On Saturday 02 August 2003 23:31, Mark M. Hoffman wrote: >Hi Gene: > >* Gene Heskett <gene.heskett at verizon.net> [2003-08-02 20:42:50 -0400]: >> On Saturday 02 August 2003 17:17, Jean Delvare wrote: >> >I suggest you start from the very beginning with the separate >> >compilation option. This is the way I personnally use, and the >> > one I recommend too. >> > >> >You need to have a kernel tree matching your running kernel. >> >> This has been one of the sore spots with me, this requirement that >> I make a new kernel, and sensors doesn't work because its >> referenceing the one that running, not the one thats building >> courtesy of the use of the uname facility to discover the version >> numbers. I'd much rather it reached back into /usr/src/linux to >> discover this since that link *will* be valid for the kernel under >> construction. > >Having a kernel tree that matches the running kernel is just the >easiest way for our users to get it right. However, if you want >to build against some other kernel just point the "LINUX :=" line >of the i2c and lm_sensors2 top-level makefiles to whatever kernel >you want to build against. Then continue with Jean's instructions. I see. I think. So to build against the kernel being built, I would make this line read LINUX :=/usr/src/linux since this link will be valid for the kernel being built. >> > This >> > kernel must not have any I2C element hard-compiled. If it has >> > some as modules, it's OK, they'll be overriden. >> > >> >Then, install i2c-2.8.0. Don't patch your kernel, just compile >> >separately. It's as easy as: make, make install. >> > >> >Then, install lm_sensors-2.8.0. Same as for i2c, don't patch your >> >kernel, compile separately. It's as easy as: make, make install, >> >ldconfig. I did this with 2.7.0 last night after disabling lm_sensors with chkconfig, and commenting out anything related in rc.local. Then I ran sensors-detect, Everything worked except it detected an lm75 instead of an adm1021 for the temp stuffs. As I already had an old line to bring it in with an option adm1021 line in modules.conf so it would ignore one potential address, canceling the lm75 and re-enabling the adm1021 modload in rc.local made everything work again. >Also note it's important to follow this order - don't try to cheat >by building lm_sensors before *installing* i2c. > >> ldconfig? Something tells me I really do need to rtfm better than >> I have recently... > >This just tells your system to update its list of where to find >shared libraries. There will be notes about this at the end of >the second "make install" - read them and you should be all set. Amanda is doing her nightly thing atm, but I'll try this with the 2.8.0 versions in the morning. But, I think I've got some hardware problems, and suspect they are only going to be cured by replacing my athlon cpu with a barton cored unit, this DX-1600 has been running at 170F or more, .100 volts low on its core vcc for going on 3 years now, and several of my reboots hung either at the post stage just before the memory scan, and once it spontainously rebooted when the viapro module was modprobed. Gawd I hate flaky hardware... I have a friend with a barton cored athlon DX-2400 that rarely exceeds 120F, and then only when doing mpeg2 smunching. Not to mention the northbridge fan has only run maybe 10% of the time since the mobo was new, what a POS. I have a much taller heat sink for it, fanless, but I'd need to get up close and personal to put it on, and the tower is trapped under the desk, its 1/2" taller then the steel angle that re-inforces the front edge of that home-made desk. I really need to get both of them out from under and take them to an air hose, the old 500 towers psu is pretty well plugged with dust bunnies. -- Cheers Mark, and thanks for the clue, Gene AMD K6-III at 500mhz 320M Athlon1600XP at 1400mhz 512M 99.27% setiathome rank, not too shabby for a WV hillbilly Yahoo.com attornies please note, additions to this message by Gene Heskett are: Copyright 2003 by Maurice Eugene Heskett, all rights reserved.
