[avr-libc-dev] Re: [avr-gcc-list] Odd itoa/strrev problem with ATmega8535

[Dmitry K.]

At 15 Jun 2003 03:39 Josh Pieper wrote:
> I haven't tried reduce this problem to the most simple test case yet,
> but was wondering if anyone has seen something similar and has a tip.
> I'm using itoa on an ATmega8535 with the UART receive and timer 0
> compare interrupt enabled.  Every 20 or 30 itoa's, it returns a string
> that is reversed.  i.e. instead of "-49" it will return "94-".
>
> I'm using a cvs snapshot of avr-gcc and avr-libc dated 20030512.
>
> I have already verified that there is no way that the stack is
> overrunning data, and I am using SIGNAL to define the interrupt
> handlers.  Any quick tips, or should I reduce it to a simple test case
> and post that?
>

Good afternoon!

I have checked up work "itoa" on the plate with ATmega163.
Works correctly, including "-49".

But on realization (itoa.S) there are two remarks:

1.  There is no clearing "T" in a case when "radix" it is not equal 10. It 
results, for example, in the following:
        itoa(0, s, 10)  -->  "0"
        itoa(-2, s, 8)  -->  "177776"   (T was 0)
        itoa(-2, s, 10)  --> "-2"               (T set 1)
        itoa(-2, s, 8)  --> "-177776"

2.  Erroneous "radix" results in unpredictable consequences: a call "strrev" 
with "value" as the address of a string.  Apparently, instead of:
        cpi     r_radix, 2      ; no radix < 2
        brlt    terminate
it is necessary to put:
        cpi     r_radix, 2      ; no radix < 2
        brlt    n_terminate

Best wishes.