On Wed, Aug 30, 2006 at 01:14:14PM +0200, Jean Delvare wrote: > > Care to work on all of the __must_check warnings that the hwmon code now > > spits out in the -mm and my trees? It's quite ugly :( > > Mark M. Hoffman and others are already working on that: > http://lists.lm-sensors.org/pipermail/lm-sensors/2006-August/017463.html > http://lists.lm-sensors.org/pipermail/lm-sensors/2006-August/017204.html > http://lists.lm-sensors.org/pipermail/lm-sensors/2006-August/017462.html Great. > Note that we decided to abuse the fact that removing a sysfs file which > doesn't exist is not considered an error. It makes our code much more > simple. I hope there is no plan to change this behavior. Hm, that works? Ah, nice, I need to remember it, as it would make my other cleanups easier too. > I still wonder how the sysfs file creation functions did qualify for > __must_check while nothing bad actually happens if the files are not > created. OK, the files are missing, but kernel-wise this doesn't cause > any broken state as far as I can see. Why would you not check the return value of a function that provides you that information? :) All of the driver core functions that return errors were marked in this manner due to a wide range of problems that we have been having for a while with drivers ignoring problems that the driver core has been reporting. We need to handle error conditions better, and this is finally forcing it to happen. thanks, greg k-h
