RTEMS test failures

[Joel Sherrill]

Kenneth Peters wrote:
> 
> I am running RTEMS 4.5.0 built for Sparc ERC32, both on Jiri Gaisler's
> instruction set simulator and on custom hardware with a very slightly
> tweaked bsp. The distributed tests sp26 and tm27 do not complete execution
> on either platform. I suspect the sp26 failure would also happen on other CPUs.
> 
> Test sp26 attempts to create tasks with four times the minimum stack, but
> system.h does not provide extra task stack space, so rtems_task_create()
> does not succeed. This I can fix easily enough, but I'd like an "official"
> patch so I don't have to explain to management how I tweaked the RTEMS
> verification suite.

Goto
http://www.oarcorp.com/cvsgraph/cvsgraphwrapper.php3?r=/usr1/CVS&m=rtems&f=c/src/tests/sptests/sp26/system.h

and pull out the patch that is difference between 1.4 and 1.5 of the
file sp26/system.h.  It was never merged on the 4.5.0 branch but
fixed in Nov 2001.

> Test tm27 is more complicated and uses internal RTEMS variables, so I
> haven't fully tracked down the problem and hope an RTEMS expert can. The
> test prints out the two "returns to interrupted task" messages, but then
> hangs, seemingly waiting for the interrupt to occur. Looks likely to be
> related to things in the specific bsp (MUST_WAIT_FOR_INTERRUPT,
> ERC32_Force_interrupt).

Use CVSweb to browse the history for the ERC32 BSP.  In particular look 
at the .h files.  I think the fix was tracked as RTEMS GNATS PR80 at
http://www.oarcorp.com/cgi-bin/gnatsweb.pl (use guest/guest).  

I saw this fix as between version 1.7.2.1 and 1.7.2.2 of
erc32/include/bsp.h
on the 4.5 branch and between 1.12 and 1.13 on the mainline.

I hope getting the fixes from RTEMS source code control repository makes
your
management happy. :)  

What are you guys doing with RTEMS?

> Ken Peters
> Ken.Peters at jpl.nasa.gov

--joel