Question About Unstable "Host system requirements"

Ken Moffat ken at kenmoffat.uklinux.net
Tue Jun 15 04:25:25 PDT 2004


On Tue, 15 Jun 2004, Randy McMurchy wrote:

> On Tue, June 15, 2004 at 00:53 -0500, DJ Lucas wrote:
>
> > Randy McMurchy wrote:
> >
> > > I suggest in the "Host requirements page" to mention
> > > something about glibc's failure of the NPTL tests if your
> > > kernel is compiled with 2.95.3. As Zack mentioned, it's
> > > probably not prudent to say it's OK and not to worry about
> > > it, but it needs to be mentioned.
> > >
> >
> > Already in the works sorta.  Take a look at bz851.
>
> "Sorta" is a stretch. There's no mention of this problem in the
> bug, or on the Wiki page (at least not that I saw). In fact,
> the Wike references the two documents I used when I upgraded to
> 2.6.4. Both of these docs have instructions to use gcc-2.95.3.

eugghh.  Another example of outside-the-book documents falling behind
(or being used for more than was intended - building 2.6 kernels with
gcc-2.95.3 is probably fine for most things except nptl).

>
> Perhaps a mention on the glibc page in Chapter 5 where there's
> known issues on test failures (As an aside, I don't understand
> explaining the test failures in Chapter 5, they really belong
> in Chapter 6. The book is fairly adamant about skipping the
> tests in Chapter 5, why explain the tests here?).
>

The 'pure lfs' method started out by running all of the tests.
Eventually the non-toolchain tests in chapter 5 were discouraged because
of unexplained problems on certain hosts / distros (tests failed in
chapter 5, but chapter 6 was ok).  Anybody who cares about the book
should prolly be running the chapter 5 tests, at least to look for new
failures.  But, if the book suggests you should skip the toolchain tests
in chapter 5 (I say 'if' because I'm not following CVS at the moment)
then the book is _wrong_.

> Anyway, it's something that could be fixed in 5 minutes and
> save folks lots of time. Why not put it in the book right now?
> There's really no reason not to. A one line command to discover
> the compiler used and a couple of lines explaining the problems
> one will encounter if gcc-2.95.3 is used.
>
> Seems like a no-brainer to me.
>

Agreed.  As you've shown us, "ignore the tests if you used 2.95.3" isn't
always a viable answer.

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




More information about the lfs-dev mailing list