glibc make check error (LFS 5.1-pre1)

jeremy at jutley.org jeremy at jutley.org
Thu Mar 25 17:44:17 PST 2004


> On 2004.03.25 20:11, Ken Moffat wrote:
>> Only that you maybe aren't on i686, or else you aren't on the right
>> kind of i686 ;-)  I'm only testing on i586 and ppc, but I think
>> glibc-2.3.3 is still somewhat suspect, certainly in the December
>> versions.  Whether failing the test matters is a different question.
>>
>> Among other things, we lack data points.  I've seen a few people say
>> 5.1-pre1 was good for them, and one or two with severe failures, but
>> not much discussion of test results.  Certainly I've seen nothing
>> mentioning this particular test.
>>
>>  My only observation is that you say you cleaned out the build
>> directory and the source is unchanged.  I hope you mean you also
>> cleaned out the glibc-2.3.3 directory and re-extracted the same
>> source ?  Just because we use a build directory doesn't necessarily
>> mean the source tree is unaffected by compilation (or maybe I'm being
>> pessimistic).
> I'm on an AMD Athlon XP-2000+, so maybe you have the right idea about
> the wrong processor.

Which glibc are you using - the date-tagged one that's currently listed in
5.1-pre, or the new -rc1 one that Ryan put together.  At any rate, using a
roughly similar machine to yours (Athlon XP 2500+), I have no test
failures in either one of those glibc tarballs.  See what happens when you
start with a clean extract of the tar.bz2 file.

>
> Also - I did use the same source directory as for the previous build.
> I suppose I can try again with a clean extract from the tar.  I suppose
> it is sometimes a good thing to be pessimistic.  For now, I'll wait to
> hear if anyone has any more ideas about how important this is.
> Can anoyone say whether this type of error might end up changing which
> glibc is used in the book before 5.1 is formally released?

It could, if we can determine it to be a problem in the glibc tarballs
instead of "user error" - but with so many of us building on i686
successfully, it seems questionable about whether it's a glibc problem.

-J-



More information about the lfs-support mailing list