Possible Bad _HAssert in cpukit/score/src/heapgetinformation.c ?

Phillip Sorensen pas37 at cornell.edu
Tue Oct 2 20:10:13 UTC 2007


Hello All,

I am have problems with an assertion in the heap code when I compile 
against RTEMS built with --enable=debug.  I am trying to get both the 
free space and the allocated space from the malloc heap using the 
rtems_region_get_information command.  This is the code:

long rtems_mem_get(double *total, double *used, double *free)
{
      region_information_block heap;


      if(!rtems_region_get_information(RTEMS_Malloc_Heap, &heap)) {
	  *used = (double)heap.Used.total / (1024 * 1024);
	  *free = (double)heap.Free.total / (1024 * 1024);

	  *total = *used + *free;

	  return 0;
      } else {
	  *total = 0;
	  *used = 0;
	  *free = 0;

	  return -1;
      }
}


When I run the code I get an assertion from the code 
cpukit/score/src/heap.c. The first _HAssert call in 
_Heap_Get_information() causes the assertion.

Heap_Get_information_status _Heap_Get_information(
   Heap_Control            *the_heap,
   Heap_Information_block  *the_info
)
{
   Heap_Block *the_block = the_heap->start;
   Heap_Block *const end = the_heap->final;

   _HAssert(the_block->prev_size == HEAP_PREV_USED);


If I check with a debugger the_block->prev_size is always 4.

If I compile RTEMS without --rtems-debug (which makes the _HAssert code 
due nothing), the code runs as it should.  Is this a bug in the 
_HAssert's, or am I lucky that my code runs.

I can work around this by using RTEMS compiled without --rtems-debug, 
but I would prefer to have the asserts and other debugging code active 
for development.

Phil Sorensen
CHESS




More information about the users mailing list