, this worked fine but I cant' seem to use math.h as I've had to use option -nostdlib ???

#include <stdio.h>
#include <math.h>

int main() {
  double i,offset1,offset2,temp; 
  int j;
  char *funky="sunspot2014";
  int funkylen=strlen(funky);
  while (1) {
    i+=0.1;
    offset1=sin(i)*36 + 36;
    offset2=cos(i*1.1)*36 + 36;
    if (offset2<offset1) {
      temp=offset1;
      offset1=offset2;
      offset2=temp;
    }
    for (j=0; j<offset1; j++) {
      printf(" ");
    }
    for (; j<offset2; j++) {
      printf("\x1b[3%cm%c",(char) ((((int)(10*sin(i*0.9)+10)+j)&7)+49),funky[j%funkylen]);
    }
    printf("\n");
  }
}

[thomas]

Who does force you to use -nostdlib ?

[Yesideez]

Not sure to be honest. I have no idea what I'm doing andI'm just going by presious examples I've found. It's worked so I've carried on with it

Thanks massively for the -lauto, it's made a huge diference

Attached is my latest program as-is and here's how I've compiled it:

vc -sc -sd -O2 -lmsoft -lauto //targets/m68k-amigaos/lib/minstart.o test.c -o test -nostdlib

[Leffmann]

Just FYI, you may want to avoid -O1 and higher optimization modes with VBCC, as they are not entirely stable and will sometimes generate erroneous code. Only the default optimization mode seems to work correctly.

[thomas]

Why don't you keep it simple? All the options you use are for experts. You are not yet an expert are you?

vc test.c -o test

undefined reference to math functions -> add -lmsoft
undefined reference to library bases -> add -lauto

vc test.c -o test -lmsoft -lauto

Done.

Don't use options which you don't understand.

BTW, undefined reference to OpenScreenTags, OpenWindowTags, Printf and the like -> add -lamiga

[Yesideez]

Thanks for the input so far but now it doesn't work :/

vc -lamiga -lauto test.c -o test

When it did compile without errors, it crashed the Amiga. I ran it through monam and it went from opening dos.library to tryign to use OpenScreenTagList without opening any other libraries :/

[thomas]

How could you compile it at all. I get an error about DOSBase being defined twice:

6> vc -lamiga -lauto test.c -o test
Error 19: T:t_6_0.o: Global symbol _DOSBase from t_6_0.o is already defined in startup.o.
vlink fehlgeschlagen Rückgabewert 20
vlink -bamigahunk -x -Bstatic -Cvbcc -nostdlib -Lvlibos3: vlibos3:startup.o "T:t_6_0.o"   -lamiga -lauto -s -Rshort -lvc -o test failed
QNAS:Sources
6>

That's exactly your problem: you removed GfxBase, IntuitionBase and the other libraries, but why did you keep DOSBase?

[phx]

Originally Posted by Leffmann

Just FYI, you may want to avoid -O1 and higher optimization modes with VBCC, as they are not entirely stable and will sometimes generate erroneous code. Only the default optimization mode seems to work correctly.

No software of this complexity is without any errors, but compiling with -O1 is as stable and as tested as compiling without optimization (maybe even more). -O2 and higher are less tested and also difficult to debug.

[Leffmann]

I didn't mean to imply that VBCC was not a good compiler, it's all above my head and I'm impressed by how Volker and you have made a complete C99-capable tool-chain with such a small footprint. It was just a helpful advice since these kind of errors are very hard to debug.

I did send you a bug-report about it some time ago. You may have misinterpreted it as VBCC just generating a bit of inefficient code, but it's an actual case of erroneous code from using -O1. Let me know if you need me to resend the mail.

[phx]

Originally Posted by Leffmann

I did send you a bug-report about it some time ago. You may have misinterpreted it as VBCC just generating a bit of inefficient code, but it's an actual case of erroneous code from using -O1. Let me know if you need me to resend the mail.

Did I reply to it? It certainly doesn't hurt to remind me, when something takes too long. Although I guess that I already forwarded the bug to Volker, but I didn't hear anything from him since months... (which currently delays the planned V0.9c release).