From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Thu, 22 Jan 2015 15:26:02 +0100 From: Gilles Chanteperdrix Message-ID: <20150122142602.GE26731@hermes.click-hack.org> References: <1537147.Wm9AbiuW5D@soho> <2296691.2JpS0rAJbA@soho> <20150122114038.GE23461@hermes.click-hack.org> <4743020.6q5RsBy8Ia@soho> <54C0F262.6010806@siemens.com> <20150122134400.GA26731@hermes.click-hack.org> <54C10043.2060608@siemens.com> <20150122135840.GC26731@hermes.click-hack.org> <54C10791.2020609@siemens.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <54C10791.2020609@siemens.com> Subject: Re: [Xenomai] Xenomai-3 creating debian packages List-Id: Discussions about the Xenomai project List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Jan Kiszka Cc: xenomai@xenomai.org On Thu, Jan 22, 2015 at 03:22:09PM +0100, Jan Kiszka wrote: > On 2015-01-22 14:58, Gilles Chanteperdrix wrote: > > On Thu, Jan 22, 2015 at 02:50:59PM +0100, Jan Kiszka wrote: > >> On 2015-01-22 14:44, Gilles Chanteperdrix wrote: > >>> On Thu, Jan 22, 2015 at 01:51:46PM +0100, Jan Kiszka wrote: > >>>> On 2015-01-22 12:59, Leopold Palomo-Avellaneda wrote: > >>>>> El Dijous, 22 de gener de 2015, a les 12:40:38, Gilles Chanteperdrix va > >>>>> escriure: > >>>>>> On Thu, Jan 22, 2015 at 12:39:10PM +0100, Leopold Palomo-Avellaneda wrote: > >>>>>>> El Dijous, 22 de gener de 2015, a les 12:31:05, Gilles Chanteperdrix va > >>>>>>> > >>>>>>> escriure: > >>>>>>>> On Thu, Jan 22, 2015 at 12:17:15PM +0100, Leopold Palomo-Avellaneda > >>>>> wrote: > >>>>>>>>> I have renamed the subject according with the theme. > >>>>>>>>> > >>>>>>>>> El Dijous, 22 de gener de 2015, a les 11:48:13, Gilles Chanteperdrix > >>>>>>>>> va > >>>>>>>>> > >>>>>>>>> escriure: > >>>>>>>>>>>> Well, the fact that the compilation error happens when trying to > >>>>>>>>>>>> compile a Xenomai file would seem to indicate that > >>>>>>>>>>>> prepare-kernel.sh > >>>>>>>>>>>> has put the wrong link into init/Kconfig. > >>>>>>>>>>>> > >>>>>>>>>>>> You pass --arch=x86 to prepare-kernel.sh, right, not > >>>>>>>>>>>> --arch=x86_64 ? > >>>>>>>>>>> > >>>>>>>>>>> yes, I have tried both, but I put --arch=x86 or the default option > >>>>>>>>>>> as > >>>>>>>>>>> I > >>>>>>>>>>> said it in my email from yesterday ( :-P) > >>>>>>>>>> > >>>>>>>>>> Well, maybe the default option does not work, so please try > >>>>>>>>>> specifying --arch=x86. The error you have is not in the compiled > >>>>>>>>>> files, but in the Kconfig files. So, the fact that you can compile > >>>>>>>>>> does not really prove that prepare-kernel.sh did not fail. > >>>>>>>>> > >>>>>>>>> Ok, from a clean sources I have run from the linux tree sources: > >>>>>>>>> > >>>>>>>>> $ ../xenomai-3.git/scripts/prepare-kernel.sh --linux=. -- > >>>>>>>>> ipipe=../xenomai-3.git/kernel/cobalt/arch/x86/patches/ipipe-core-3.16- > >>>>>>>>> x86-1.patch --arch=x86 > >>>>>>>>> > >>>>>>>>>> Now, prepare-kernel.sh adds the scripts/Kconfig.frag file to > >>>>>>>>> > >>>>>>>>>> init/Kconfig. Kconfig.frag contains: > >>>>>>>>> I don't understand that sentence. Yes, the Kconfig.frag now, contains > >>>>>>>>> this: > >>>>>>>>> > >>>>>>>>> source "arch/$ARCH/xenomai/Kconfig" > >>>>>>>> > >>>>>>>> Ah OK. Could you try to revert that commit ? > >>>>>>>> > >>>>>>>> http://git.xenomai.org/xenomai-3.git/commit/scripts/Kconfig.frag?h=next& > >>>>>>>> id=2 6cc64313e2e695ac227e5a02992c5fe72df2e5f > >>>>>>> > >>>>>>> Gilles, I did it. I explained it before. As I just run the script, I don't > >>>>>>> run any autoconf, so, I then have this part as @LINUX_ARCH@ so it doesn't > >>>>>>> work. > >>>>>>> > >>>>>>> prepare-kernel script doesn't touch (substitute) that. > >>>>>> > >>>>>> prepare-kernel.sh before that commit substitutes @LINUX_ARCH@. You > >>>>>> just need to revert the whole commit. > >>>>> > >>>>> Ok, the whole commit touch also the prepare-kernel script. I have tested and > >>>>> now, I got in linux-tree/init/Kconfig: > >>>>> > >>>>> if XENOMAI > >>>>> source "arch/x86/xenomai/Kconfig" > >>>>> endif > >>>>> > >>>>> so, I'm sorry Jan, but I think that at least in debian platforms it wasn't a > >>>>> good change. > >>>> > >>>> See > >>>> http://git.xenomai.org/xenomai-jki.git/commit/?h=for-forge&id=5d104b1de84c8f4cbfb9d6b79cfaa633ef550e1c > >>>> for the fix. The bug was the improper variable use in the Kconfig > >>>> fragment. You only had to revert the whole commit as you were falling > >>>> back to the obsolete prepare-time patching then, which involved the script. > >>> > >>> I do not see what is so obsolete about it. As long as patches are > >>> provided separately for each architecture, the prepare-kernel script > >>> has to know the architecture anyway. > >> > >> The reason is given in the original patch: it allows merging all patches > >> and preparation into a single source tree. There is no technical need to > >> patch the arch in anymore. > > > > Providing separate patches for architectures has a huge advantage: > > if you need to release patch because one bug was fixed for one > > architecture, you only need to release the patch for this > > architecture. Besides, most Xenomai users (as opposed to developers) > > use one architecture at a time. > > Misunderstanding: This is about integrating the separate patches and > Xenomai fragments over a single source tree for local build purposes, > not about obsoleting per-arch I-pipe patches. It's a small added feature > to the Xenomai kernel preparation process, no change of the maintenance > approach. Then again, I think calling the method we used until now "obsolete" is a poor choice of word. Since it is the method that Xenomai use and will continue to use, there is nothing obsolete about it. -- Gilles.