[lfs-support] Chapter 6.9 errors

Ken Moffat zarniwhoop at ntlworld.com
Fri Jun 29 16:46:01 PDT 2012


On Fri, Jun 29, 2012 at 11:12:22PM +0100, spiky wrote:
> The host is there beta of next release, 32bit , kernel 3.2.0-26-generic-pae
> I installed the 3.5rc4 in previous chapter 6.7
> I was in 2 minds weather to carry on, or rebuild an earlier kernel ( if 
> thats a possible fault) Or rebuild with Debain host using 3.5rc4 kernel
> 
 I think you should carry on.  A few people seem to get perfect
results in the toolchain tests for years on end, the rest of us
often get a few minor failures from time to time.  This error almost
certainly does not matter.

 I've been thinking about suggesting that we change the language in
the book re toolchain tests, but haven't got a round tuit. It was
valid and useful when we started running the tests, but I think it
is a little too 'black and white' for the variety of results we've
seen over the years - and anyway, the tests can only attempt to check
known failure cases and may themselves be buggy.

 If you are still unconvinced, take a look at what that test does,
if you are able to work that out [ some testsuites are almost
impossible to understand, I haven't found it necessary to look at
the glibc tests for ages and can't remember whether they are
straightforward or obscure ].

 We have seen batches of errors in glibc tests in the past,
particularly on ubuntu!  But AFAIK nobody ever found an error in the
behaviour of the resulting LFS system which was related to those
failures.

 If you are *really* concerned about this, with CPU cycles and
partitions to spare, build your new LFS, install a 3.5-rc4 kernel
(since you are interested enough to use the -rc headers) and
whatever BLFS/beyond packages you need to make it a comfortable
host, then use it as the host for a fresh build (as far as the glibc
tests) to see if the results differ.  You might even stop before
installing that glibc from the second build, then repeat the glibc
build and tests using a 3.2.21 kernel to see if the problem is
ubuntu-specific, or perhaps related to 3.5 headers on older kernels.

 OTOH, although that might be interesting I doubt that many of us -
obsessive though we usually are on matters linux - will be
particularly concerned about the results.  For the moment, we only
know that one ubuntu 32-bit host had a new failure.  Even if
everyone gets the same single failure with 3.5.0 headers, it won't
prevent the book being upgraded.  For me, 3.5-rc4 is a good kernel
(unlike -rc{2,3} : thanks, Trond!), but a lot could still change
before 3.5.0 is released.  I often try to build at least one system
using a *late* kernel -rc, if time permits, but I think there will
probably be some more releases in the 3.5 cycle.  That's just
guesswork, but I haven't seen comments that it is nearing a release.

 Your system (and time), your decision ;)

ĸen
-- 
das eine Mal als Tragödie, das andere Mal als Farce



More information about the lfs-support mailing list