Replying to myself... On Mon, 26 Jan 2009 15:57:10 +0100, Jean Delvare wrote: > On Mon, 26 Jan 2009 22:47:36 +1300, Volker Kuhlmann wrote: > > A wrapper for the library, which outputs device names to /sys/class/... > > pairs one per line in a format easily parsable by a shell script? > > Either that, or with a command line interface so that you can ask > directly for what you want. This will save some parsing in > user-space, and also avoids computing things that the caller will > just discard. BTW, I forgot to mention: this won't solve the problem completely, as the i2c bus numbers are still not stable. libsensors only addresses this issue during configuration file parsing, where bus number references can be bound to actual buses. But these statements have no effect on the device names as returned by libsensors to applications. This was the point of ticket #2238: http://www.lm-sensors.org/ticket/2238 Unfortunately, the implementation I have in mind for ticket #2174 is incompatible with the fact of letting bus statements have any effect beyond the configuration file they come from: http://www.lm-sensors.org/ticket/2174#comment:6 And I consider #2174 much more important to have than #2238, so the latter will most certainly be closed as WONTFIX. I tend to think that i2c bus number stability is better achieved at kernel level anyway, maybe using the same approach used for multimedia devices: a module parameter can be passed that specifies the desired device number. -- Jean Delvare