Bug In Signed Comparison

[Dark.Rider 0]

Hello!

 

Here is my problem, probably a bug:

 

Bug description:

a comparison between two signed chars seems not to work correclty

 

Steps to reproduce:

here my code sniplet:

 

signed char DefaultAction(signed char cIst, signed char cSoll)

{

signed char cAccel;

if(cIst == cSoll) return(0);

 

cAccel=1;

if(opposite(cIst,cSoll)) cAccel=4;

 

if(cSoll==0) cAccel=2;

 

 

//HERE IS THE PROBLEM

if(cIst>cSoll) cAccel=-cAccel;

//Example: cIst=-90 cSoll=0 and the statement gets TRUE

//(cAccel old value=2 new value=-2) 100% tested with the debugger

//Assembler code for this statement

//006E 0832 MOVF DefaultAct_00005_arg_cIst, W

//006F 1003 BCF STATUS,C

//0070 1BB2 BTFSC DefaultAct_00005_arg_cIst,7

//0071 0233 SUBWF DefaultAct_00005_arg_cSoll, W

//0072 1FB3 BTFSS DefaultAct_00005_arg_cSoll,7

//0073 0233 SUBWF DefaultAct_00005_arg_cSoll, W

//0074 1803 BTFSC STATUS,C

//0075 2879 GOTO label268436421

//0076 0934 COMF DefaultAct_00005_1_cAccel, W

//0077 00B4 MOVWF DefaultAct_00005_1_cAccel

//0078 0AB4 INCF DefaultAct_00005_1_cAccel, F

//0079 label268436421

 

if(bSpinTurn) cAccel+=cAccel;

 

return(cAccel);

}

 

 

Expected behaviour:

It should do the opposithe thing

 

Is the problem 100% reproduceable:

It happens EVERY time at this statement

 

IDE version: SourceBoost IDE version 6.01

Compiler: BoostC

Target device: PIC16F873

OS: Win2k

 

Comments:

Please help! I would like to change from cc5x because of limitations and price. But the compiler should work correctly

[Pavel 0]

What is 'opposite'?

 

Regards,

Pavel

[Dark.Rider 0]

Hello!

 

What is 'opposite'?

<{POST_SNAPBACK}>

 

signed char opposite(signed char a, signed char B )

{

if((a^ B ) <0 && a && B ) return TRUE;

return FALSE;

}

 

I commented it out, no change.

 

 

 

I just tested a simple program:

 

#include <system.h>

 

signed char a,b,c;

 

main()

{

a=-90;

b=0;

c=0;

if( a>b ) c=1;

for(;;

}

 

c becomes '1'!!! The comparison seems be done unsigned.

 

Here the casm:

main()

 

{

a=-90;

0003 30A6 MOVLW 0xA6

0004 1283 BCF STATUS, RP0

0005 1303 BCF STATUS, RP1

0006 00A0 MOVWF gbl_a

 

b=0;

0007 01A1 CLRF gbl_b

 

c=0;

0008 01A2 CLRF gbl_c

 

if(a> c=1;

0009 0820 MOVF gbl_a, W

000A 1003 BCF STATUS,C

000B 1BA0 BTFSC gbl_a,7

000C 0221 SUBWF gbl_b, W

000D 1FA1 BTFSS gbl_b,7

000E 0221 SUBWF gbl_b, W

000F 1803 BTFSC STATUS,C

0010 2813 GOTO label268436063

0011 3001 MOVLW 0x01

0012 00A2 MOVWF gbl_c

0013 label268436063

 

for(;;

0013 2813 GOTO label268436063

 

}

 

regards

Thomas Kusch

Edited June 23, 2005 by Dark.Rider

[Pavel 0]

Yes this looks like a bug Will work on it today. A compiler fix will be released as soon as this one gets fixed. Thanks for reporting.

 

Regards,

Pavel