the trouble is that 'make install' on the i2c side puts the headers in /usr/local/include... Especially now that we are requiring an upgrade to i2c-2.7.0, if we don't point to the same place on i2c and lm_sensors, then people couldn't compile i2c outside the kernel, they would have to mkpatch... Philip Edelbrock wrote: > > > I would recommend that I2C_HEADERS be set to LINUX_HEADERS by default. > (basicly just swap the comment hash between the two lines in the > Makefile.) I think it is odd to have it defaulted to /usr/local/include > since that isn't the normal or usual place for it any more. Now, > kernels have for some time had the i2c headers in them, so if you define > the include search paths as including both kernel headers and > /usr/local/include, there is a significant risk of having two seperate > sets of i2c headers of different versions being referenced > similtaniously. Right? > > We could get fancy by checking if the kernel's i2c is compiled and used, > and make that set of i2c headers the default if so? Probably not > needed, though. > > > Phil > > Mark Studebaker wrote: > >> I just found and fixed a couple of things but nothing major. >> Khali did you have anything you wanted to test? >> >> Also, Phil & Khali, any proposal on I2C_HEADERS? >> >> Philip Edelbrock wrote: >> >>> >>> >>> That's how I did it (mkpatch'ed i2c against my kernel), although my >>> sensors bin is segfaulting... ;') Keep in mind that the I2C header >>> variable in the lm_sensors makefile is hardcoded to >>> /usr/local/include, which is not great imho. >>> >>> >>> Phil >>> >>> >>> Mark Studebaker wrote: >>> >>>> did anybody test mkpatched kernels? >>>> here's one ticket... >>>> perhaps we should do some mkpatch testing before putting the >>>> announcement on freshmeat? >>>> >>>> http://www2.lm-sensors.nu/~lm78/readticket.cgi?ticket=1087 >>>> >>> >> >> >
