On 09/07/2012 01:12 AM, Tobias Gasser wrote: > Baho Utot schrieb: >> I always build all of LFS with -j4 or -j8 and it has not failed me. >> > so you just had very very much luck on your way... > > there are packages which CAN fail with parallel builds.
yes i know that, but I do not attribute my success using -j4 -j8 as luck as my builds are constent over many complete runs and four different machines two x86_64 and two i686 builds. As I am currently developing my build ayatem I start from a clean position and build/rebuild 4 to six times on each box both i686 and x86_64 before I make a decision upon it. > > whenever i run into a problem, i just restart the package with -j1 which > often solved the problem magically in the past... Yes I will drop down to -j1 if necessary. > > i'm used to set -jX to the number of cores i have (currently -j8 on my > i7) globally which speeds up the build-time significally. > > my system might be a little special, as i use a SSD and the working > diretory for the builds is a tmpfs as i have 16g of ram available. thus > disk-io is very very very fast. > > from the base LFS i have to overwrite -j8 for > make grub > make groff > make / test udev > test tar > test patch > test binutils For the above list I only had trouble with the systemd version of udev. All the udev versions used in LFS-6.8 to 7.1 worked with -j4 without error. Dropping to -j1 in the build phase fixed it. But I am investigating it > recently i had intermittent failures on both xulrunner and firefox where > i build with -jX since a very long time. with version 15.0 the build > fails at about half of the builds on my system. giving -j1 runs fine. i > can use 8 cores in the mk_add_options MOZ_MAKE_FLAGS="-j8" but NOT in > the MAKEFLAGS or the make-commandline! I just set MAKEFLAGS and if I get an error I just add make -j1 to the build script for that package. I have not built BLFS so I may be in for some trouble? -- http://linuxfromscratch.org/mailman/listinfo/lfs-support FAQ: http://www.linuxfromscratch.org/lfs/faq.html Unsubscribe: See the above information page
