From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: <nhorman@tuxdriver.com> Received: from smtp.tuxdriver.com (charlotte.tuxdriver.com [70.61.120.58]) by dpdk.org (Postfix) with ESMTP id CDE1E44CF for <dev@dpdk.org>; Mon, 22 Jan 2018 02:34:58 +0100 (CET) Received: from cpe-2606-a000-111b-4011-215-ff-fecc-4872.dyn6.twc.com ([2606:a000:111b:4011:215:ff:fecc:4872] helo=localhost) by smtp.tuxdriver.com with esmtpsa (TLSv1:AES256-SHA:256) (Exim 4.63) (envelope-from <nhorman@tuxdriver.com>) id 1edR0e-00064Z-9g; Sun, 21 Jan 2018 20:34:55 -0500 Date: Sun, 21 Jan 2018 20:34:17 -0500 From: Neil Horman <nhorman@tuxdriver.com> To: Thomas Monjalon <thomas@monjalon.net> Cc: dev@dpdk.org, Ferruh Yigit <ferruh.yigit@intel.com>, john.mcnamara@intel.com, bruce.richardson@intel.com Message-ID: <20180122013417.GB5415@neilslaptop.think-freely.org> References: <20171201185628.16261-1-nhorman@tuxdriver.com> <a945deb9-ecee-9538-c3a4-dd3905bd0fa4@intel.com> <20180112124459.GA20015@hmswarspite.think-freely.org> <2231669.8C7kkI4IjB@xps> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <2231669.8C7kkI4IjB@xps> User-Agent: Mutt/1.9.1 (2017-09-22) X-Spam-Score: -2.9 (--) X-Spam-Status: No Subject: Re: [dpdk-dev] [PATCHv4 3/5] Makefiles: Add experimental tag check and warnings to trigger on use X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions <dev.dpdk.org> List-Unsubscribe: <https://dpdk.org/ml/options/dev>, <mailto:dev-request@dpdk.org?subject=unsubscribe> List-Archive: <http://dpdk.org/ml/archives/dev/> List-Post: <mailto:dev@dpdk.org> List-Help: <mailto:dev-request@dpdk.org?subject=help> List-Subscribe: <https://dpdk.org/ml/listinfo/dev>, <mailto:dev-request@dpdk.org?subject=subscribe> X-List-Received-Date: Mon, 22 Jan 2018 01:34:59 -0000 On Sun, Jan 21, 2018 at 07:54:44PM +0100, Thomas Monjalon wrote: > 12/01/2018 13:44, Neil Horman: > > On Fri, Jan 12, 2018 at 11:49:57AM +0000, Ferruh Yigit wrote: > > > On 1/11/2018 8:50 PM, Neil Horman wrote: > > > > On Thu, Jan 11, 2018 at 08:06:43PM +0000, Ferruh Yigit wrote: > > > >> On 12/13/2017 3:17 PM, Neil Horman wrote: > > > >>> --- a/app/test-eventdev/Makefile > > > >>> +++ b/app/test-eventdev/Makefile > > > >>> @@ -32,6 +32,7 @@ include $(RTE_SDK)/mk/rte.vars.mk > > > >>> > > > >>> APP = dpdk-test-eventdev > > > >>> > > > >>> +CFLAGS += -DALLOW_EXPERIMENTAL_APIS > > > >> > > > >> Do we need this internally in DPDK? For application developers this is great, > > > >> they will get warning unless explicitly stated that they are OK with it. > > > >> > > > > I'm not sure what you're asking here. As I mentioned in the initial post, I > > > > think we should give blanket permission to any in-tree dpdk library to allow the > > > > use of experimental API's, but that doesn't really imply that all developers > > > > would wan't it disabled all the time. That is to say, I could envision a > > > > library author who, early in development would want to get a warning issued if > > > > they used an unstable API, and, only once they were happy with their design and > > > > choice of API usage, turn the warning off. > > > > > > I got your point, but I think whoever using an experimental API in another > > > component in DPDK is almost always the author of the that experimental API, so > > > not sure this check is ever really needed within dpdk. > > > > > I would have thought so too, but it doesn't really bear up. The example I used > > to convince myself of a more granular approach was commit > > 9c38b704d280ac128003238d7d80bf07fa556a7d where the rte_service API was > > introduced as experimental by Nikhil Rao, and then later used in the eal library > > as part of commit a894d4815f79b0d76527d9c42b23327de1501711 by Harry van Haaren. > > Its no big deal because, as we agree, internal usage should be considered safe, > > but it seemed clear that differing authors were using each others code > > (potentially oblivious to the experimental nature of those APIs) > > > > > But OK, I guess it won't hurt to have more granular approach. > > > > > > > > > > >> Do we have any option than allowing them in DPDK library? And when experimental > > > >> API modified the users in the DPDK library internally guaranteed to be updated. > > > >> Why not globally allow this for all DPDK internally? > > > >> > > > > For the reason I gave above. We certainly could enable this in a more top-level > > > > makefile so that for in-library systems it was opt-in rather than opt-out, but I > > > > chose a more granular approach because I could envision newer libraries wanting > > > > it on. I also felt, generally speaking, that where warning flags were > > > > concerned, it generally desireous to have them on by default, and make people > > > > explicitly choose to turn them off. > > I think DPDK developpers look at the EXPERIMENTAL warning in the doxygen > above the prototype. I'm not sure I agree with that, but regardless, my initial reasoning for writing this tag was to call attention to experimental API's during review, rather than their use during development, so I didn't gripe about ABI changes on expemted code. Additionally, weather they look at the docs or not, they can pre-emptively turn off the warning if they choose. > And when API will be switched to stable, we probably won't remove the flag > in the makefile to disable allowing experimental. Well, that remains to be seen I suppose. > So at the end, we could just allow experimental API for all internal libs. Thats a rather bootstrapping argument. You are effecitvely saying that no one developing will ever want to be warned of using experimental APIs in DPDK, so lets just turn it off everyone. I would really rather let individual developers make that call at the time they author something new. >