Bug in TivaWare/driverlib/adc.c? (ADCIntUnregister)

Veikko Immonen

Hi,

I'm looking at the ADC module, and when checking what the driverlib functions actually do, I stumbled across this thing that I think is a bug. It seems that the (internal to the driverlib) function _ADCIntNumberGet always returns the interrupt number from ADC0:

static uint_fast8_t
_ADCIntNumberGet(uint32_t ui32Base, uint32_t ui32SequenceNum)
{
    uint_fast8_t ui8Int;

    //
    // Determine the interrupt to register based on the sequence number.
    //
    if(CLASS_IS_BLIZZARD)
    {
        ui8Int = ((ui32Base == ADC0_BASE) ?
                  (INT_ADC0SS0_BLIZZARD + ui32SequenceNum) :
                  (INT_ADC0SS0_BLIZZARD + ui32SequenceNum));
                   ^^^^^^^^^^^^^^^^^^^^
    }
    else
    {
        ui8Int = 0;
    }

    return(ui8Int);
}

Shouldn't the caret-marked constant be INT_ADC1SS0_BLIZZARD instead?

Interestingly, that function is only used once, in ADCIntUnregister. In ADCIntRegister, where similar functionality is required, it is provided inline, without a call to _ADCIntNumberGet. In ADCIntRegister, the alternative result of the ternary operator has INT_ADC1SS0_BLIZZARD, as is expected.

I'm looking at TivaWare version 1.1.

Best,

Veikko

over 12 years ago

0 Chester Gillon over 12 years ago

Guru 92251 points

Veikko Immonen said:
Shouldn't the caret-marked constant be INT_ADC1SS0_BLIZZARD instead?

From looking at the code agree.

Veikko Immonen said:
I'm looking at TivaWare version 1.1

This appears to a bug introduced by TivaWare, since StellarisWare version 9453 doesn't have this bug.

In StellarisWare there is no _ADCIntNumberGet function and both ADCIntRegister and ADCIntUnregister use

    //
    // Determine the interrupt to unregister based on the sequence number.
    //
    ulInt = ((ulBase == ADC0_BASE) ? (INT_ADC0SS0 + ulSequenceNum) :
             (INT_ADC1SS0 + ulSequenceNum));

0 Veikko Immonen over 11 years ago

Expert 2415 points

Okay, now I've got to say I'm puzzled and disappointed. This bug has escalated, it now affects also ADCIntRegister, in the newest revision of TivaWare (2.1.0.12573).

ADCIntRegister too now calls _ADCIntNumberGet, as can be seen in the excerpt below:

void
ADCIntRegister(uint32_t ui32Base, uint32_t ui32SequenceNum,
               void (*pfnHandler)(void))
{
    uint_fast8_t ui8Int;

    //
    // Check the arguments.
    //
    ASSERT((ui32Base == ADC0_BASE) || (ui32Base == ADC1_BASE));
    ASSERT(ui32SequenceNum < 4);

    //
    // Determine the interrupt to register based on the sequence number.
    //
    ui8Int = _ADCIntNumberGet(ui32Base, ui32SequenceNum);
    ASSERT(ui8Int != 0);

    //
    // Register the interrupt handler.
    //
    IntRegister(ui8Int, pfnHandler);

    //
    // Enable the timer interrupt.
    //
    IntEnable(ui8Int);
}

_ADCIntNumberGet has been updated with new labels since my last post, but unfortunately the bug has not been corrected:

static uint_fast8_t
_ADCIntNumberGet(uint32_t ui32Base, uint32_t ui32SequenceNum)
{
    uint_fast8_t ui8Int;

    //
    // Determine the interrupt to register based on the sequence number.
    //
    if(CLASS_IS_TM4C123)
    {
        ui8Int = ((ui32Base == ADC0_BASE) ?
                  (INT_ADC0SS0_TM4C123 + ui32SequenceNum) :
                  (INT_ADC0SS0_TM4C123 + ui32SequenceNum)); // <-- problem!
    }
    else if(CLASS_IS_TM4C129)
    {
        ui8Int = ((ui32Base == ADC0_BASE) ?
                  (INT_ADC0SS0_TM4C129 + ui32SequenceNum) :
                  (INT_ADC1SS0_TM4C129 + ui32SequenceNum));
    }
    else
    {
        ui8Int = 0;
    }

    return(ui8Int);
}

Interestingly enough, the added code for the 129 series is correct. Not much proofreading done there?

I'd hope someone from the TivaWare team would care to comment. I've just spent a day debugging my uDMA-enabled ADC code that works perfectly fine with ADC0 but hangs in the interrupt handler when using ADC1 - no wonder anymore!

0 Luis Afonso over 11 years ago in reply to Veikko Immonen

Guru 20670 points

Agree, some more direct feedback from TI and preferably a member of TivaWare team would be appreciated.

0 Amit Ashara over 11 years ago in reply to Veikko Immonen

TI__Guru**** 244440 points

Hello Veikko,

I can confirm the issue (was raised in the forum a couple of months back) and this has been marked for fixing in the next TivaWare release.

Regards

Amit

0 Veikko Immonen over 11 years ago in reply to Amit Ashara

Expert 2415 points

Thanks Amit, good to know.

0 Luis Afonso over 11 years ago in reply to Amit Ashara

Guru 20670 points

Good to know bug reports get through :D

Arm-based microcontrollers

Arm-based microcontrollers forum

Bug in TivaWare/driverlib/adc.c? (ADCIntUnregister)