This happened twice tonight that I know of. Did something change in the 2.4.17 kernel sources? <kainz> 00:47 < PreZ> compiling a kernel, ksyms.c bombs out inside a .h file (kernel_stat.h). Its failing to find the symbol smp_num_cpus .... <kainz> now kernel_stat.h includes linux/smp.h, which has in it, IF its an smp box, then have smp_num_cpus as an extern int. <kainz> Otherwise #define smp_num_cpus 1 <kainz> 00:47 < PreZ> by rights the preprocessor should have optimized that out, and it should have succeeded <kainz> 00:48 < PreZ> but I ended up having to modify kernel_stat.h to check for smp support, and change its behavior accordingly ... bizarre <kainz> 00:51 < PreZ> seem -- scroll back, see if you can explain the behavior I just saw
OK, so it seems that people aren't making mrproper. For a lot of people, that's standard procedure during kernel compiles, but people read the install docs to the letter, and somehow ignore that step. Doing that, however, makes this problem go away...
yes, found it and fixed it a day ago. the ebuilds now do make mrproper automatically in pkg_postinst() for safety.