[an almost identical e-mail as the one I've sent to i2c; you seem to kind of share docs and Makefile structure...] Ugh, I hate it when different pieces of documentation say things that are completely contrary to each other... :( The kernel README and IIRC also the glibc docs specifically say that /usr/src/linux _shouldn't_ contain the header files of the current kernel, but of the kernel that was installed at the moment the (glibc) library was compiled... Or in other words: "Don't fsck with /usr/src/linux unless you're installing a new glibc" Now the docs of lm_sensors mention "/usr/src/linux" in the role of "current kernel sources" multiple times, which, as alredy said, _isn't_ correct according to the README of the kernel itself... :( Some small comments about the Makefile: would the makefile language allow for a kind of $(shell if [ ! -d $(LINUX_HEADERS) ]; \ then <abort the Makefile here>; fi) type of construct to be inserted to do an early abort of the Makefile if someone is silly enough to run the Makefile while /lib/modules/$(KERNELVERSION)/build is a dangling symlink, so the 'grep's, 'cut's etc. don't go horribly wrong for lack of input(files) in such a case?.. Fortunately $(shell if [ -L also sees _dangling_ symlinks as symlinks, so there's no disastrous fall-back to /usr/src/linux if /lib/modules/$(KERNELVERSION)/build is dangling... (my $(KERNELVERSION) is 2.4.20 , /usr/src/linux/include/ contains 2.4.19 headers (see rant above)) Marco van Loon
