On Tue, Dec 12, 2017 at 02:07:52PM +0000, Bruce Richardson wrote:
> On Mon, Dec 11, 2017 at 02:36:15PM -0500, Neil Horman wrote:
> > Hey all-
> > A few days ago, I was lamenting the fact that, when reviewing patches I
> > would frequently complain about ABI changes that were actually considered
> > safe
> > because they were part of the EXPERIMENTAL api set. John M. asked me then
> > what
> > I might do to improve the situation, and the following patch set is a
> > proposal
> > that I've come up with.
> >
> > In thinking about the problem I identified two issues that I think we
> > can improve on in this area:
> >
> > 1) Make experimental api calls more visible in the source code. That is to
> > say,
> > when reviewing patches, it would be nice to have some sort of visual
> > reference
> > that indicates that the changes being made are part of an experimental API
> > and
> > therefore ABI concerns need not be addressed
> >
> > 2) Make experimenal api usage more visible to consumers of the DPDK, so that
> > they can make a more informed decision about the API's they consume in their
> > application. We make an effort to document all the experimental API's, but
> > there is no guarantee that a user will check the documentation before
> > making use
> > of a new library.
> >
> > This patch set attempts to achieve both of the above goals. To do this I've
> > added an __experimental macro tag, suitable for inserting into api forward
> > declarations and definitions.
> >
> > The presence of the tag in the header and c files where the api code resides
> > increases the likelyhood that any patch submitted against them will include
> > the
> > tag in the context, making it clear to reviewers that ABI stability isn't a
> > concern here.
> >
> >
> > Also, This tag marks each function it is used on with an attibute causing
> > any
> > use of the fuction to emit a warning during the build
> > with a message indicating that the API call in question is not yet part of
> > the
> > stable interface. Developers can then make an informed decision to suppress
> > that warning or not.
> >
> > Because there is internal use of several experimental API's, this set also
> > includes a new override macro ALLOW_EXPERIMENTAL_APIS to automatically
> > suprress these warnings. I think its fair to assume that, for internal
> > use, we
> > almost always want to suppress these warnings, as by definition any change
> > to
> > the apis (even their removal) must be done in parallel with an appropriate
> > change in the calling locations, lest the dpdk build itself break.
> >
> > Neil
> >
> > ---
> > Change Notes:
> > v2)
> > * Cleaned up checkpatch errors
> > * Added Allowance for building experimental on BSD
> > * Swapped Patch 3 and 4 so that we didn't have a commit level that issued
> > warnings/errors without need
> >
> > v3)
> > * On suggestion from Bruce, modify ALLOW_EXPERIMENTAL_APIS to be defined in
> > CFLAGS rather than a makefile variable. This is more flexible in that it
> > allows us to suppress this specific feature rather than all uses of the
> > deprecated attribute, as we might use it for other features in the furute
> >
>
> Despite the fact that this is making yet more work for porting to a new
> build system, I think this is a good idea to have. As such,
>
> Acked-by: Bruce Richardson <bruce.richard...@intel.com>
>
>
Thomas-
I just noticed that the ci tests are failing on the intel compiler, which
makes very little sense to me, as the error is a permission error on a bash
script that added in this series, which works during the gcc compilation. Can
you take a look at that please?
thanks
Neil
