[PATCH 2/5] testsuites/libtests: Remove floats from libdl tests

[Sam Price]

Just curious instead of doing #ifdef __microblaze__
could you do
#ifndef __extendsfdf2

Or something similar to determine if the floating point symbols are known?

This would make it more portable.

On Mon, Aug 28, 2023 at 10:11 AM Alex White <alex.white at oarcorp.com> wrote:
>
> On Sun, Aug 27, 2023 at 11:48 PM Chris Johns <chrisj at rtems.org> wrote:
> >
> > On 28/8/2023 2:12 pm, Alex White wrote:
> > > On Sun, Aug 27, 2023 at 8:38 PM Chris Johns <chrisj at rtems.org> wrote:
> > >> The floats need to stay to make sure there are no issues with this type. I see
> > >> the tests with floats as valid.
> > >>
> > >> Is the only missing function `__extendsfdf2`?
> > >
> > > Yes, that is the only missing function. If I change the floats to doubles, the
> > > need for `__extendsfdf2` goes away and the tests pass. Is that an acceptable
> > > solution?
> >
> > I would happy with doubles being added so they are tested as well but that is
> > also outside the scope of your effort :)
> >
> > Is it easier to add a Microblaze conditional for the use of a double to the base
> > image and to forced the symbol to be linked in?
>
> I don't think it's too hard to add a conditional to the base image. For
> instance, this modification to dl07 is enough to make it pass on MicroBlaze:
>
> diff --git a/testsuites/libtests/dl07/init.c b/testsuites/libtests/dl07/init.c
> index 8db106ba29..7fb1500660 100644
> --- a/testsuites/libtests/dl07/init.c
> +++ b/testsuites/libtests/dl07/init.c
> @@ -67,6 +67,14 @@ static void Init(rtems_task_argument arg)
>
>    TEST_BEGIN();
>
> +#ifdef __microblaze__
> +  /* This is just here to link `__extendsfdf2` into the base image. */
> +  {
> +    float f = 0.3333;
> +    printf("f: %f\n", f);
> +  }
> +#endif /* __microblaze__ */
> +
>    te = rtems_tarfs_load("/", (void *)TARFILE_START, (size_t)TARFILE_SIZE);
>    if (te != 0)
>    {
>
> Is that what you were thinking?
>
> Alex
> _______________________________________________
> devel mailing list
> devel at rtems.org
> http://lists.rtems.org/mailman/listinfo/devel



-- 
Sincerely,

Sam Price