From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: <512222AE.9050308@xenomai.org> Date: Mon, 18 Feb 2013 13:46:38 +0100 From: Gilles Chanteperdrix MIME-Version: 1.0 References: <511C911B.2080609@zultron.com> <511CD558.2000409@xenomai.org> <511D2777.107@zultron.com> <511D3D04.7080902@xenomai.org> <511EACB8.30000@zultron.com> In-Reply-To: <511EACB8.30000@zultron.com> Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Subject: Re: [Xenomai] Announcing beta x86 Xenomai kernel packages: squeeze, precise, lucid List-Id: Discussions about the Xenomai project List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: John Morris Cc: emc-developers@lists.sourceforge.net, Xenomai On 02/15/2013 10:46 PM, John Morris wrote: > On 02/14/2013 01:37 PM, Gilles Chanteperdrix wrote: >> On 02/14/2013 07:05 PM, John Morris wrote: >>> On 02/14/2013 06:15 AM, Gilles Chanteperdrix wrote: >>> I'm hoping that someone else will volunteer to take over package >>> maintenance once the dust settles a little. >> >> >> To be completely frank, I hope we find a volunteer for the debian >> packages too... So, actually, we could add such an announce to the wiki. > > There's a request for volunteer package maintainers on the LinuxCNC > wiki, but probably not in a prominent enough position. > > Once the packages are tested and working, I'll be around to help fix > problems. However I'm quite eager to pass basic maintenance along to > someone else, so I'm trying to make maintenance as easy as possible to > help encourage volunteers. > > When a new Xenomai is released, the steps to update the kernel packages > are simple (I might forget something small): > > - Rebuild xenomai package; just a few minor tweaks > - Unpack new vanilla kernel and check out /debian directory from git > - Fix version numbers in new debian/changelog entry > - Update the vanilla kernel .config, either: > - pull a config verbatim from a similar Debian/Ubuntu kernel package > - or else update the existing config by running 'make fooconfig' > - Rebuild. Xeno patch is grabbed from /usr/src, and Xeno config is > automatically overlaid upon the vanilla config. > > I'm also working on a Makefile that will automatically assemble all the > ingredients and build the packages in a chroot (using > pbuilder/debootstrap) for all combos of squeeze/precise/lucid and > i386/amd64. Whether the next maintainer uses it as is, or sets up a > buildbot or buildd, it should make the job trivial. The thing that worries me is that I have to test the debian packages as part of each release, which means booting on the kernel, and running xeno-regiression-test. But you are right: I should automate building and testing the debian packages. I have everything needed at hand to do it. >>>> We will also try and address the issues which were identified since you >>>> started working on this, namely: >>>> - allowing enabling the SMI workaround without kernel re-compilation >>> >>> Great! I'm working around this for now by packaging Jan's smictrl >>> utility. That could actually be sufficient; I'd rather put effort into >>> a more user-friendly wrapper, and maybe a config file and init script >>> for persistent settings across reboots. >> >> I guess I have some little piece of code somewhere to handle >> configuration files with the "windows ini" style, if you are interested. > > I'm interested. It may sit on the back burner for a little, but if it's > easy for you to find, send it over and I'll certainly borrow from it > when the time comes. http://sisyphus.hd.free.fr/~gilles/conf.tar.bz2 conf.[ch] is the "configuration parser" bm_conf.[ch] is an example. I thought I had another version with the windows ini "[sections]", but I do not find it, you only have "key = value", but it may be sufficient. > >> The kernel-based approach has a big advantage though: recent kernels >> have a driver for the LPC part of the Intel ICH, which is where the >> register controlling SMI is found (drivers/mfd/lpc_ich.c), so, basing >> our work on this driver in the I-pipe kernel, we could share the table >> and avoid the double maintenance. > > Is the advantage that the lpc_ich driver has more control than Jan's > smictrl? Anyway, I don't expect to get to the config/init script for a > bit unless there's a lot of demand for it, so for now maybe I can be > lazy here and wait for you to make the first move. ;) The advantage of the lpc_ich is that it has an exhaustive list of all the Intel LPCs product IDs, which has always been the weak point of Xenomai's SMI workaround module. -- Gilles.