[arm-gnu] Cortex-M3 and Floating Point Problems

[Mark Mulrooney]

I have manually implemented the underlying _sbrk function. I've pasted the code below just in case I've made some silly mistake but as far as I know it seems to be working.


/**
 ********************************************************************************************************
 * @fn      _sbrk(int incr)
 * @brief   Increase program data space. Malloc and similar functions depend on this
 * @param   incr:   The amount of space to grab
 * @return  Returns the address to the allocated space on success
 * @return  Returns -1 on failure
 ********************************************************************************************************
 */
caddr_t _sbrk(int incr)
{
    unsigned char *prev_heap;
    static unsigned char *heap = NULL;

    /* If the heap var is NULL, set it to the start of the heap */
    if(heap == NULL)
    {
        heap = (unsigned char *)&_heap_start;
    }

    /* Save the current heap location */
    prev_heap = heap;

    /* Make sure the stack and the heap don't collide */
    if((heap + incr) > (unsigned char*) &_sstack)
    {
        return (caddr_t) -1;
    }

    /* Increase by the increment amount and return the address */
    heap += incr;
    return (caddr_t) prev_heap;
}

________________________________________
From: Krzysztof Wesołowski [krzysztof.wesolowski at rainlabs.pl]
Sent: Thursday, June 21, 2012 3:32 PM
To: Carlos O'Donell
Cc: Mark Mulrooney; arm-gnu at codesourcery.com
Subject: Re: [arm-gnu] Cortex-M3 and Floating Point Problems

> * And lastly... the CodeBench printf has a bug.
>

One more possible reason is that AFAIR printf with %f uses malloc for
memory allocation, and often malloc or underlying sbrk are not
prepared (by user) to work on MCUs (in many example projects I have
seen there was no such support and compiler reports undefined
reference, and then, sometimes not functional code is added just to
stop errors ;)).

Regards,
Krzysztof Wesołowski,