From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jean Delvare Date: Mon, 12 Nov 2007 13:32:55 +0000 Subject: Re: [lm-sensors] 2.10.5 on 2.4 kernel requires sysfs/libsysfs.h? Message-Id: <20071112143255.4fc3dad2@hyperion.delvare> List-Id: References: <20071110095638.GA29181@puariko.nirvana> In-Reply-To: <20071110095638.GA29181@puariko.nirvana> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: lm-sensors@vger.kernel.org Hi Axel, On Mon, 12 Nov 2007 01:05:17 +0200, Axel Thimm wrote: > On Sun, Nov 11, 2007 at 11:25:39PM +0100, Jean Delvare wrote: > > On Sat, 10 Nov 2007 11:56:38 +0200, Axel Thimm wrote: > > > up to 2.10.4 sysfsutils was only required for 2.6 kernels. Now the > > > build fails on 2.4 kernel system w/o sysfsutils with > > > > > > make: *** No rule to make target `sysfs/libsysfs.h', needed by `lib/sysfs.ad'. > > > > > > Is that a bug in the Makefiles or is sysfsutils now really required > > > for 2.4. kernels as well? > > > > > > (I hit this while building rpms for RHEL3). > > > > No, sysfsutils is still not required for 2.4 kernels. > > > > The only change in 2.10.5 that could cause this problem is this one: > > http://www.lm-sensors.org/changeset/4738 > > I see, I will probably need to divert KERNELVERSION. > > > But I tested this with a 2.4 kernel and it works as intended. > > > > Hmmm, maybe I see what's going on. Are you cross-building the RHEL3 > > package, on a system itself running a 2.6 kernel? > > Yes. > > > And how are you pointing the build system to the right kernel tree > > to build against in this case? > > Something along > make LINUX=/srv/atrpms.net/atrpms/kernelsrc/el3-x86_64/2.4.21-52.EL-x86_64 I2C_HEADERS=/usr/include/i2c COMPILE_KERNEL=0 > > I'll add an KERNELVERSION=2.4.21-52.EL for good measure ;) Yes, that should solve the problem. I admit that my fix introduced a regression, as it was previously possible to only set LINUX and you now have to set both LINUX and KERNELVERSION. However we have a chicken and egg problem, because the "UTS release" number, which in practice is the kernel version number, is defined in a different header depending on the kernel version. But well, I'm not exactly sure what my change is fixing, so I might as well revert it, and wait for a real bug report before I change it again. -- Jean Delvare _______________________________________________ lm-sensors mailing list lm-sensors@lm-sensors.org http://lists.lm-sensors.org/mailman/listinfo/lm-sensors