MQX 3.8 _time_get() ->_time_to_date defect

[timias]

I noticed the following behavior in my code using the MQX 3.8 and the  K60 BSP -didn't confirm other processors

occasionally the following pair of calls will fail, and the _time_to_date function will return FALSE. Though you wouldn't notice it since most MQX internal calls ignore the _time_to_date function return value. This seems to be the source of a ton of file time stamp glitches I have noticed.

I was using the following code and had to change it, but the problem still exist all over MQX. I will probably patch MQX myself.

_time_get(&time_mqx);

_time_to_date(&time_mqx, &clk_time);

My fix for the snippet was :

_time_get(&time_mqx);

/* Fix for a wonderful little MQX bug*/

    if (time_mqx.MILLISECONDS ==1000){

	time_mqx.MILLISECONDS =0;
	time_mqx.SECONDS++;

    }

_time_to_date(&time_mqx, &clk_time);

[timias]

FYI the fix was NOT included in MQX 4.0

[BielikM]

FYI: the fix has been applied and will be included in MQX 4.0.1 release.

[billnd]

Any hint as to when 4.0.1 will be released?

[BielikM]

Ideally 22.4.2013.

[JuroV]

Hi Robert,

you are right. On the other hand there are plans and already some progress in remaking timer concept in MQX.

[c0170]

Hello,

yes, it's not in MQX 4.0 as you stated. This will be fixed but first this bug requires thorough code review and perhaps to rewrite more parts of the code :smileywink:

Regards,

MartinK

[c0170]

Hello Robert,

Thank you for reporting the bug, I will dig into it during this week and I will return here with an outcome.

Regards,

MartinK

[c0170]

Hello,

can you replace function _psp_ticks_to_time in psp_tkti.c and rebuild psp and test this:

  boolean _psp_ticks_to_time

   (

      /* [IN] Pointer to the tick struct to store the results in */

      PSP_TICK_STRUCT_PTR tick_ptr,

      /* [OUT] Pointer to the time struct to convert */

      TIME_STRUCT_PTR     time_ptr

   )

{ /* Body */

   uint_64                tmp, hwt;

   uint_32                tps;

   KERNEL_DATA_STRUCT_PTR kernel_data;

   _GET_KERNEL_DATA(kernel_data);

   tps = kernel_data->TICKS_PER_SECOND;

   /* Saturate if ticks go out of range of time struct */

   if ( (tick_ptr->TICKS[0] / tps) > MAX_UINT_32) {

      time_ptr->SECONDS      = MAX_UINT_32;

      time_ptr->MILLISECONDS = 999;

      return FALSE;

   } /* Endif */

   /* */

   tmp = (tick_ptr->TICKS[0] * 1000) + (tick_ptr->HW_TICKS[0] * 1000 / kernel_data->HW_TICKS_PER_TICK);

   tmp = tmp / tps;

   time_ptr->SECONDS      = tmp / 1000;

   time_ptr->MILLISECONDS = tmp - time_ptr->SECONDS * 1000;

   return TRUE;

} /* Endbody */

Regards,

MartinK

[audi_mcavoy]

MartinK -- That seems to have fixed the problem for me too.  Thanks!

[timias]

That seemed to solve my problem, without adding any discernible side effects.

Thanks for the help