Ken Moffat wrote:
I've started on a build, with current LFS plus a few variations (bison-3.0.4, libtool-2.4.5, patch-2.7.3, linux-3.19-rc6 headers) and I've also changed my build to untar and compile in a tmpfs - does not seem to make any significant difference to the build times I can compare, but perhaps the untarring and rm -rf is faster.
I created a ram disk for all of /mnt/lfs and found the speedup time negligible. I didn't run tests though. The test was on Jan 12 with all packages current as of then.
Anyway, I mention the tmpfs because attr's testsuite failed - you can't have extended attributes in tmpfs - and it is possible that the new gcc failures are related to this, or, of course, to the newer kernel headers. I'm seeing errors, beyond those I expected to see, in the testsuites for: gcc: 125 unexpected failures in g++, 658 in gcc, 22 in libstdc++ instead of the usual handful of failures. The last time I saw those sorts of numbers was a little while before my AmigaOne expired. perl: unexpected error number in getnameinfo.t, invalid argument in io_multihomed.t and io_sock.t : I think I've seen something like that in the last few months, but I thought it had gone away (didn't run any tests on real hardware for my previous build, and the logs from the i686 VM tests are in qemu images).
I've not seen significant differences between VM and real HW. -- Bruce -- http://lists.linuxfromscratch.org/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
