From: Paul H. Hargrove (PHHargrove_at_lbl_dot_gov)
Date: Mon Sep 10 2007 - 13:56:35 PDT
Neal Becker wrote: > Here is a (first try) at patching blcr-0.6 to use dkms: > Neal, Thanks you the patch. I hope to be able to integrate DKMS support into the BLCR specfile for 0.6.x, for some small 'x'. However, as I mentioned the last time we discussed this, I'd like to continue building the "normal" binary RPM of kernel modules for those users not using DKMS. It is certainly possible for me to try to do this based on your patch, but I'd not (currently) be able to test the result. So, if you could, please resubmit the patch suitably modified to build both a DKMS and "normal" modules rpm, I'd appreciate it. Also as we agreed by e-mail once before, you almost certainly need a line like 'CLEAN="[ -e Makefile ] && make distclean"' to ensure builds after the first will start with a clean slate. However, I think we concluded that something in my Makefiles kept that from working. If that is still the case, I'd suggest something like the following: MAKE[0]="rm -rf BLD && mkdir BLD && cd BLD && ../configure --with-components=modules && make" CLEAN="rm -rf BLD" BUILT_MODULE_LOCATION[0]=BLD/cr_module/kbuild BUILT_MODULE_LOCATION[1]=BLD/blcr_imports/kbuild BUILT_MODULE_LOCATION[2]=BLD/vmadump4/kbuild Finaly, in our past emails on the subject I think we agreed on the need for "--with-linux=$kernel_source_dir" (or similar) passed to configure to support the -k option to dkms. -Paul -- Paul H. Hargrove PHHargrove_at_lbl_dot_gov Future Technologies Group HPC Research Department Tel: +1-510-495-2352 Lawrence Berkeley National Laboratory Fax: +1-510-486-6900