Last night I built shared-mime-info-1.5 on my Kaveri. All previous builds there had used earlier versions. Trying to build using make -j5 -O it failed :
config.status: executing po/stamp-it commands GEN create-pot Making check in . make[1]: 'shared-mime-info.pot' is up to date. make[1]: Entering directory '/scratch/working/shared-mime-info-1.5' CC test_subclassing-test-subclassing.o make[1]: Leaving directory '/scratch/working/shared-mime-info-1.5' make[1]: Entering directory '/scratch/working/shared-mime-info-1.5' CC test_tree_magic-test-tree-magic.o make[1]: Leaving directory '/scratch/working/shared-mime-info-1.5' CC update_mime_database-update-mime-database.o make[1]: Entering directory '/scratch/working/shared-mime-info-1.5' CC update_mime_database-update-mime-database.o mv: cannot stat '.deps/update_mime_database-update-mime-database.Tpo': No such file or directory And it repeatedly fails like that on the Kaveri, but is ok with -j1. Oddly, I have built it at least three times on the haswell, all using make -j8 -O. For the moment, I'm merely remarking this and switching to -j1 for all *my* future builds (I cannot be bothered to do test runs on the other two boxen when I eventually get to them, the time lost by building with -j1 will only be a few seconds). And since -j8 works on my haswell I don't - for the moment - feel inclined to try to change the book. But I never expected that a "middling" processor (top-end AMD APU) would highlight something like this. Why do I build with -j5 instead of -j4 there ? I begin to wonder ;-) Actually I tried -j5 (cores+1) when I got an SSD, to make sure it was fully loaded. Building the current kernel there with -j4 took six minutes, so I figured that any gains offered by -j5 should be grabbed with both hands. I'm sure the kernel build on the Kaveri, albeit without all the RAID options, was *much* faster back in November. ĸen -- This email was written using 100% recycled letters. -- http://lists.linuxfromscratch.org/listinfo/blfs-dev FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
