Phew. Let's see... Jeronimo Pellegrini <[EMAIL PROTECTED]> writes: JP> So... Even though I set MODULE_LOC correctly, it is unset by the JP> script. heodd thing is that KSRC is not changed. The only part of the JP> script that seems to change these variables is in the tpo of the file: JP> JP> KSRC:=/usr/src/linux JP> MODULE_LOC:=/usr/src/modules
Can you try locally changing the second line to 'MODULE_LOC?=/usr/src/modules'? If that works, I can make the change in the package, too. JP> gcc -M -MG -I. -Ikernel/include -I/home/jeronimo/Linux/kernel/modules/i2c \ JP> [...] JP> make[2]: *** No rule to make target `linux/i2c-proc.h', needed by `kernel/chips/via686a.d'. Stop. JP> make[2]: Leaving directory `/home/jeronimo/Linux/kernel/modules/lm-sensors' JP> make[1]: *** [build-modules] Error 2 JP> make[1]: Leaving directory `/home/jeronimo/Linux/kernel/modules/lm-sensors' JP> Module /home/jeronimo/Linux/kernel/modules/lm-sensors failed. JP> JP> JP> Because it is loking for headers in the i2c dir, and not in i2c/linux, JP> and I'd expect (that's where the headers are). No, that doesn't seem right; it's looking for <linux/i2c-proc.h> under $(MODULE_LOC)/i2c, so if you concatenate these two pieces in the correct order, you should find the header file. It looks like what's actually happening here is that the dependency-detecting code[1] is losing. Does running 'make clean' before trying the build fix this? (Should probably add a comment to README.Debian about this, since it seems a not-too-infrequent failure mode.) (And I checked, the .d files aren't included in the source tarball.) JP> What's going on? I feel like I missed something really basic and I JP> shouldn't need to change the rules script, but things don't seem to JP> work here... Errm, yeah, it's pretty clear that there's a bug with not-in-the-default-place module building. :-( [1] As much as it's a pain to put all of your header file dependencies in both your source file *and* the Makefile, as far as I can tell, this sort of automated dependency checking causes insane amounts of pain when it doesn't realize that its assumptions about the world are completely wrong, because the source is being built on a different machine or it guessed wrong the first or something like that. *sigh* -- David Maze [EMAIL PROTECTED] http://people.debian.org/~dmaze/ "Theoretical politics is interesting. Politicking should be illegal." -- Abra Mitchell
