[dpdk-dev] [PATCH v2 2/6] eal: add header files to support windows

Thomas Monjalon thomas at monjalon.net
Thu Mar 7 12:33:25 CET 2019


07/03/2019 11:24, Bruce Richardson:
> On Thu, Mar 07, 2019 at 09:45:57AM +0100, Thomas Monjalon wrote:
> > 07/03/2019 04:27, Anand Rawat:
> > > On 3/6/2019 3:31 AM, Thomas Monjalon wrote:
> > > > 06/03/2019 05:16, Anand Rawat:
> > > >> Added header files to support windows on x86 platforms.
> > > >> Updated rte_config to include rte_windows.h for windows
> > > >> build.
> > > > [...]
> > > >> --- a/config/rte_config.h
> > > >> +++ b/config/rte_config.h
> > > >> +/* windows specific*/
> > > >> +#ifdef RTE_EXEC_ENV_WINDOWS
> > > >> +#include <rte_windows.h>
> > > >> +#endif
> > > > 
> > > > Include in the config file looks wrong. >
> > > >> --- /dev/null
> > > >> +++ b/lib/librte_eal/windows/eal/include/exec-env/rte_windows.h
> > > > 
> > > > I think we could remove the sub-directory exec-env.
> > > > Could we include this file from rte_common.h?
> > > rte_windows.h defines types and substitution macros which are
> > > needed to support common code on windows. So it should be
> > > included as a global include for every library on windows.
> > > rte_common.h is not included in all the source code and headers
> > > we currently build for windows.
> > 
> > I think it is not an issue adding some rte_common.h include
> > here and there.
> > 
> 
> I'd be hesitant about putting it in rte_common.h without knowing the
> scope of the changes - "here and there" could end up being "everywhere".

It's already almost everywhere, and I think it's normal,
it is the meaning of "common".

% git grep rte_common.h | wc -l
459

% git grep rte_common.h lib | fgrep .h: | wc -l
79

If some .c files doesn't have rte_common.h included directly or
indirectly from other includes, it may be fixed.

> Another alternative is to see if most/all the definitions could actually be
> put in dpdk_conf and thereby go into the standard config.h file generated
> at build time. Anything that couldn't be done via macros or defines in the
> config could then look to be put maybe in rte_common.h or other locations,
> perhaps.

I think we should keep the semantic of what a config file is.
I don't see the need of adding more than strict config in it.




More information about the dev mailing list