LOG10 library function problem

[aclamb]

Hi all - I have done some C programming a while ago, so I combined my
interests in PICs with some exercises to refresh C in my mind as well as
hopefully allow me to do some things I would like to with PICs.

All has gone well (several working projects - simple but working) _UNTIL_
I sez self why don't you add a 'simple' calculation of dB(W) to that
display? I can't seem to get it (log10())to do the most basic of
calculations. I get WILDly errant answers. Here's the latest (boiled down
to DEAD simple): The display output follows.

MPLAB 7.51
CCS C PCM 4.023
PIC16F877
ICD2 (as programmer)
Development board - DHMicro 40pin generic breadboard

Code sample:

#include <16F877.h>
#fuses HS,NOWDT,NOPROTECT,NOLVP
#use delay(clock=4000000)

#include <math.h>
#include <lcd.c>


void main() {
float fwd, pf;

lcd_init();

while(TRUE)
{
pf = 120.0;
fwd = log10(pf);

lcd_gotoxy(1,1);
printf(lcd_putc,"%3.0g lg%7.3g",pf, fwd);
delay_ms(50);

pf = 10.0;
fwd = log10(pf);
printf(lcd_putc,"\n%3.0g lg%7.3g",pf, fwd);
delay_ms(50);

}
}

LCD Display shows (16X2):

120 lg 0.172
10 lg -32.942

The LCD is a generic 16X2 that I have been using successfully for a while.
(it seems to work fine with the 'canned' driver)

I searched the forem and couldn't find anyone else complaining that the
log() or LOG10() is completely busted so I'm sure its me! I can't see what
I'm doin wrong to save me... PLEASE help!!

[PCM programmer]

Here is a thread on it. This one will help:
http://www.ccsinfo.com/forum/viewtopic.php?t=29740&highlight=log

[aclamb]

AWSOME! Thanks!

Gotta make an editorial comment - sorry!
Quality Control is expensive! (regression testing for example)
I worked at a VERY large computer / software company.
The budget / manpower required for the QC group nearly
equalled the compiler group (not to mention how thankless a
job it is). Unfortunately, QC and customer services are
early on the list for downsizing and low on the list for
increases come budget time. As a customer, I'm not
pleased when I get stung. As a business man, I'm afraid
I understand the issues.
In my limited experience, I like the PCM compiler very much.

[PCM programmer]

Report the bug to CCS support. I'm not sure if anyone has done that yet.

[Ttelmah]

Reasonably comprehensive 'QA', is hard. However basic functionality testing for a compiler should be very easy indeed. The chip can do a lot of it 'itself'. All that is needed is a small suite of programs. A typical one, would comprise a simple set of maths operations, together with a constant declaration of the input values, and output values. It simply runs through the input values, and checks that the value returned by the function is 'as expected'. Similar programs can test the operation of many 'core' parts of the compiler. Quite a while ago, there were complaints about some other functions being wrong in some versions, and it was said that CCS were improving such 'self testing', and going to apply these to new releases. Obviously these 'good intentions', went 'by the board' at some point...
A a user, I have written several such programs to test 'core bits' that I need to work. Currently, serial I/O, pointer operations, array accesses, some arithmetic, basic looping and control flow. I recently added some for pointers to functions, constant string handling, typemod/addressmod. Now the 4.xxx compilers, are just starting to show signs of being able to pass many of these tests. a few though (like the arithmetic), have gone 'backwards'. However as has been said many times here before, the 'best' compiler at present, unless you 'need' something offered by V4, is 3.249...

Best Wishes

[thomasj50]

One last message - I reported the float problem to CCS and they sent me a new math.h - but it is still wrong ! Using the corrections I got from this forum works fine though. CCS used a cast of (int8), but the correct one should be (int8*).
Best regards,
Thomas Johansson

[thomasj50]

The LOG problem, sorry.
thomasj