> IMHO it could be stressed a bit more that all three (running kernel,
> kernel source and lm_sensor) should be the same to avoid trouble. I
> somehowhad a different source than the running kernel. Well, more idot
> proof thing will created more creative idiot ;-)
>From i2c/QUICKSTART and lm_sensors2/QUICKSTART:
* Make sure you have the kernel tree corresponding to your current
kernel in /lib/modules/(uname -r)/build.
Even if the source is there, it may not correspond to your running
kernel. If so, you will have to recompile your kernel.
>From lm_sensors2/INSTALL:
This package REQUIRES i2c-2.8.1 or later!!!
So it's already in there.
> > > - sensors-detect produced a line # no driver for Winbond W83L785TS
> > > yet, ask us for one! and so I do ;-)
> >
> > OK, will write a driver if I can find some spare time.
>
> Great! Thank you.
BTW, could you please provide a dump of your chip? The data sheet isn't
always clear about how things should be handled, so a sample dump would
help:
i2cdump 0 0x2e (or i2cdump 1 0x2e if it's on the second bus).
> > BTW, your chipset isn't really an AS99127F, but an ASB100. We now
> > have a separate driver for it in CVS, you should give it a try. The
> > i2c patch you used is OK, you just need to download and install
> > lm_sensors2 CVS: http://secure.netroedge.com/~lm78/download.html#cvs
> > We still need to update sensors-detect to reflect that, so don't
> > trust it when it says you should use the w83781d driver, and use
> > asb100 instead.
>
> This seems to be already done.
Yes, this was done right before releasing 2.8.2.
> One more note on making lm_sensors more "package-friendly". You could
> add -rpath $PREFIX/lib (or -R, depending on the linker) to make sure
> that libsensors is always found by the package binaries. For testing,
> I use prefix=/usr/local/lm_sensors, so this would help me.
I'm not sure I understand what you need. Could you please provide a
patch against CVS that does what you want?
> Also a make uninstall target (of course, automatically in sync with
> make install, in the worst case via a generated file list) would IMHO
> encourage people to try things.
I agree. That might not be easy to do though, since our Makefile spreads
over many directories. I don't think I'll have the time to take care
about that. Maybe MDS will, or you can work it out yourself and submit a
patch against CVS. Idealy, patches would be needed for both i2c and
lm_sensors2 (i2c being easier, of course).
--
Jean Delvare
http://www.ensicaen.ismra.fr/~delvare/
