From Bugzilla Helper: User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1) Description of problem: At least Two problems currently prevent the Fedora core 1 kernel from building. Hardware platform is PII in HP Omnibook Xe2DB. I am using standard config for i686 from the configs directory having upgraded kernel source from up2date to 2129 version. 1) kernel/sched.c fails to compile reporting 'undefined reference to active_load_balance'. This is easily fixed by editing the code to move the offending line ~217 outside the #ifdef, but seems to be a bit basic 2) Later a report of 'undefined reference to init_rootfs' is reported in function mnt_init I realise I'm reporting two issues here, but there is a lot of discussion of this in a number of forums including some old references to similar problems on redhat versions. In several cases these were rejected as 'not a bug'. They definately are bugs, and I don't think I'm the only person experiencing them, so they should be either re-logged or re-opened. Version-Release number of selected component (if applicable): Kernel linux-2.4 ... 2129 nptl How reproducible: Always Steps to Reproduce: 1. Build kernel using the regular i686 config file 2. 3. Actual Results: Errors as described above Expected Results: Kernel should build Additional info:
did you start with make mrproper ?
Sequence to build was make clean make mrproper copy standard config file file from configs directory to .config make xconfig (here I make no changes just save and exit - also have repeated whole sequence using 'make config' and 'make menuconfig') make dep make bzImage
make xconfig is broken; you need to do a make oldconfig after it...
Ok have run make oldconfig after make xconfig and the build now works ok. Have a working kernel, although interestingly its not __exactly__ the same (size) as the distributed one. Are you aware whether the distributed kernel is made from the distributed configs or from some other config file ? ps. I used kernel-2.4.22-i686.config as the base for my build. Suggest if you have a mechanism to make the 'missing' oldconfig step more widely known you use it, it will save a lot of 'bad mouthing' of Fedora thats currently flying around on the web. Thanks for the help
we use that exact config (as you can see in the spec file). The size difference can be due to the fact that the kernel builds the hostname and date/time of build into itself (see the first line of dmesg) and is then compressed, obviously different strings yield slightly different compression results..
Ok, I think that's it for now. Thanks again