Hi Jean: > On 2005-08-31, Mark M. Hoffman wrote: > > As I pointed out on #linux-sensors, there is what I consider to be a bug > > in the /sys/class/hwmon handling: if /sys/class/hwmon is not present, > > SPD/eeproms will be read normally by sensors(1); but if it *is* present, > > then SPD/eeproms will not be found or displayed. * Jean Delvare <khali at linux-fr.org> [2005-08-31 11:30:23 +0200]: > It's a feature :') > On 2005-08-31, Mark M. Hoffman wrote: > > Even if we plan to get rid of SPD/eeprom reading from sensors(1) (which > > I think is sensible), the current behavior is surprising, to say the > > least. * Jean Delvare <khali at linux-fr.org> [2005-08-31 11:30:23 +0200]: > This is true and is likely to raise an unreasonable number of support > requests. We need to anticipate that, you are plain right. > On 2005-08-31, Mark M. Hoffman wrote: > > I'm sorry I haven't had time to work up a fix - I have a four-day > > weekend coming up during which I should have some time. At any rate, > > IMO it's a release-stopper. * Jean Delvare <khali at linux-fr.org> [2005-08-31 11:30:23 +0200]: > I have been thinking of it, and came to the conclusion that it was merely > a documentation issue. I was planning to add en entry in the FAQ, a > news item on our website when we release, and a note to kernel26.html. > Don't you think it is sufficient? What do you suggest? I would like > to avoid adding temporary code to libsensors if possible, as we will > ultimately drop non-hwmon support anyway. I don't think just documenting it is enough. Here is an example scenario that better demonstrates the present "surprising" behavior: (with just i2c-i801 and eeprom modules loaded using Linux 2.6.12-rc6-mm2) # sensors (shows eeprom info) # modprobe lm78 # sensors (shows lm78 info but *not* eeprom info) I know you don't want a lot of temporary code shoved in there, and for that matter neither do I. So here is a minimal patch. It simply disables eeprom display for all 2.6.x kernels. We can rip out the guts of the eeprom code for the next release - my libsysfs patchset in progress already does some of that anyway. This should be easier to document and may also generate fewer bug reports. Index: lib/proc.c =================================================================== RCS file: /home/cvs/lm_sensors2/lib/proc.c,v retrieving revision 1.45 diff -U4 -r1.45 proc.c --- lib/proc.c 19 Aug 2005 02:59:52 -0000 1.45 +++ lib/proc.c 5 Sep 2005 17:51:22 -0000 @@ -92,8 +92,12 @@ /* We don't care about subclients */ if (len >= 10 && !strcmp(x + len - 10, " subclient")) return 0; + /* also, ignore eeproms for all 2.6.x kernels */ + if (!strcmp(x, "eeprom")) + return 0; + /* Fill in the entry fields */ entry.name.prefix = strdup(x); if (entry.name.prefix == NULL) return -SENSORS_ERR_PARSE; /* No better error :( */ Regards, -- Mark M. Hoffman mhoffman at lightlink.com
