From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrick McHardy Subject: Re: nftables: improve build system Date: Mon, 13 Jan 2014 09:56:46 +0000 Message-ID: <20140113095646.GC15457@macbook.localnet> References: <1389604070-12068-1-git-send-email-jengelh@inai.de> <20140113091250.GB14601@macbook.localnet> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: netfilter-devel@vger.kernel.org To: Jan Engelhardt Return-path: Received: from stinky.trash.net ([213.144.137.162]:55611 "EHLO stinky.trash.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750991AbaAMJ5E (ORCPT ); Mon, 13 Jan 2014 04:57:04 -0500 Content-Disposition: inline In-Reply-To: Sender: netfilter-devel-owner@vger.kernel.org List-ID: On Mon, Jan 13, 2014 at 10:47:24AM +0100, Jan Engelhardt wrote: >=20 > On Monday 2014-01-13 10:12, Patrick McHardy wrote: > > > >> build: remove unused checks > > > >Why aren't we instead evaluating the results? >=20 > I don't know - I did not write those lines in the first place. A > reasonable hypothesis is that this is just old boilerplate copied > from somewhere else rather than actually looking for what the > commands executed by make really need. That was not my question. I asked why we're not evaluating *instead* of removing them. > Furthermore, like a lot of NF software, it is tied to Linux where the > existence of arpa/inet.h, fcntl.h, malloc(3), and so on is > pretty much guaranteed. >=20 > The only point to check for absence of stdint.h is when at the same > time, the code provides alternative ways to get what it needs (like a > hand-crafted typedef unsigned char uint8_t) - which are not present > either. So checking for stdint.h is rather useless. >=20 > >> build: rename conflicting parser.h instances > > > >The changelog mentions something about -I but no further explanation= about > >why this change is done and the effects. >=20 > How so? Commit says: >=20 > If -I. is on the command line =E2=80=A6 >=20 > and -I. is on the command line with autoconf (more precisely, > -I${srcdir} -I${builddir} -I${path_to_config_h} is). >=20 > =E2=80=A6 #include becomes (=3Dmatches) ./parser.h > instead of ../include/parser.h. >=20 > Basically you never ever want to have a header with the same include > path. Even #include is prone to ambiguity if you had > -I/usr/include/boost (which you should not of course, but people > do all kinds of weird things sometimes), suddenly matching > /usr/include/boost/regex.h instead of /usr/include/regex.h. >=20 > It's all about distributions, and it takes a regular packager to know= =2E Jan, the changelog simply needs a description what is wrong, why it is wrong and what is changed. You can't expect people to know what parameters autoconf uses on the command line and I guess you know that. -- To unsubscribe from this list: send the line "unsubscribe netfilter-dev= el" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html