beta3 test results for efi332 bsp

Joel Sherrill joel.sherrill at OARcorp.com
Sat May 13 14:20:03 UTC 2000



"John S. Gwynne" wrote:
> 
> I've run the test programs included with 4.5.0-beta3 with the efi332
> bsp, and am happy to say most ran fine. Most if not all of the problem
> are a direct result of a limited amount RAM in this target (512k). In
> short, unlimited.nxe, psxfile01.nxe, psxmount.nxe, and psxstat.nxe
> appear too large, and rtems++.nxe and psxsem01.nxe _may be_ too
> large. It would be helpful to small board without much memory to split
> those tests into two or more smaller tests that would run on all
> boards. The failure of sp28.nxe, rtems++.nxe, and pscsem01.nxe are not
> as clear. More details follow....

Do you have itron and posix enabled? A key part of the memory
consumption
is that each API defaults to configuring 10 tasks.  With 3 APIs enabled,
that is 30 tasks and 30 TCBs.  On top of that, most tests do not require
20 open file descriptors.  

Perhaps adding a "minimal memory defaults" configuration option would
help. All of the above are simply convenience choices in confdefs.h
to allow MOST applications to run OK.

I do not expect that unlimited or some of the file tests will run in 
low RAM targets.  In particular, the filesystem test which attempts
to extend a file until it runs out of memory will only work if the
IMFS is configured to use small "block sizes".  Otherwise, the 
maximum filesize is VERY large. :)
 
> Thanks,
> john gwynne
> jsg at donet.com
> 
> These tests failed without a clear indication of the problem:
> -------------------------------------------------------------
> 
> sp28.nxe: fails ((int)taskvar == localvar) test near line 58 of init.c
>  prints:
>    Task:3 taskvar:300 localvar:300
>    Task:2 taskvar:200 localvar:200
>    Task:1 taskvar:100 localvar:100
>  then suspends.
> 
>  Question: should rtems_task_variable_delete clear these? does this
>            test really work?
> 
> rtems++.nxe: Bus Error....
>  prints:
>    *** RTEMS++ TEST ***
>    INIT - Task.create() - RTEMS[19] invalid thread priority
>    INIT - Task.create() - RTEMS[00] successful completion
>    INIT - Task.create() - RTEMS[16] illegal on calling thread
>    INIT - Task.restart() - RTEMS[14] thread is in wrong state
>    INIT - Task.start(0xDEADDEAD) - RTEMS[00] successful completion
>    INIT - Destroy it's self
> 
>    RTEMS: Spurious interrupt: Bus Error
>    RTEMS:    Vector: 0x00000002 sp: 0x000d21f8
> 
>  cause: The cause at the moment is unknown, but could be the result
>         insufficient memory and failed malloc/constructor. This test use to
>         work; is it still working on other targets?
> 
> 
> psxsem01.nxe: bus error
>   prints:
>     *** POSIX SEMAPHORE MANAGER TEST 1 ***
>     Init: sem_init - SUCCESSFUL
> 
>     RTEMS: Spurious interrupt: Bus Error
>     RTEMS:    Vector: 0x00000002 sp: 0x000bed88
> 
>  cause: The cause at the moment is unknown, but could be the result
>         insufficient memory and failed malloc. Is this working on other
>         targets?
> 
> These tests are believed to use too much RAM:
> ---------------------------------------------
> 
> unlimited.nxe: HEAP error, not enough RAM
> 
> psxfile01.nxe: lack of memory?
>   fails part way through with...
>   assertion "memory" failed: file "../../../../../../rtems-4.5.0-beta3/c/src/lib/libc/memfile.c", line 340
> 
> psxmount.nxe: lack of memory?
>   fails part way through with...
>   assertion "memory" failed: file "../../../../../../rtems-4.5.0-beta3/c/src/lib/0
> 
> psxstat.nxe: lack of ram?
>   prints:
>     <delete>
>     Attempting to mount IMFS file system at /dir1/my_mount_point
>     rmdir /dir1/my_mount_point should fail with EBUSY
>     Unmount /dir1/my_mount_point
>     assertion "status == 0" failed: file "../../../../../../../rtems-4.5.0-beta3/c/src/tests/psxtests/psxstat/test.c", line 669

-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel at OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
   Support Available             (256) 722-9985



More information about the users mailing list