[PATCH 0/3] Fixed ability to enable console polling for certain BSPs

[Sebastian Huber]

On 26/03/2019 19:11, Joel Sherrill wrote:
>
>
> On Tue, Mar 26, 2019 at 11:55 AM Sebastian Huber 
> <sebastian.huber at embedded-brains.de 
> <mailto:sebastian.huber at embedded-brains.de>> wrote:
>
>     ----- Am 26. Mrz 2019 um 16:17 schrieb joel joel at rtems.org
>     <mailto:joel at rtems.org>:
>
>     > On Tue, Mar 26, 2019 at 9:36 AM Sebastian Huber <
>     > sebastian.huber at embedded-brains.de
>     <mailto:sebastian.huber at embedded-brains.de>> wrote:
>     >
>     >> On 26/03/2019 14:56, Lou Woods wrote:
>     >> > From: Lou Woods <Lou.Woods at OARCorp.com>
>     >> >
>     >> > I discovered that I was not able to turn off interrupt-based
>     console
>     >> mode on the
>     >> > Xilinx Zynq BSP under Qemu via the configure command.  The
>     fix simply
>     >> exchanges
>     >> > #ifdef XXX_CONSOLE_USE_INTERRUPTS for
>     >> > #if XXX_CONSOLE_USE_INTERRUPTS for the BSP specific console code.
>     >>
>     >> You can disable it via the configure command line, just use
>     >>
>     >> XXX_CONSOLE_USE_INTERRUPTS=
>     >>
>     >> I have never seen setting it to nothing ever documented anywhere in
>     > the RTEMS pantheon of material. All examples I have ever seen set
>     > these to a value
>     >
>     >
>     >> These feature defines are not uniformly used, some use #if
>     other use
>     >> #ifdef.
>     >>
>     >
>     > Changing these ifdef CONSOLE_USE to #if ensures they work if the
>     value
>     > is set to nothing or 0. I don't see how this isn't an improvement.
>
>     I just wanted to point out that this patch changes something that
>     works into something else that works also. 
>
>
> Works for more cases. The configure.ac <http://configure.ac> sets it 
> to 1. Setting it to 0 is a logical choice looking at the configure.ac 
> <http://configure.ac> for what to do.
>
> If I didn't know XXX= results in undefined, I doubt that is common 
> knowledge.

If you need to tinker with the BSP options as a normal user you already 
have a problem. How do you know that you can call

./c/src/lib/libbsp/arm/xilinx-zynq/configure --help

to get the available options? Only a hand full of BSPs have a 
documentation section in

https://docs.rtems.org/branches/master/user/bsps/index.html

Anyway, the patch set is fine, but it would be nice to think also about 
improving the documentation since this is the root cause for the problem 
from my point of view.

>     The change is somewhat arbitrary.  It should be accompanied with a
>     recommendation in the BSP developer documentation.  There should
>     be also a ticket for this to change everything from #ifdef XYZ to
>     #if XYZ, if we agree that this is actually better.  Only for BSP
>     options, for stuff which is available via <bsp.h>, or does it
>     stop, when we change all the #if defined(RTEMS_SMP) to #if RTEMS_SMP?
>
>
> Setting something to 0 and using ifdef do not evaluate to the same 
> thing so for BSP options, I would tend to lean to #if since setting it 
> to 0 and nothing are equivalent. Plus may of the BSP options do take 
> values so this helps make them consistent.
>
> For the RTEMS feature flags, there is no associated value and never 
> has been. It is a simple defined=true, not defined=false. I know there 
> is no functional difference, but I would say that those should stay to 
> ifdef since that's the intent. They also are directly from configure 
> arguments which are explicitly enable/disable so no oddity with user 
> set values.

This means the user interface to set the defines determines if a feature 
define uses undefined/defined or a zero/non-zero value.

-- 
Sebastian Huber, embedded brains GmbH

Address : Dornierstr. 4, D-82178 Puchheim, Germany
Phone   : +49 89 189 47 41-16
Fax     : +49 89 189 47 41-09
E-Mail  : sebastian.huber at embedded-brains.de
PGP     : Public key available on request.

Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.