rtems-tester coverage patches break or1k/sis runs

[Chris Johns]

On 2/10/2014 3:44 am, Joel Sherrill wrote:
> Krzysztof/Hesham,
>
> And devel@ since this needs to be public now.
>
> I am trying to test or1ksim and sis using rtems-tester.
> We need to get Krzysztof's patches merged and I am
> just trying to run down all the issues I can.
>
> You both have made some changes to the repository
> and I am having issues. Krzysztof's are on a branch
> for me and I have attached them.  Except for the note
> before the remains of the old email, all issues are
> with Krzysztof's changes.
>
> Hesham's are all committed.
>
> $ ~/rtems-4.11-work/rtems-tools/tester/rtems-test --log=or1ksim.log
> --rtems-bsp=sis --rtems-tools=/users/joel/rtems-4.11-work/tools
> sparc-rtems4.11/c/sis/testsuites/samples/hello/
> RTEMS Testing - Tester, v0.2.0
> error: gdb.cfg:59: macro '%{_coverage}' not found
> warning: switched to dry run due to errors
> error: gdb.cfg:59: invalid if bool value:  %if %{_coverage}
> [1/1] p:0 f:0 t:0 i:0 | sparc/sis: hello.exe
>

There is something missing in the default configuration. The pattern 
often used in this case is to define a default if the macro is not 
defined before any logic that uses it.

> When I use Krzysztof's branch with the or1ksim as shown below,
> it adds a 1 to the end of the qemu-system-or32 invocation.
>
> On 10/1/2014 11:19 AM, Hesham Moustafa wrote:
>> I cloned a vanilla RTEMS tester, and it works fine with me with the
>> following command
>> "~/development/rtems/test/rtems-tester/tester/rtems-test
>> --log=or1ksim.log --rtems-bsp=or1ksim
>> /home/hesham/build/or1k-rtems4.11/c/or1ksim/testsuites/"
>>
> But it seemed to run a long time and then I killed it when
> I used your command.
>
> And when I try to run it on sis, you do need the --rtems-tools
> argument to find the simulator. I don't know what the difference
> for this is when doing the or1ksim. I think it should be required
> for consistency.

Yes, dependence on environment variables should not occur.

Chris